![Say you use a trojan to test your sandbox. You So how can you reliably test your sandbox
notice that your machine is not compromised. without risking infection, or even worse, have
But is it because your sandbox contained the malware escape into your corporate network?
trojan, or because the trojan failed to execute
properly? It might surprise you, but there is a You can do this with shellcode. Here is an ex-
lot of unreliable malware out in the wild - mal- ample of simple shellcode that will create a file
ware that will crash more often than not, mal- in the directory of your choice (This shellcode
ware that will flat-out refuse to run in certain includes a library that is not discussed in this
environments, like virtual machines. article):
! segment .text
! ! call geteip
! ! geteip:
! ! pop ebx
! ! ; Setup environment
! ! lea esi, [KERNEL32_FUNCTIONS_TABLE-geteip+ebx]
! ! push esi
! ! lea esi, [KERNEL32_HASHES_TABLE-geteip+ebx]
! ! push esi
! ! push KERNEL32_NUMBER_OF_FUNCTIONS
! ! push KERNEL32_HASH
! ! call LookupFunctions
! ! ; CREATEFILEA and CLOSEHANDLE
! ! push 0x0
! ! push 0x80
! ! push 0x2
! ! push 0x0
! ! push 0x0
! ! push 0x0
! ! lea eax, [FILENAME-geteip+ebx]
! ! push eax
! ! call [KERNEL32_CREATEFILEA-geteip+ebx]
! ! push eax
! ! call [KERNEL32_CLOSEHANDLE-geteip+ebx]
! ! ret
Let us assume you sandboxed your preferred This method is very reliable, especially com-
browser, Firefox, and now you want to test if pared with the use of real (unreliable) mal-
Firefox is restricted from writing to the sys- ware. If you need to test access to other re-
tem32 directory. sources, like the registry, you just need to use
shellcode that writes to a particular key in the
For this, we use shellcode that creates file registry.
c:windowssystem32testfile.txt and inject this
shellcode in process firefox.exe. Example 2: Enforcing Permanent DEP
If the test file was not created in the system32, DEP is an important security feature intro-
you have successfully verified that your sand- duced with Windows XP SP3. But not all ap-
box prevents Firefox from writing to the sys- plications use DEP, so here is how you can
tem32 directory. You can also start Sysinter- enforce it. DEP can be enabled by setting a
nal's procmon and look for “access denied” flag in the executable file (the NO_EXECUTE
messages from Firefox. This is further proof flag) or by calling WIN32 API function Set-
that the shellcode tried to write to system32 ProcessDEPPolicy.
but was denied.
www.insecuremag.com 27](https://image.slidesharecdn.com/insecure-mag-33-120201200807-phpapp01/85/Insecure-mag-33-27-320.jpg)
![SetProcessDEPPolicy has one advantage ProcessDEPPolicy with argument 1 - some-
over the NO_EXECUTE flag – it can enable thing you can do with shellcode.
Permanent DEP. Once Permanent DEP has
been enabled, it cannot be disabled anymore. Shellcode to enable Permanent DEP is rather
The only way to enforce Permanent DEP is to simple: it only has to call SetProcessDEP-
make the application (like calc.exe) call Set- Policy with argument 1:
; Enable permanent DEP in current process
push PROCESS_DEP_ENABLE
call [KERNEL32_SETPROCESSDEPPOLICY-geteip+ebx]
When you inject this shellcode in your applica- pose you to turn in on again. To get rid of this
tion, Permanent DEP will be turned on. But nag screen, I developed a patch: replace byte
how can you modify your application so that it sequence
calls SetProcessDEPPolicy each time it is 50A16CBF9323FF90C805000039750859 with
launched? You can inject the shellcode per- 50A16CBF9323B8020000009039750859 in
manently in the application with a PE-file edi- file EScript.api. If you cannot change file
tor such as LordPE. EScript.api, you can still change the code di-
rectly in memory.
First you make a copy of the application (e.g.
calc.exe) and you open it with LordPE. Then I have developed shellcode to search and re-
you create a new section with the shellcode, place a sequence of bytes in the virtual mem-
and make the entrypoint point to the shell- ory of an application. This shellcode can be
code. When finished, the shellcode jumps to used to apply the Adobe Reader patch I de-
the original entrypoint. You rebuild the PE file scribed. To achieve this, you inject this shell-
and save it. code (together with the search and replace
byte sequences) in Adobe Reader.
When you execute this copy of calc.exe, your
shellcode will be the first thing to run. This Another advantage of patching dynamically in
shellcode will enable Permanent DEP, and memory with shellcode, is that the patch will
then jump to the start of the calculator pro- not be lost when you update your application
gram. to a new version (Adobe Reader in our exam-
ple).
Example 3: Patching an application
Example 4: Preventing heap sprays with
Patches are changes to the binary code of an shellcode
application. They typically fix bugs, security
vulnerabilities or change features. But when Shellcode is often used in attacks and mal-
you make changes to the files of an applica- ware together with heap sprays: the heap is
tion (.exe or .dll), you invalidate the digital sig- filled with shellcode (preceded by a long NOP
nature and you are probably breaking the sled), and then the vulnerability is triggered.
EULA. EIP jumps to the heap, hits a NOP sled and
slides to the shellcode. The shellcode exe-
If you want to change an application but are cutes, and typically downloads and installs a
not in a position to change the binary files, trojan.
shellcode designed to patch in memory can
help you. Successful heap sprays can be prevented by
pre-allocating memory, so that the heap spray
Two years ago I developed a patch to fix an cannot write shellcode to the pre-allocated
annoying “feature” of Adobe Reader 9.1. If you memory. If we pre-allocate memory and fill it
disabled JavaScript in Adobe Reader, each with our own NOP sled and shellcode, we can
time you opened a PDF document with em- intercept the attack and block it.
bedded JavaScript, Adobe Reader would re-
mind you that JavaScript is disabled and pro-
www.insecuremag.com 28](https://image.slidesharecdn.com/insecure-mag-33-120201200807-phpapp01/85/Insecure-mag-33-28-320.jpg)
![Below the identities selector is a tree of mod- spective. The progress bar will adjust in size
ules that can be selected or deselected for in- as the scanner discovers more of the applica-
clusion in the scan. Modules typically repre- tion during its recursive crawl. Vega performs
sent individual vulnerability checks and each various tests on each accessed path, trying to
module is a single Javascript file in the Vega determine if it is a file or directory. Vega also
modules directory (scripts/scanner/modules/ does 404 analysis to fingerprint the server re-
injection or response). Vega supports two sponse in cases where a path that does not
types of modules: "basic" (active) and "re- exist is accessed. As Vega identifies vulner-
sponse processing" (passive). abilities, the summary table in the "Scan Info"
central view will be populated and correspond-
The basic modules, which are also known as ing alerts will be added to the Alerts view.
injection modules, run on each injection point When an alert is selected for review from
identified by the crawler: all files, directories, within the Alerts view, it will be rendered in the
and parameters. The basic modules do the Scan Info central view. The alert contents will
fuzzing: they generate multiple new requests be described in more detail below.
and process the responses using a callback Clicking on the top level node of the scan in-
function that they register. The response stance in the Alerts view will switch the con-
processing modules run on all HTTP re- tents of the central view back to the scan
sponses received by Vega, grepping for pat- summary.
terns corresponding to security vulnerabilities.
Both modules can generate alerts and store/ Scan alerts
retrieve data in an internal database. Adding a
module to the list is as easy as dropping a file Vega modules generate alerts when they de-
in the right directory. Modules can also be ed- tect possible vulnerabilities. It is up to the
ited and reloaded without restarting Vega. module developer to decide when and why to
generate an alert, and which alert should be
To continue the setup of a new scan, the user generated. The alerts are generated when the
can click "Next" to continue to the second module invokes a specific method in the mod-
step, or skip it. For the purpose of this tutorial ule's context, specifying the XML template to
we will proceed by clicking "Next". The second use for the alert. It also passes parameters
step allows for the user to input a custom such as a HTTP request and response, some
cookie value, as well as any paths that they relevant content, a link to the vulnerable re-
do not wish the crawler to access. This is use- source on the target server, and a unique key
ful if there are logout links in the application for the alert to prevent duplicate instances.
that will clear the authenticated session if they Vega assembles the final alert using static
are accessed. Clicking "finish" will start the content from the XML template file and dy-
scan. namic content from the module.
Once the scan has started, the progress is in- Example - Module "vinfo-1918.js" invoking an
dicated in the central view of the scanner per- alert (the first parameter is the XML template):
[..]
ctx.alert("vinfo-1918", request, response, { // XML file, request
object, response object
output: result.join(" "), // output included in
alert
resource: request.requestLine.uri, // vulnerable resource
(link)
key: "vinfo-1918" + request.requestLine.uri + result.join(" ") // unique
key
});
[..]
Example - XML template "vinfo-1918.xml":
<?xml version="1.0" encoding="UTF-8"?>
www.insecuremag.com 53](https://image.slidesharecdn.com/insecure-mag-33-120201200807-phpapp01/85/Insecure-mag-33-53-320.jpg)
![<alert>
<title>Internal Addresses Found</title>
<class>Information</class>
<severity>Low</severity>
<impact>May reveal internal network structure to outside attackers.</
impact>
<impact>Internal IP addresses that have been disclosed could be used
as targets in otherwise blind attacks.</impact>
<discussion>
Vega has discovered references to internal hosts or networks in
publicly accessible content. These addresses may reveal information to an
attacker about the internal network structure, increasing the likelihood of
success for blind attacks involving other vulnerabilities.
</discussion>
[..]
These will be discussed in further detail later in this article.
Request viewer requests saved by alerts are logged). The
user is advised to review the settings in the
Most Vega modules save a specific request preferences menu option of the "Window"
and response pair for inclusion in an alert. menu bar.
This is useful for users who want to verify that
the possible vulnerability is present or investi- Proxy
gate it further. This request and response pair
is made available to the user in a link within The Vega intercepting proxy is meant for use
the alert. Clicking on the "Request" link will with a HTTP client, such as a web browser,
open up the request viewer fast view, with the and allows for close observation and manipu-
request and response in message viewers be- lation of client-server interaction. When the
low a request table. The user can then inspect proxy is enabled, Vega opens a listening TCP
the full HTTP request and response associ- port on a configurable port number (default is
ated with the alert. Right clicking on the re- 8888). HTTP clients can be configured to use
quest log entry in the request table above the the proxy on this port. Firefox is a good choice
message viewers allows for it to be selected of browser for use with the proxy because it
for replay. maintains its own proxy settings, distinct from
system-wide proxy settings.
If a request is selected for replay, a request
editor tab will open in the Scan Info view. The The Vega intercepting proxy can be accessed
user can then modify the request and click the by clicking the proxy button at the top right,
"play" button above the editor region to trans- which will open the proxy perspective. The
mit it to the server. The server response will proxy can be enabled by clicking the "Play"
be rendered in the message viewer below. button in the top left corner of the proxy inter-
The user may modify and send as many re- face, and can be stopped by clicking the stop
quests as they like from this view and then icon. A status indicator in the bottom right cor-
close it when they are finished. ner of the Vega UI will indicate that the proxy
is listening. The proxy perspective is com-
Scanner preferences prised of three major views: the website view,
the request table, and the HTTP message
The scanner preferences allow for resource viewers. The website view in the proxy is iden-
limits to be set. This can constrain the scope tical to the small version embedded in the top
of scans. One useful debugging feature during left corner of the scanner perspective. The re-
module development is the logging of all quest table is a list of all requests saved by
scanner requests (by default, only scanner Vega.
www.insecuremag.com 54](https://image.slidesharecdn.com/insecure-mag-33-120201200807-phpapp01/85/Insecure-mag-33-54-320.jpg)
![These parameters are: an object representing Response processing modules are invoked
the HTTP request, the HTTP response, and when a response is processed. The logic of
the context. The context object connects the the module occurs within the run() function.
module to Vega and exposes the scanner API
to the module developer. These objects are In this example the module is analyzing the
documented in detail on the Subgraph docu- body of the responses using Javascript regu-
mentation website: lar expressions to try and identify email ad-
https://support.subgraph.com/trac/wiki/Respon dresses:
seProcessingModules
function run(request, response, ctx) {
var atDomainRegex = /@(?:[^s.]{1,64}.)+S{2,6}/,
mailRegex = /w[^s@]*@(?:[^s.]{1,64}.)+S{2,6}/g,
strictMailRegex =
/[w!#$%&'*+-/=?^`{|}~.]+@(?:(([a-z0-9]{1}[a-z0-9-]{0,62}[a-z0-9]{1})|[a-z])
.)+(?:aero|arpa|biz|com|coop|edu|gov|info|int|mil|museum|name|net|org|pro|tra
vel|mobi|asia|xxx|[a-z][a-z])/i,
body = response.bodyAsString,
emails = [],
r, sr, i, found;
// First the module attempts to find a basic match of characters@domain, if it
does not, it returns
if (!atDomainRegex.test(body)) return;
// It then attempts to match a more strict regular expression. Any matches are
converted to lowercase and uniquely stored in an array
while (r = mailRegex.exec(body)) {
sr = strictMailRegex.exec(r[0]);
if (sr && emails.indexOf(sr[0]) == -1) {
found = 0;
for (i = 0; i < emails.length; i++) {
if (emails[i] == sr[0].toLowerCase()) {
found = 1;
}
}
if (!found) {
emails.push(sr[0].toLowerCase());
}
}
}
The processing of the response is complete, and the alert can now be generated
if e-mail addresses were identified:
if (emails.length) {
A unique key is then constructed for this alert. come up with a scheme that prevents too
The key in this example is constructed by sort- many duplicate alerts while still providing use-
ing all of the discovered e-mail addresses and ful findings. In this case, the key should pre-
delimiting them with a space. The key is arbi- vent other alerts from being generated for the
trary - it is up to the module developer to same precise type of finding:
var key = emails.sort().join(" ");
www.insecuremag.com 57](https://image.slidesharecdn.com/insecure-mag-33-120201200807-phpapp01/85/Insecure-mag-33-57-320.jpg)
![var uristr = String(request.requestLine.uri);
var uripart = uristr.replace(/?.*/, "");
The alert() function is exposed through the context object:
ctx.alert("vinfo-emails", request, response, {!! // The XML file, the
request, and response objects
"output": emails.join(" "),!// The output to be rendered in the alert
"resource": uripart, ! ! // The URI for the resource field of the
alert
key: "vinfo-emails" + uripart + key! // The unique key
});
}
Regular expressions are not the only tool that from a response body. If it succeeds, a DOM
can be used to analyze response content. The object will be returned. Otherwise
Vega Javascript API comes with JQuery to response.document will be null. JQuery can
analyze content at the DOM level. A DOM ob- then be used on the DOM. An example of this
ject can be obtained by accessing module in use is in
response.document. When the module does scripts/scanner/modules/response/vautocompl
this, Vega will lazily attempt to parse a DOM ete.js:
[..]
if (response.document) {
var form = jQuery("form", response.document);
form.children().each(function() {
if ((this.getAttribute("type") != null) && (this.getAttribute("type") ==
"password")) {
if ((this.getAttribute("autocomplete") == null) ||
(this.getAttribute("autocomplete").toLowerCase() != "off")) {
found++;
}
}
});
}
[..]
Writing a basic module The entry point of a basic module is a function
named initialize() which accepts a single pa-
This guide will explain a very simple example rameter, the "context" object. As with re-
of a basic module. The module is located at sponse processing modules, the context ob-
scripts/scanner/modules/injection/header-injec ject connects the module to Vega, exposing
t.js. the API. The context object for basic modules
is distinct from the object of the same name
As with response processing modules, every for response processing modules.
basic module has a metadata object called
"module": function initialize(ctx) {
var module = { Vega populates a tree-like data structure
name: "HTTP Header Injection known as a path state as it crawls a website.
checks", Basic modules run on path state nodes, which
category: "Injection Modules" may either be files, directories, or parameters.
};
www.insecuremag.com 58](https://image.slidesharecdn.com/insecure-mag-33-120201200807-phpapp01/85/Insecure-mag-33-58-320.jpg)
![Vega handles identifying and iterating over the of the responses to try and identify instances
parameters while still providing great flexibility of header injection.
to the module developer.
The first parameter to this fuzzing example is
In a very simple example, the following API the callback function, listed in the next code
function accepts only fuzzed parameter values snippet, followed by an array of parameter
and requires no knowledge of where in the values to be injected. The last parameter is an
application the module is. This module at- optional Boolean indicating whether the in-
tempts to inject values into the application. jected values are to be appended to a pa-
The callback function examines the headers rameter seen by the crawler, in this case it is
set to true:
ctx.submitMultipleAlteredRequests(process, ["bogusnVega-Inject:bogus",
"bogusrVega-Inject:bogus"], true);
}
There are many functions for generating new https://support.subgraph.com/trac/wiki/PathSt
requests exposed through the context object ate
for basic modules. The reader is invited to
view them all at the Subgraph documentation When a basic module queues requests for the
website: crawler (for example, to fuzz parameters), it
must register a callback function that Vega will
https://support.subgraph.com/trac/wiki/BasicM run for each of the responses. The callback
oduleContext function is passed three parameters by Vega:
the request object, response object, and con-
The path state node of the module is accessi- text object. In this example, the callback func-
ble through the context object, if the module tion checks the response headers to see if in-
writer wishes to know information about the jected values are present, generating an alert
path state node on which it is running. The if it finds them:
path state structure is explained here:
function process(req, res, ctx) {
if (res.hasHeader("Vega-Inject")) {
ctx.alert("vinfo-header-inject", request, response, {
message: "Injected Vega-Inject header into response",
resource: request.requestLine.uri
});
}
More advanced analysis and shell injection. It is also possible to ana-
lyze the timing of responses. Readers inter-
The basic module explained in this tutorial is ested in experimenting with Vega modules are
among the simplest examples. It is possible advised to read the API documentation for ba-
for basic modules to send many requests to sic modules on https://support.subgraph.com.
perform more complex logical analysis.
It is useful to note that the ctx.debug() function
One facility for doing this provided by Vega is can be used to print output to the console dur-
page fingerprinting, where Vega distills page ing module development. It is also possible for
contents to a simpler representation such that modules to store data in a key-value database
two fingerprints can be efficiently compared to share data between modules. The reader
for page differences. This is used as the basis should refer to the documentation on the con-
for determining positive or negative results in text objects for more information.
several injection modules, including blind SQL
www.insecuremag.com 59](https://image.slidesharecdn.com/insecure-mag-33-120201200807-phpapp01/85/Insecure-mag-33-59-320.jpg)
Insecure mag-33
- 4. With this issue of (IN)SECURE Magazine, we enter our seventh year of publication. This time around we focus on Android security, we bring you the thoughts of the Facebook CSO and THE man behind Metasploit. To top if off, there are articles on web security, shellcode, mobile security, and more! February is going to be a busy time for every information security company. The monumental RSA Conference is opening its doors later this month, and we'll be there to cover all the news and meet with companies and readers. I'm looking forward to the expo floor safari, there's always interesting technologies to discover. Look out for our camera, you might just be featured in an upcoming issue! Mirko Zorz Editor in Chief Visit the magazine website at www.insecuremag.com (IN)SECURE Magazine contacts Feedback and contributions: Mirko Zorz, Editor in Chief - mzorz@net-security.org News: Zeljka Zorz, Managing Editor - zzorz@net-security.org Marketing: Berislav Kucan, Director of Operations - bkucan@net-security.org Distribution (IN)SECURE Magazine can be freely distributed in the form of the original, non-modified PDF document. Distribution of modified versions of (IN)SECURE Magazine content is prohibited without the explicit permission from the editor. Copyright (IN)SECURE Magazine 2012. www.insecuremag.com
- 5. Online scam susceptibility of Regionally, respondents who indicated they are from the Southwest are the most American consumers susceptible, while respondents from the Midwest and Pacific are the least. PC Tools, in collaboration with the Ponemon Institute, announced the findings of its online The survey results also indicated that scam susceptibility study of 1,858 American respondents from the following demographics consumers. are more susceptible to online scams: The results of the survey show that close to • 18-25 year olds half of US respondents think that they would • Females be likely to provide personal or financial • Less than a high school diploma information online in each of the test • Household income of $25,000 - $50,000 scenarios presented: • Reside in the Southwest. Unfortunately, many consumers don’t realize that some online scams don’t involve malware. Traditional internet security is essential to maintain protection against viruses or The survey results also indicate that certain malicious files and websites, but demographic groups are more susceptible cybercriminals are changing their methods by than others. For example, respondents who tricking consumers into revealing their indicated they are Independent supporters are personal information, so this requires a very the most susceptible to online scams, while different protection approach. supporters from the Green Party are the least. www.insecuremag.com 5
- 6. Brazen Brazilian hackers opening spotted another business venture initiated by the criminals. cybercrime schools "To help new 'entrepreneurs' or beginners interested in a life of cybercrime, some Brazilian bad guys started to offer paid courses," he revealed. "Others went even further, creating a Cybercrime school to sell the necessary skills to anyone who fancies a life of computer crime but lacks the technical know-how." A number of different courses are offered, and while some seem like legitimate ones - how to Brazilian hackers are known for their become a designer, a Web designer, a hacker, preference for stealing and misusing phished a programmer - other not so much as they banking credentials and credit card numbers, offer to teach how to become a "banker", a but also for their penchant to openly brag defacer or a spammer. online about their illegal activities. The courses can be bought online but - as This relaxed attitude regarding the possibility unbelievable as it sounds - aspiring of getting caught and tried for their illegal cybercriminals can also attend real-world actions is due to the country's extremely classes at a location that is shared freely and, inadequate anti-cybercrime laws, explained obviously, without any fear of law enforcement Kaspersky Lab's Fabio Assolini, who recently reactions. Hackers steal $6.7 million in bank and compromising an employee computer in the Rustenburg Post Office. cyber heist Once the offices were closed for the New Year holidays, the gang put their plan in motion. They accessed the computer from a remote location and used it to break into Postbank's server system and transfer money from various accounts into the ones they opened. Having also raised the withdrawal limits on those accounts, money mules had no problem withdrawing great amounts of money from ATMs in Gauteng, KwaZulu-Natal and the Free State during the next few days, stopping A perfectly planned and coordinated bank completely when the offices were opened robbery was executed during the first three again on January 3. days of the new year in Johannesburg, and left the targeted South African Postbank - part Unfortunately, the Postbank's fraud detection of the nation's Post Office service - with a loss system hasn't performed as it should, and the of some $6.7 million. crime was discovered only after everyone returned to work after the holiday break. According to the Sunday Times, the cyber Apparently, it should not come as a surprise - gang behind the heist was obviously very well according to a banking security expert, "the informed about the post office's IT systems, Postbank network and security systems are and began preparing the ground for its shocking and in desperate need of an execution a few months before by opening overhaul." accounts in post offices across the country www.insecuremag.com 6
- 7. Mozilla offers alternative to OpenID "Many web sites store extensive user data and act on behalf of the user. While the browser may be fully under the user’s control, many of the services that users enjoy are not. Sometimes, these web services handle data in ways that are of questionable value to the user, even detrimental," says Ben Adida, Mozilla's Tech Lead on Identity and User Data. Mozilla has been working for a while now on a "It’s clear that Mozilla needs to step up and new browser-based system for identifying and provide, in addition to the Firefox browser, authenticating users it calls BrowserID, but its certain services to enhance users’ control only this January that all of its sites have over their online experience and personal finally been outfitted with the technology. data." Mozilla aims for BrowserID to become a more Apart from BrowserID, Mozilla is also looking secure alternative to OpenID, the to launch Boot to Gecko (B2G), a standalone decentralized authentication system offered to mobile web-based operating system, and an users of popular sites such as Google, app store. Yahoo!, PayPal, MySpace and others. Stratfor hack exposes UK, US and exchanged and worked on by the organization's employees; and around 75,000 NATO officials to danger, phishing credit card details complete with security codes required for no card present transactions. The Guardian has hired cyber-security expert John Bumgarner to rifle through the information already leaked by the hacker group, and he has ascertained that thousand of emails and passwords belonging to UK, US and NATO officials were thusly made public. 19,000 email addresses and passwords and During the last days of 2011, Anonymous other personal data belonging to US military attacked Stratfor, a US-based research group personnel were revealed, as well as those of that gathers intelligence and produces seven officials of the UK's Cabinet Office, 45 political, economic and military reports that of the Foreign Office, 14 of the Home Office, help government organizations and major 67 police officers of the London Metropolitan corporations asses risk. Police and other officials, two employees with the royal household, 23 workers/members of Among the data they have managed to steal the Houses of Parliament, and a number of from its servers were names, home intelligence officers. 242 Nato staffers have addresses, credit card details and passwords also had their emails revealed. of Stratfor clients, 17,000 of which they have immediately shared with the public in order to British officials and the government are still prove the veracity of their claims. not worried about the revealed information posing any threat to national security. To be All in all, the hackers said that they have sure, the revealed (easily decryptable) managed to put their hands on around passwords are those used by Stratford 860,000 usernames, emails, and hashed customers to access the content offered by passwords; internal emails and documents the think-tank and not their email accounts. www.insecuremag.com 7
- 8. Researchers demonstrate tragic The remaining five (General Electric's D20ME, Koyo's Direct LOGIC H4-ES, Rockwell state of SCADA security Automation's Allen-Bradley ControlLogix and Allen-Bradley MicroLogix, Schneider Electric's Modicon Quantum, and Schweitzer's SEL-2032) displayed a dazzling array of back door accounts, old hardware and firmware, lousy security controls, configuration files easily obtainable by attackers, buffer overflow and remotely exploitable vulnerabilities, unexpected crashes, weak password implementation and authentication protection, and inability to upload custom firmware. Despite the reservations of some security experts that have questioned the researchers' action of making this information public before sharing it with the vendors, most industrial control security experts are satisfied that someone has finally pointed out these things they knew for years. "A large percentage of these vulnerabilities the vendor already knows about and has chosen to live with, so this is not news to them," commented Dale Peterson, CEO of SCADA security firm Digital Bond, which At the SCADA Security Scientific Symposium organized the project, and said that the best held in Miami, visitors had the opportunity to way to avoid uncomfortable disclosures is to hear a damning presentation held by do a better job making secure products. researchers grouped around Project Basecamp which revealed that their testing of He expressed his belief that this presentation six widely used programmable logic should be the moment when SCADA systems controllers (PLCs) resulted in the discovery of and PLC vendors finally realize that they have alarming security bugs that are mostly design to take security more seriously. For their part, flaws and (even!) features, and of the fact that the researchers collaborated with Rapid 7 and some of them can't even take a probing Tenable in order to create test modules for the without crashing. Metasploit Framework and the Nessus scanner for these vulnerabilities, in the hope One of the devices, the Control Microsystems' that vendors will be pushed to make changes SCADAPack, bricked early on into testing. with security in mind. Qualys expands its FreeScan service help SMBs audit and protect their web sites from security vulnerabilities and malware infections. The new FreeScan service allows SMBs to scan their web sites for of malware, network and web application vulnerabilities, as well as SSL certificate validation, helping web site owners identify risk before hackers do in order Qualys announced its new and improved to prevent data beaches and protect online FreeScan service (freescan.qualys.com) to visitors from infections. www.insecuremag.com 8
- 9. Pwn2Own 2012: Changed rules, browser will be awarded 32 points," say the rules. "When the contest begins we will be bigger prizes, no more mobile hacks announcing 2 vulnerabilities per target that were patched in recent years. The first contestant (or team) who is able to write an exploit for the announced vulnerabilities will be awarded 10, 9, or 8 points depending on the day the exploit is demonstrated." For exploiting the already known vulnerabilities, contestants will only have to Pwn2Own, one of the most anticipated overcome DEP, and don't have to escape from hacking contests that takes place each year at a sandbox or protected mode. The browsers the CanSecWest conference in Vancouver, will be installed on Windows XP and Snow British Columbia, is set to unfold under Leopard, and their versions will be made dramatically different rules this year. public at the beginning of the contest. First and foremost, smartphone hacking is no For the zero-days, hackers will be targeting longer on the table. This year edition will also browsers on fully patched Windows 7 and reward the three most successful participants Mac OS X Lion machines. Also, one with cash prizes of $60,000, $30,000 and requirement that contestants must fulfill in $15,000, respectively (plus the laptops they order to win is to demonstrate at least one manage to compromise). zero-day vulnerability on one of the targets. Also, a successfully compromised target will As the in the previous year, Google is offering not be pulled from the competition as in special prizes for Chrome "ownage": $20,000 previous years. All contestants can attack all for a set of bugs present only in Chrome that targets during the whole three days of the allow full unsandboxed code execution, and contest, and the contest will be point-based. $10,000 for a compromise that used bugs both in Chrome and the OS for the same type "Any contestant who demonstrates a working of code execution. 0day exploit against the latest version of the Entrust Discovery now offers "Understanding that today's organization often manages complex certificate environments, Microsoft CAPI query capabilities we provide more methods of discovering certificates and enhance the policy options once under management," said Entrust President and CEO Bill Conner. Entrust Discovery assists organizations in gaining a complete perspective of deployed certificates. The solution finds, inventories and manages digital certificates across diverse systems to help prevent outages, data breach Entrust expands its certificate discovery and non-compliance. solution, Entrust Discovery, by broadening search capabilities for digital certificates The solution now offers more policy alert residing within Microsoft's Cryptographic APIs fields, including issuer DN, expiry status, (CAPI). And now with more than 25 basic or subject DN, key (e.g., RSA 2048), time valid, custom policy alert fields, Entrust Discovery subject alt names (SAN) and certificate offers stronger compliance tools. signature method. www.insecuremag.com 9
- 10. Targeted attacks will change the than 250 employees will have to appoint a privacy officer. economics of security Corporations risk being fined up to 2 per cent of their global turnover for failure to adequately secure citizens’ information. In addition, in a new “right to be forgotten” ruling, customers can request details of the information that companies hold about them and ask for it to be amended or removed. European Justice Commissioner, Viviane Bruce Green, Chief Operating Officer at M86 Reding, unveiled the new European Privacy Security, commented: “While we applaud the Directive, designed to safeguard personal, move to strengthen safeguards around identifiable information that is stored by individuals’ private information, we recognize private and public sector organizations. that this harmonization of data privacy rules across Europe will increase the data All 27 European member states will be management overhead for companies of all governed by the new rules, which could see sizes. The prospect of being fined two per companies being fined 2 per cent of global cent of turnover will change the economics of turnover if their customers’ privacy is security, because the cost of compliance breached. compared to the financial risk of a breach will now fall firmly in favor of security for global Under the new rules, all UK companies that enterprises. This will make information suffer a security breach will have to inform the security a discussion for the boardroom, not Information Commissioner within 24 hours of just the domain of compliance specialists and discovering a breach. Companies with more privacy officers.” Symantec advises customers to stop control sessions and, thus, access to systems and sensitive data. If the cryptographic key using pcAnywhere itself is using Active Directory credentials, they can also carry out other malicious activities on the network. - If the attackers place a network sniffer on a customer’s internal network and have access to the encryption details, the pcAnywhere traffic - including exchanged user login credentials - could be intercepted and decoded. In a perhaps not wholly unexpected move, The white paper also contains security Symantec has advised the customers of its recommendations for minimizing the potential pcAnywhere remote control application to stop risk of using the software, since some using it until patches for a slew of customers cannot stop using it because its of vulnerabilities are issued. According to a critical importance to their business. company white paper, the risks for the users are the following: Martin McKeay, Security Evangelist at Akamai Technologies, pointed out that most remote - Man-in-the-middle attacks (depending on the desktop applications are directly exposed to configuration and use of the product) because the Internet because they are used by service of vulnerable encoding and encryption providers for troubleshooting their clients' elements within the software. network equipment, and that that is unlikely to - If the attackers get their hands on the change in the near future. cryptographic key they can launch remote www.insecuremag.com 10
- 12. The popularity of Android-based devices is driving their increased adoption in enterprise mobile applications, where security is a significant concern. In ad- dition, designers of embedded systems are considering using Android for all forms of human-machine interfaces (HMI) in practically all major industri- es—automotive center stacks, medical device graphical interfaces, and home smart energy management panels, just to name a few. Android brings to electronic products the based devices in order to make them more power of open source Linux augmented with suitable for enterprise, government, and other the graphical interfaces and app store infra- mission-critical environments. structure of one of the world’s most popular mobile operating systems. Android security retrospective In addition, the rapidly emerging market for As part of Android’s original introduction in Android Mobile Device Management (MDM) 2008, Google touted improved security in its solutions provides developers with the prom- smartphones. Google’s website ise of a world-class remote device manage- (code.google.com/android) lauded the plat- ment infrastructure that can seamlessly tie form’s security: “A central design point of the into traditional back-end IT systems. MDM Android security architecture is that no appli- functions include remote monitoring and audit- cation, by default, has permission to perform ing, firmware updates, application configura- any operations that would adversely impact tion management and control, data-at-rest en- other applications, the operating system, or cryption, VPN services, remote wipe (e.g., the user.” Days after the release of the first when an embedded device is believed to be Android phone, the G1, a well-publicized, se- compromised), and more. vere vulnerability was found in the phone’s Web browser. But the G1’s security woes This article discusses the challenges and so- didn’t end there. lutions for improving the security of Android- www.insecuremag.com 12
- 13. In November, hackers discovered a way to These vulnerabilities are specific to the An- install arbitrary programs on the phone, droid stack that runs on top of Linux. Android prompting this lament from Google: "We tried is, of course, susceptible to Linux kernel vul- really hard to secure Android. This is definitely nerabilities as well. The rapid development a big bug. The reason why we consider it a and monolithic architecture of Linux has been large security issue is because root access on well publicized. Lead Linux kernel authors the device breaks our application sandbox." have published multiple installments of a Linux kernel development statistical overview, In fact, the Android bug would silently and in- and the numbers are staggering. visibly interpret every word typed as a com- mand, and then execute it with superuser With 20,000 lines of code modified per day, privileges. 6,000 unique authors, and rapid growth in its overall code base, it should come as no sur- In late 2010, security researchers uploaded to prise that dozens of Linux kernel vulnerabili- the Android market a spoofed Angry Birds ties are reported each year, and that a steady game application that surreptitiously down- stream of undiscovered vulnerabilities are la- loaded other apps without the user’s approval tent in every Linux distribution deployed to the or knowledge. field. The extra downloads were malicious, stealing While a significant portion of the growth and the phone’s location information and contacts, churn in the Linux kernel code base is due to and sending illicit text messages. As part of the continual adding of support for new micro- their work, the researchers reported numerous processors and peripherals, the core kernel weaknesses in Android, including a faulty use itself, including networking and file system of SSL, a lack of application authentication, an support, also undergoes rapid change. easy method of breaking out of the Android Dalvik virtual machine sandbox via native CVE-2009-1185 documents a flaw in the Linux code, and the focus of the attack—a weak netlink socket implementation, and is but one permissions architecture. example of a Linux vulnerability that has al- legedly been used to compromise Android de- Next, we visit our favorite website, the U.S. vices. CVE-2009-2692, informally known as CERT National Vulnerability Database. A the proto-ops flaw, is a set of bugs in the Linux search on Android turns up numerous vulner- kernel’s management of file and network ac- abilities of varying severity. Here is a sampling cess objects. of the worst offenders: ! A trivial user mode program can be used to • CVE-2011-0680: Allows remote attackers to subvert an Android system using this vulner- read SMS messages intended for other re- ability. The proto-ops flaw was latent in the cipients. Linux kernel for eight years before research- • CVE-2010-1807: Allows remote attackers to ers discovered it. execute arbitrary code. • CVE-2009-2999, -2656: Allows remote at- Because its architecture for kernel object tackers to cause a denial of service (applica- management is so entrenched, Linux remains tion restart and network disconnection). susceptible to the vulnerability as new device • CVE-2009-1754: Allows remote attackers to drivers and communication mechanisms are access application data. added to the code base. • CVE-2009-0985, -0986: Buffer overflows al- low remote attackers to execute arbitrary Android device rooting code. Android rooting (also known as jailbreaking) is We point out these particular vulnerabilities the process of replacing the manufacturer- because they fall into the most serious sever- installed kernel (Linux) and/or its critical file ity category of remote exploitability. system partitions. Once a device is rooted, the hacker can change Android’s behavior to suit his or hers particular desires. www.insecuremag.com 13
- 14. The term rooting originates from the UNIX In this case, rooting is usually accomplished concept of root privilege, which is needed to with a form of side-loading/booting using an modify protected functions. The goals of An- SD card or USB to host or install the custom droid hackers range from the hobbyist’s desire ROM. The manufacturer-installed boot loader to overclock a CPU for better performance (at does not cryptographically authenticate the the expense of battery life) and install custom Android firmware, paving the way for ROM applications, to more malicious pursuits, such execution. as illegally obtaining carrier network services, and installing key loggers and SMS snoopers. Some device makers have gone to great lengths to prevent rooting for various reasons. The collection of new and replaced files in- Obviously, many developers using Android will stalled by the hacker is referred to as a cus- want to lock down the Android OS completely tom ROM, another imperfect reference to the to prevent unauthorized modification and ma- concept of firmware that is often deployed in licious tampering. read-only memory. One of the most high-profile secure boot fail- Android vulnerabilities are often used by ures in this realm is the Amazon Kindle. The hackers to root Android phones. The rate of presumed aim of locking down the Kindle is to vulnerability discovery is such that practically force users to access Amazon content and every Android consumer device has been require use of the Kindle e-reader software. rooted within a short period of time, some- The Amazon secure boot approach attempted times within a day or two of release. to authenticate critical system files at startup using digital signature checks. Hackers used In addition to software vulnerabilities, secure vulnerabilities in Linux to circumvent these boot problems are another major source of checks and run malicious boot code, rooting Android rooting attacks. Some Android device the device. makers, such as Barnes and Noble with its Nook Color, have permitted (if not encour- Yes, we paint a grim picture of Android secu- aged) rooting in order to facilitate a wider de- rity. However, the picture is based on a simple veloper community and device sales. fact that shouldn’t be surprising—Android was never designed to provide a high assurance of security. ANDROID VULNERABILITIES ARE OFTEN USED BY HACKERS TO ROOT ANDROID PHONES Mobile phone data protection: A case can use a layer four (OSI model) SSL VPN study of defense-in-depth client to establish a protected data communi- cation session. An IPsec VPN application, Android’s tremendous popularity, juxtaposed running at layer three, can be used to create a with its lack of strong security, has sparked a second, independent connection between the rigorous scramble by software vendors, de- smartphone and the remote endpoint (Figure vice OEMs, systems integrators, and govern- 1). ment security evaluators to find ways to retro- fit Android-based devices with improved sys- This secondary connection uses independent tem security. public keys to represent the static identities of the endpoints. The data in transit is doubly One approach to raising the level of assur- encrypted within these two concurrent con- ance in data protection within an Android- nections. This layered security approach is an based device is to employ multiple encryption example of defense-in-depth. layers. For example, an Android smartphone www.insecuremag.com 14
- 15. Figure 1 - Multiple layers of encryption within Android. The concept of defense-in-depth originated in attacker must break both the SSL and IPsec the military—multiple layers of defense, such encryption layers. as a combination of mines and barbed wire, rather than just mines or barbed wire alone, to Clearly, this layered approach depends on the increase the probability of a successful de- independence of the layers. Most importantly, fense, as well as potentially to slow the pro- the SSL and IPsec private keys must be inde- gress of an attacker. pendently stored and immune to a single point-of-failure compromise. However, in a Defense-in-depth has been successfully ap- typical Android environment, both the SSL and plied in war since ancient times, and the con- IPsec long-term private keys are stored within cept is alive and well in the information secu- the same flash device and file system. Fur- rity age. thermore, the key stores are not protected against physical attacks. Let’s consider a few of the threats against an SSL data protection application. An attacker This environment provides numerous single can attack the application directly, perhaps points of compromise that do not require so- exploiting a flaw in the SSL software stack, to phisticated attacks. A single Android root vul- disable encryption entirely or steal the encryp- nerability or physical attack on the storage de- tion keys residing in RAM during operation. An vice can compromise both sets of keys and attacker can try to steal the static public SSL encryption layers. keys stored on disk. If these keys are com- promised, the attacker can impersonate the The run-time environment must provide strong associated identity to gain access to the re- isolation of the SSL and IPsec application lay- mote client over a malicious SSL session. ers, and the run-time environment itself must not provide an attack surface through which to Malware elsewhere in the Android system can break that isolation. Much of the research and use side channel attacks to break the SSL en- product development aimed at Android secu- cryption and recover its keys. rity has focused, in one form or another, on providing sandboxes for data isolation and the Layered SSL/IPsec data protection is a sensi- protected execution of critical functions. Those ble application of defense-in-depth to counter sandboxes are used to realize the layered en- these threats. If an attacker is able to break cryption approach. the SSL encryption, the IPsec layer will con- tinue to protect the data. An attacker may be Let’s now compare and contrast the various able to steal the SSL keys but not the IPsec approaches for Android sandboxing. Develop- keys. The attacker may be able to install mal- ers considering the adoption of Android in ware into the SSL application but not the their next-generation designs can use this IPsec application. The SSL application may comparison to make sensible security exhibit side channel weaknesses to which the choices. IPsec application is immune. To succeed, the www.insecuremag.com 15
- 16. Android sandboxing approaches Multi-boot Separate hardware The multi-boot concept has been attempted on a handful of laptops and netbooks over the One sandboxing approach is to have multiple years. In a dual boot laptop scenario, a sec- microprocessors dedicated to the differing ondary operating system, typically a scaled- tasks. While Android smartphone OEMs are down Linux, can be launched in lieu of the unlikely to add additional hardware cost to main platform operating system. The scaled- their designs, custom electronic product de- down system is typically only used for Web velopers may have more options depending browsing, and the primary goal is to enable on many factors, including form-factor flexibil- the user to browse within a handful of seconds ity. from cold boot. The secondary operating sys- tem resides in separate storage and never For example, a PCI-capable design may be runs at the same time as the primary platform able to host an IPsec VPN card that wraps the operating system. In some cases, the light- second layer encryption around the main weight environment executes on a secondary processor’s Android SSL. In some cases, microprocessor (e.g., an ARM SoC independ- however, the additional hardware size, weight, ent of the netbook’s main Intel processor). On power, and cost will be prohibitive for this ap- an Android mobile device, the primary Android proach. can be hosted on internal NAND flash, and a secondary Android can be hosted on an in- serted microSD card (Figure 2). Figure 2 – Dual-boot Android. The secondary operating system provides webtop runs as a set of applications on top of good isolation from a security perspective. the primary operating system. However, the inconvenience of rebooting and In the case of the Motorola Atrix Android the inability to seamlessly switch between en- smartphone released in 2011, the webtop vironments has severely limited adoption. The sandbox is an independent file system parti- multi-boot option is also impractical for the tion that contains a limited Ubuntu Linux- layered encryption use case that requires based personality (Figure 3). concurrent execution of the sandboxes. The primary Android partition is located on the Webtop same internal NAND flash device within the phone. The Atrix webtop is intended to provide The webtop concept provides a limited brows- a desktop-like environment for users that dock ing environment (the webtop), independent the phone on a separately purchased KVM from the primary operating system environ- (keyboard/video/mouse) apparatus. ment. However, instead of a dual boot, the www.insecuremag.com 16
- 17. Figure 3 - Android webtop environment. While webtop was most likely not intended as and its own encryption of the e-mail folders a security capability, one mapping of this ap- resident on the phone. proach to the layered encryption use case is to execute IPsec from the primary Android en- Some MDM solutions use Android profiles to vironment and an SSL-based Web session divide the Android system into two sets of ap- from the webtop sandbox. plications—one for the user’s personal envi- ronment and one for the enterprise-managed The problem with this approach is that the en- environment (Figure 4). tire Linux kernel, including its TCP/IP stack, is depended upon for the isolation of the When the enterprise profile is invoked, the webtop’s SSL from the Android IPsec. MDM product may automatically turn on en- cryption for data associated with that profile. Mobile Device Management (MDM) Numerous other Linux controls can be used to encrypted containers improve the isolation of profiles, including chroot jails and operating system-level re- The growing popularity of Android mobile de- source grouping techniques like OpenVZ. vices and the desire to use them in the work- place has spawned dozens of MDM products Clearly, this approach can be used to imple- and companies. The two main purposes of ment the layered encryption use case—the MDM are to provide mobile data protection MDM application can create an SSL connec- and IT management services. tion on top of the underlying Android’s IPsec connection. Manageability includes application configura- tion (ensuring that all employees have an ap- However, once again, the underlying Android proved set of preloaded software), auditing, operating system is relied upon for the secu- document management, and remote wipe rity of both layers. (disabling the handset when an employee leaves the company). Remoting Data protection covers both data at rest and One approach to enterprise data protection in data in transit (e.g.VPN to the corporate net- Android is to not allow any of the enterprise work). data on the mobile device itself. Rather, the only way to access enterprise information is Android MDM solutions often use application- using a remote desktop and/or application vir- level encryption. For example, an enterprise tualization. When the device is not connected e-mail client may implement its own encryp- to the enterprise (e.g. offline operation), en- tion protocol for the connection between a terprise applications and services are unavail- mobile device and an enterprise e-mail server, able. www.insecuremag.com 17
- 18. Figure 4 - MDM containers. While the result is a neutered device that de- Type-2 hypervisor feats the purpose of having such a powerful hardware platform with multiple cores and Type-2 hypervisors are similar to webtops and multimedia accelerators, there are certainly MDM containers in that the secondary envi- use cases that can take advantage of remot- ronment runs as an application on top of the ing. primary operating system. However, instead of hosting only a browser, the secondary per- Remoting precludes the requirement for local sona is a full-fledged guest operating system data protection; however, our use case for running within a virtual machine created by layered data-in-motion protection remains. the hypervisor application (Figure 6). The remoting application (Figure 5) provides SSL encryption while the underlying Android The hypervisor uses the primary operating runs IPsec. Once again, the underlying An- system to handle I/O and other resource droid operating system is relied upon for the management functions. security of both layers. Figure 5 - Remoting. Figure 6 - Type-2 hypervisor. www.insecuremag.com 18
- 19. Type-2 mobile hypervisor products, such as Type-1 hypervisor VMware MVP, are used to provide an enter- prise management persona on top of the pri- Type-1 hypervisors also provide functional mary Android environment. The virtualized completeness and concurrent execution of a Android can use an SSL connection to the en- secondary enterprise persona. However, be- terprise while the underlying Android’s IPsec cause the hypervisor runs on the bare metal, is also used to wrap the communication be- persona isolation cannot be violated by weak- tween endpoints. nesses in the persona operating system. Thus, a Type-1 hypervisor represents a prom- However, once again, the Type-2 model fails ising approach from both a functionality and to provide strong isolation. Faults or security security perspective. But the hypervisor vul- vulnerabilities in the primary general-purpose nerability threat still exists, and not all Type-1 operating system will impact the critical func- hypervisors are designed to meet high levels tions running in the virtual machine. Further- of safety and security. more, Type-2 hypervisor applications de- ployed in the enterprise space have been One particular variant, the microkernel-based found to contain vulnerabilities that break the Type-1 hypervisor, is specifically designed to sandbox. meet high-assurance, security-critical re- quirements. Microkernels are well known to Sandboxes built on sand provide a superior architecture for safety and security relative to large, general-purpose op- Constant reader, hopefully you observe as erating systems such as Linux and Android. obvious the common weakness among all of the sandboxing approaches previously de- In a microkernel Type-1 hypervisor, system scribed. Multiple Android applications, MDM virtualization is built as a service on the mi- containers, remoting applications, webtops, crokernel. Thus, in addition to isolated virtual and Type-2 hypervisors all attempt to retrofit machines, the microkernel provides an open security to the Android kernel itself. standard interface for lightweight critical appli- cations, which cannot be entrusted to a The Android/Linux system, while providing rich general-purpose guest. For example, SSL can multimedia functionality of which mobile and be hosted as a microkernel application, pro- embedded designs can take good advantage, viding the highest possible level of assurance is riddled with security vulnerabilities that sim- for this encryption layer. IPsec packets origi- ply cannot be avoided. High-assurance secu- nating from Android are doubly encrypted with rity must be designed from the beginning. the high-assurance SSL layer service before transmission over the wireless interface (Fig- But while high assurance cannot be retrofitted ure 7). to Android itself, it can be retrofitted at a sys- tem level. Let’s take a look at how. Figure 7 - Microkernel Type-1 hypervisor approach to layered data-in-motion encryption. www.insecuremag.com 19
- 20. The real-time microkernel is an excellent comparable to today’s Trusted Platform Mod- choice for practically any mobile and embed- ules (TPMs) found in laptops and PCs. ded system since the microkernel can host any real-time application not appropriate for Most of the work being done in this area is in the Android/Linux environment. its infancy; full specifications are not complete, and commercial products that incorporate The microkernel Type-1 hypervisor typically these standards are not yet on the market. uses the microprocessor MMU to isolate the memory spaces of the primary Android envi- However, the concept of the MTM can be ronment and the native SSL encryption appli- combined with the functionality of a smartcard cation. However, device drivers in Android to provide a mobile hardware root of trust with may use DMA that can violate the memory secure key store capability. partitioning by bypassing the MMU entirely. This approach offers a single element that can Running the hypervisor in TrustZone on an provide a secure trust anchor for secure boot applicable ARM-based microprocessor, using and remote attestation, as well as a secure an IOMMU, or using the hypervisor itself to key store for device, user, and application mediate all DMA bus masters are all potential keys and certificates. approaches to guarding this attack vector. For example, a smartcard chip can be incor- The isolation properties of some secure mi- porated into a microSD device and attached to crokernels can even protect against sophisti- a smartphone (Figure 8). This approach pro- cated covert and side-channel, software-borne vides the physical security benefits of a se- attacks. cure element while allowing credentials to move with the user by removing and then in- Physical security serting the microSD into another device. Now that we have an approach that prevents Of course, implementations will vary depend- software attacks from breaking the sandbox ing on the types and sophistication of physical between protection layers, let’s take defense- protections available. But a hardware-based in-depth a step further and consider how the root of trust enables a higher-level FIPS-140 layered encryption system can be protected certification and provides an important addi- from physical attacks. For example, a lost or tional layer of security independent of the stolen mobile device in the hands of a sophis- microkernel-based runtime environment isola- ticated attacker is susceptible to memory tion. snooping, power analysis, and other invasive and non-invasive physical attacks. Summary While physical protection of the entire device Layered encryption as a defense-in-depth may not be practical, targeted physical protec- strategy is a sensible approach to increasing tions can make a huge difference in overall the assurance of Android-based data protec- system security. A secure element can be tion services. However, it is not sensible to run used to provide physical protection of critical both layers within the Android environment parameters, including private keys. Several itself. There is simply too much vulnerability to industry standards bodies are examining this prevent both layers from being simultaneously requirement and offering solutions. subverted. Designers considering Android must also carefully sandbox critical security For example, GlobalPlatform functions outside of the Android system. Mod- (www.globalplatform.org) recommends the ern microprocessors and system software so- use of TrustZone, coupled with some form of lutions provide the requisite features to get the secure element, to protect critical parameters best of both worlds—the power of Android’s and cryptographic functions used for mobile multimedia and applications deployment infra- payments. The Trusted Computing Group structure alongside, but securely separated (www.tcg.org) is working on the specification from critical system security functions. for a Mobile Trusted Module (MTM) that is www.insecuremag.com 20
- 21. Figure 8 - Adding physical security protection via attached smartcard to the microkernel Type-1 hypervisor. Kirk Spring is the VP of Technology for SafeNet. Currently he oversees SafeNet’s strategic development of security solutions that includes technology sharing of both its commercial and government products. Mr. Spring earned his bachelor of science in Computer Engineering from Oakland University and has been with SafeNet since 2001. Prior to SafeNet, Mr. Spring was at Harris Corporation, Allied Signal Corporation, and Hughes Ground Systems. David Kleidermacher is CTO at Green Hills Software where he is responsible for technology strategy, platform planning, and solutions design. Kleidermacher is a leading authority in systems software and security, includ- ing secure operating systems, virtualization technology, and the application of high robustness security engi- neering principles to solve computing infrastructure problems. Kleidermacher earned his bachelor of science in computer science from Cornell University and has been with Green Hills Software since 1991. www.insecuremag.com 21
- 22. Joe Sullivan is the Chief Security Officer at Facebook, where he manages a small part of a company-wide effort to ensure a safe internet experience for Facebook users. He and the Facebook Security Team work internally to de- velop and promote high product security standards, partner externally to promote safe internet practices, and coordinate internal investigations with outside law enforcement agencies. Being the CSO of Facebook certainly puts skills, the ability to develop creative solutions you into the spotlight. How have your prior to new and unique problems, and the ability to positions prepared you for your work at stay focused on addressing real risks and Facebook? threats while under great scrutiny, are critically important for succeeding in my role at Face- I can think of two important ways my prior po- book. sitions have helped prepare me for my current responsibilities. Before Facebook I worked as Facebook has partnered with the National a federal prosecutor working on cybercrime Cyber Security Alliance on the STOP. cases that were in the media every day and THINK. CONNECT. campaign over two then worked at eBay during the early part of years ago. What are your thoughts on how the 2000s when that company was celebrated public-awareness-raising campaigns can and scrutinized. be improved in the future? In both of those places I was challenged to If you look at internet education safety cam- develop creative solutions - because we were paigns before this effort by NCSA, you see a breaking new ground in areas where there bunch of different parallel efforts focused on was not much precedent. Likewise, in both I the same problems but using different tactics learned how to stay effective and focused and terminology. This initiative is important even when under a serious microscope. Both because it brings together an incredibly wide www.insecuremag.com 22
- 23. spectrum of technology, communication and port meaningful bugs. On the contrary, we other companies to work with government on have found that there is an incredibly vibrant developing unified messaging. entrepreneurial security community around the world that is passionate about engaging Having consistent terminology is critical to on web application security. education in a complex area and with this ef- fort the sum of our individual efforts working We have had submissions from over 16 coun- together is much greater than it would be if we tries and have already payed out over $150, invested the same in education but without 000 in bounties. In the process we have built this degree of coordination. great relationships with some amazing re- searchers from every corner of the globe. And Facebook launched its bug bounty pro- yes, we do have a summer intern coming who gram in August last year and has already we met through the program. doled out quite a sum to outside security experts. Have there been any great sur- I don't think it has influenced the way we re- prises? Has the program influenced the view code, but it does make us feel even bet- way that the security team approaches ter about the overall review process we have code reviewing? Did you offer employment in place being as complete as possible. We to a particularly successful bug hunter/are intend to keep investing in this program and you thinking about doing it? are always looking for feedback on how to make it better. The program has been successful beyond our expectations. First, it really blew up the as- Our latest iteration was to add a debit card as sumption that there are only a small number a payment option so that we can reload easily of quality researchers able and willing to re- for people who submit bugs regularly. We know that we will always be out-numbered by the bad guys, but we can overcome that by making sure that our systems are up to the challenge. As the number of Facebook users grows year. Self-XSS attacks used social engineer- seemingly exponentially, does your secu- ing to trick users into copying-and-pasting ma- rity team as well? What security-related licious javascript into their browser, thereby problems currently give you the biggest self-propagating the spam and evading our headaches? detection systems. We do continue to grow in size, but we are Before the attacks increased dramatically also constantly challenging ourselves to de- most experts would have doubted that a social velop in such a way that every employee fo- engineering scheme could work at such scale. cused on security has a greater individual im- pact tomorrow than that person did today. We Fortunately, we reacted quickly and have had can do that both by continuing to innovate on success beating it back. In addition to improv- our approaches to security and investing in ing internal detection mechanisms, we have system and infrastructure. worked with browser vendors to make it harder for spammers to take advantage of this We know that we will always be out-numbered vulnerability in the browser, and we have part- by the bad guys, but we can overcome that by nered with external companies to make our making sure that our systems are up to the malicious link detection system more robust. challenge. An example of how things change and new headaches arise the sudden in- We are still battling this but thankfully it is crease in what we call self-XSS during last much less of a headache than it use to be. www.insecuremag.com 23
- 24. I can't remember the last time I saw a bo- These changes instituted granular enforce- gus or information-collecting app being ment which selectively disables an app's abil- pushed onto users by third party develop- ity to propagate through Facebook based on ers, and I recall them being plentiful at one the amount of negative user feedback - so point in time. How did you solve that that an app that has been reported for abusing particular problem? chat will have this feature disabled until the developers have made substantial changes. We have several different teams that work closely together to ensure people have a great In the future, we are moving to more sophisti- experience when connecting with applications cated ranking models where the amount of that leverage our platform. distribution will be a function to the app's qual- ity. Good content will be seen by more people, Major props go to the platform integrity engi- while lower quality or spammy apps will be neers who have been constantly iterating on seen by fewer people or no one. the automated systems that we put in place to secure our Platform. Of particular note were We believe this will reward apps that provide the changes we made last July which made great experiences while minimizing the nega- significant improvement to the enforcement tive impact of poor quality apps. systems so we can identify and disable apps that violate our policies as quickly as possible. Zeljka Zorz is the Managing Editor of (IN)SECURE Magazine and Help Net Security. www.insecuremag.com 24
- 26. The goal of this article is to plant a seed of the idea that shellcode has a place in your defense toolbox. I do not want to teach you how to write shellcode, neither do I want to present a complete anthology of white hat shellcode. What I want is to show a few examples in order to help you be more creative, so that when you are facing a problem in your IT security job, you will also consider shellcode as a potential solution. When a system is attacked, be it by malware Shellcode is a tool, and it can be a solution to or by a human, shellcode is often involved. your problem. Shellcode is executed on the system to change its behavior, so that the system opens What is shellcode? Shellcode is a program, up to the attacker. But why couldn't you use but it has some characteristics that differenti- shellcode to change the behavior of your sys- ate it from applications like .exe files. Shell- tem, too, so that it defends itself against an code is a program that is location-independent attacker? There is no reason why you couldn't and comes as a binary file without any meta- do this. data. As the administrator of the system, you have Example 1: Testing a security setup an advantage over the attacker. While the at- tacker has to rely on exploits that often offer In the first example, we will test our security no guarantee that the shellcode will execute, setup with shellcode. People regularly ask me you, on the other hand, can use reliable meth- for malware so they can test their security ods to inject and execute shellcode. Shellcode setup. First, that is a bad idea, and second, is almost always used in attack scenarios, but you can do without. Why is using malware a it can also be used for defense. bad idea? It is dangerous and not reliable. www.insecuremag.com 26
- 27. Say you use a trojan to test your sandbox. You So how can you reliably test your sandbox notice that your machine is not compromised. without risking infection, or even worse, have But is it because your sandbox contained the malware escape into your corporate network? trojan, or because the trojan failed to execute properly? It might surprise you, but there is a You can do this with shellcode. Here is an ex- lot of unreliable malware out in the wild - mal- ample of simple shellcode that will create a file ware that will crash more often than not, mal- in the directory of your choice (This shellcode ware that will flat-out refuse to run in certain includes a library that is not discussed in this environments, like virtual machines. article): ! segment .text ! ! call geteip ! ! geteip: ! ! pop ebx ! ! ; Setup environment ! ! lea esi, [KERNEL32_FUNCTIONS_TABLE-geteip+ebx] ! ! push esi ! ! lea esi, [KERNEL32_HASHES_TABLE-geteip+ebx] ! ! push esi ! ! push KERNEL32_NUMBER_OF_FUNCTIONS ! ! push KERNEL32_HASH ! ! call LookupFunctions ! ! ; CREATEFILEA and CLOSEHANDLE ! ! push 0x0 ! ! push 0x80 ! ! push 0x2 ! ! push 0x0 ! ! push 0x0 ! ! push 0x0 ! ! lea eax, [FILENAME-geteip+ebx] ! ! push eax ! ! call [KERNEL32_CREATEFILEA-geteip+ebx] ! ! push eax ! ! call [KERNEL32_CLOSEHANDLE-geteip+ebx] ! ! ret Let us assume you sandboxed your preferred This method is very reliable, especially com- browser, Firefox, and now you want to test if pared with the use of real (unreliable) mal- Firefox is restricted from writing to the sys- ware. If you need to test access to other re- tem32 directory. sources, like the registry, you just need to use shellcode that writes to a particular key in the For this, we use shellcode that creates file registry. c:windowssystem32testfile.txt and inject this shellcode in process firefox.exe. Example 2: Enforcing Permanent DEP If the test file was not created in the system32, DEP is an important security feature intro- you have successfully verified that your sand- duced with Windows XP SP3. But not all ap- box prevents Firefox from writing to the sys- plications use DEP, so here is how you can tem32 directory. You can also start Sysinter- enforce it. DEP can be enabled by setting a nal's procmon and look for “access denied” flag in the executable file (the NO_EXECUTE messages from Firefox. This is further proof flag) or by calling WIN32 API function Set- that the shellcode tried to write to system32 ProcessDEPPolicy. but was denied. www.insecuremag.com 27
- 28. SetProcessDEPPolicy has one advantage ProcessDEPPolicy with argument 1 - some- over the NO_EXECUTE flag – it can enable thing you can do with shellcode. Permanent DEP. Once Permanent DEP has been enabled, it cannot be disabled anymore. Shellcode to enable Permanent DEP is rather The only way to enforce Permanent DEP is to simple: it only has to call SetProcessDEP- make the application (like calc.exe) call Set- Policy with argument 1: ; Enable permanent DEP in current process push PROCESS_DEP_ENABLE call [KERNEL32_SETPROCESSDEPPOLICY-geteip+ebx] When you inject this shellcode in your applica- pose you to turn in on again. To get rid of this tion, Permanent DEP will be turned on. But nag screen, I developed a patch: replace byte how can you modify your application so that it sequence calls SetProcessDEPPolicy each time it is 50A16CBF9323FF90C805000039750859 with launched? You can inject the shellcode per- 50A16CBF9323B8020000009039750859 in manently in the application with a PE-file edi- file EScript.api. If you cannot change file tor such as LordPE. EScript.api, you can still change the code di- rectly in memory. First you make a copy of the application (e.g. calc.exe) and you open it with LordPE. Then I have developed shellcode to search and re- you create a new section with the shellcode, place a sequence of bytes in the virtual mem- and make the entrypoint point to the shell- ory of an application. This shellcode can be code. When finished, the shellcode jumps to used to apply the Adobe Reader patch I de- the original entrypoint. You rebuild the PE file scribed. To achieve this, you inject this shell- and save it. code (together with the search and replace byte sequences) in Adobe Reader. When you execute this copy of calc.exe, your shellcode will be the first thing to run. This Another advantage of patching dynamically in shellcode will enable Permanent DEP, and memory with shellcode, is that the patch will then jump to the start of the calculator pro- not be lost when you update your application gram. to a new version (Adobe Reader in our exam- ple). Example 3: Patching an application Example 4: Preventing heap sprays with Patches are changes to the binary code of an shellcode application. They typically fix bugs, security vulnerabilities or change features. But when Shellcode is often used in attacks and mal- you make changes to the files of an applica- ware together with heap sprays: the heap is tion (.exe or .dll), you invalidate the digital sig- filled with shellcode (preceded by a long NOP nature and you are probably breaking the sled), and then the vulnerability is triggered. EULA. EIP jumps to the heap, hits a NOP sled and slides to the shellcode. The shellcode exe- If you want to change an application but are cutes, and typically downloads and installs a not in a position to change the binary files, trojan. shellcode designed to patch in memory can help you. Successful heap sprays can be prevented by pre-allocating memory, so that the heap spray Two years ago I developed a patch to fix an cannot write shellcode to the pre-allocated annoying “feature” of Adobe Reader 9.1. If you memory. If we pre-allocate memory and fill it disabled JavaScript in Adobe Reader, each with our own NOP sled and shellcode, we can time you opened a PDF document with em- intercept the attack and block it. bedded JavaScript, Adobe Reader would re- mind you that JavaScript is disabled and pro- www.insecuremag.com 28
- 29. If you open a PDF document with an util.printf But when we inject our own NOP sled and exploit with Adobe Reader 8.1.2, it will crash shellcode at this address (0x30303030), we because this PDF document contains an ex- achieve code execution. The exploit triggers, ploit that makes EIP jump to 0x30303030 (this but it executes our shellcode, not the shell- might be a few bytes off). Since there is no code of the attacker. code at this address, an exception is gener- ated. This is because we planted our shellcode in But if we inject defensive shellcode that dis- the application's memory before the PDF plays a warning for the user, the user will document was opened and the heap spray know he is being attacked with a malicious executed. PDF document and he will have a chance to act appropriately. The heap spray will fill memory with its attack shellcode, but it cannot overwrite our defense Conclusion shellcode. So when the exploit triggers after the heap spray filled memory, our shellcode Shellcode is just a program, and it is up to the executes instead of the attacker's shellcode. programmer to code the behavior of his pro- gram. We could also use shellcode that suspends the attacked application and warns the user. Shellcode is often programmed to attack, but For user applications, like Adobe Reader, this there is no inherent reason why it cannot be shellcode offers a huge advantage over pro- coded to defend. tection methods that just pre-allocate heap memory and do not inject defensive shellcode. I hope that these four examples give you an idea how to use shellcode to protect your sys- If you just pre-allocate heap memory, the ap- tem. If you want the shellcode of these exam- plication will just crash when it is exploited, ples so that you can test it out yourself, take a and the user will not know what happened. He look at my workshop exercises: could easily assume that Adobe reader just workshop-shellcode.didierstevens.com crashed because of a bug, and try to open the malicious PDF document again. Or even It also contains some tools (for example to in- worse, send it to a colleague so that she can ject shellcode), and I have produced a video try to open the malicious document. for the DEP exercise. Didier Stevens (Microsoft MVP Consumer Security, CISSP, GSSP-C, MCSD .NET, MCITP, MCSE/Security, RHCT, CCNA Security, OSWP) is an IT Security Consultant currently working at a large Belgian financial cor- poration. He is employed by Contraste Europe NV, an IT Consulting Services company (www.contraste.com). You can find his open source security tools on his IT security related blog at blog.DidierStevens.com. www.insecuremag.com 29
- 30. RSA Conference 2012 The Amphion Forum www.rsaconference.com/events/2012/usa www.amphionforum.com Moscone Center, San Francisco Hotel Bayerischer Hof, Munich, Germany 27 February-2 March 2012. 28 March 2012. InfoSec World Conference & Expo Cyber Defence Summit 2012 www.cyberdefencesummit.com www.misti.com/infosecworld Grand Hyatt Hotel, Muscat, Oman Disney's Contemporary Resort, Orlando 2-3 April 2012 2-4 April 2012. HITBSecConf Amsterdam 2012 conference.hitb.org Okura Hotel, Amsterdam, the Netherlands 24-26 April 2012. www.insecuremag.com 30
- 32. Like it or not, enterprise IT organizations are quickly realizing that mobile de- vices are eclipsing PCs and laptops as the devices of choice for employees in the workplace and beyond. Mobile devices such as smartphones and tablets offer incredible power and flexibility in both our business and personal lives, which is leading to great pressure to integrate them within the enterprise. Mobile computing today, when done right, tight economic conditions – they continue to creates an opportunity for workers to be more struggle to address these issues. productive and happy, while also offering a major competitive advantage for the organiza- That is why many large, medium and even tion. However, if not done right, the conse- small corporations are seriously considering a quences can be quite devastating. formalized enterprise Mobile Device Man- agement (MDM) strategy to deal with the pro- This was the main topic of conversation dur- liferation of mobile devices knocking on their ing a recent series of workshops we hosted doors. This means not only using MDM spe- for public and private companies on the im- cific applications and products, but also com- pact the proliferation of mobile devices is hav- bining them with the right mix of policy, proce- ing on enterprises. Interestingly, not a single dures and end user training. organization in attendance had a fully formu- lated Mobile Device Management strategy. Done correctly, enterprise MDM can be a practical approach that first assesses the or- Most, if not all, were still on the ground floor ganization’s challenges, and then evolves with trying to figure out what to do. They realize the dynamic, constantly changing business there are significant risk mitigation issues that needs. By working together and developing a they need to address, but because IT is often pragmatic approach with MDM, an organiza- resource-constrained – especially in today’s tion’s IT and business leaders are much more www.insecuremag.com 32
- 33. likely to embrace today’s mobile world – and • Physical access benefit from it. • Malicious code • Device and application attacks The mobility gold rush • The interception of communications • Insider threats. It's not hard to see why these devices have spurred this gold rush to mobility in the enter- Too often, the decision makers jump right to prise. Sometimes, it comes from the top. The which tools they should buy and want to know board or C-level execs may favor a certain what kinds of bells and whistles are out there device. Meanwhile, employees down the to “lock these things down.” To paraphrase chain are often adopting the latest devices, former U.S. Secretary of Defense Donald platforms and applications much faster than Rumsfeld, when it comes to mobility there are corporate IT departments can react. “known knowns,” “known unknowns,” and “unknown unknowns.” And most organizations Social media is growing as a business appli- don’t know what they don’t know when they cation as well, blurring the work and home look at how they are going to mitigate risk in a environments. Shifting business models also mobile environment. require tech-savvy employees, who are look- ing to connect to the enterprise with their So where do we begin? iPhones, iPads, Androids, Blackberries and other mobile platforms. And along the way, In our opinion, it is always best to use those employee expectations of corporate IT’s abil- tried and true methodologies, or best prac- ity to manage their mobile needs are chang- tices, that security professionals have been ing. preaching for years. But this consumerization of IT also presents An effective approach begins with a risk as- some significant challenges. Of course, the sessment that assesses, evaluates, manages cost of keeping up with the mobile world is and measures each of these security risks. It always a factor. Many companies simply can- is also important that the enterprise IT de- not afford to dedicate in-house resources to partment work with the business units to un- keep up. derstand their mobile requirements. Regardless of whether they do it themselves Without a comprehensive risk assessment, or engage outside expertise, organizations the purchasing decision will more than likely have to address the issue of integrating mo- not reflect the reality of what they are looking bile into existing business processes. This in- to protect. cludes managing the productivity of a remote workforce, determining the reliability of the Before moving forward, organizations need to mobile technologies, and most critical, secu- be able to answer several key questions: rity issues. 1. How many mobile devices are connected to For instance, a recent joint study by Carnegie our network? Mellon's CyLab and McAfee found that almost 2. How do we know how many mobile devices half of users keep sensitive data on their mo- we have? bile devices, including passwords, PIN codes 3. How are these devices connecting? and credit card details. The ramifications of 4. How often are these devices connecting? losing a device or having it compromised can 5. What data and services are these devices be devastating – not only to the individual, but accessing? to the organization whose sensitive data, or at 6. How many of these devices are managed? least the keys to it (passwords, PINs, etc.), 7. How many comply with our corporate poli- may be held within the device. cies? 8. What would be the ramifications if any of For corporate IT, there are five major security these devices are compromised, lost or sto- risks that must be addressed: len? www.insecuremag.com 33
- 34. The matrix applications, and also often leave the device with a standard root password that may grant From here, a matrix of controls can be devel- an attacker administrator-level access to the oped to help enhance the risk mitigation. For device. instance, organizations need to determine what technologies and practices need to be The threat is real. Just last year, a hacker implemented to control different classes of in- pleaded guilty to electronically stealing data formation that mobile devices can access or from more than 100,000 iPad users. Employ- store. They also need to think ahead and ex- ees need to be aware that just because data tend acceptable use policies to all current and is contained in electronic form on their phone, future mobile devices. And all mobile device it is no less confidential and should be treated users must agree to company-defined proc- no less carefully than if it were on paper. And esses and regulations before being granted ideally, this requirement needs to be written access to corporate resources. into their employment contract and reinforced through regularly scheduled training. The next step is to design effective training and communication plans. Although the over- One very simple, yet elegant, solution is to whelming majority of organizations have poli- insist that users turn on the built-in security cies in place for mobile devices, fewer than mechanisms on their devices. Even before one in three employees are aware of their establishing a thorough risk mitigation strat- company’s mobile security policy. egy, organizations can insist that users must install a PIN number on their iPhone if they Consider this: many legit iPhone and iPad plan to use it to access the network. apps leak personal data to third parties. Users don’t help – some still insist on using 0000 or Mobile devices also have location awareness 1234 as their password, making it easy to tools that can help the IT department conduct hack the device. Jailbreaking also puts a remote wipe if the devices are lost or mis- iPhone users at risk for downloading infected placed. Although the overwhelming majority of organizations have policies in place for mobile devices, fewer than one in three employees are aware of their company's mobile security policy. One size does not fit all tralized security app for the variety of phones being released by vendors to the market. It is also important to realize that one size does not fit all when it comes to mobility. In It is most likely that within any corporate envi- fact, the ability to standardize on only one ronment there will never be a "one size fits all" mobile operating platform within the enter- solution. Employees, depending on their job prise is going the way of the rotary dial with requirements, will likely require varying levels the advent of these new devices and tech- of access to data and services. Thus it makes nologies. sense to consider some form of a multi-tiered answer to the problem. One suggestion is to Users are looking to blend their personal de- segment the environment into three basic lev- vices into their work lives, and that means or- els. ganizations need to prioritize which devices they will support and at what levels. For in- Tier One would be executives and others who stance, one issue that will need to be consid- need access to very specific types of highly ered is what images will be displayed on the sensitive information and services, and who various operating systems. And security re- will use the mobile devices as a critical facet mains an ever-present concern, since nobody of their jobs. Tier Two would be those whose has yet been able to develop a universal cen- mobile devices aren’t a necessity for the cor- poration, but can benefit both themselves and www.insecuremag.com 34
- 35. the organization with some access. Finally, The apps story Tier Three would be individuals to whom a minimal level of access (perhaps email only) Over 300,000 mobile applications have been is granted, but strictly as a convenience to the developed in the last three years alone, and individual. users have downloaded 10.9 billion apps over that same time period. Clearly, the prolifera- For this scenario a multi-tiered solution may tion of apps has helped drive the consumeri- look something like this: zation of IT. • Tier One – Users qualify for corporate-liable The challenge is that most apps being pub- devices and are provisioned with Mobile De- lished to the app store are developed vice Management software and business ap- autonomously and don’t have a high level of plications. quality assurance when it comes to security. • Tier Two – Users qualify for personally- Yes, Apple and others will say they provide owned devices that are “lightly” managed and security checks, but those are mostly rudi- supported by the organization. mentary. Once the app is downloaded and in- • Tier Three – Users are free to connect their stalled, it is caveat emptor – back doors and own devices with web-based applications, but coding objection flaws probably haven’t been they don’t qualify for reimbursement of any addressed in today’s app stores. Users are at kind, nor are they supported by the organiza- the mercy of the app, and they aren’t really tion. seeing what’s being communicated and how it’s being communicated across the network. Organizations must also reserve the right to manage any and all mobile devices that re- For instance, a colleague recently accessed a quire access to corporate resources. This well-known airline’s mobile app to check in. management responsibility needs to be inde- He was shocked when he immediately re- pendent of who actually owns the mobile de- ceived a notification from his personal DLP vices, and may require the installation of the (Data Loss Protection) service that his check- firm’s security policies on the mobile devices in request had been blocked due to a violation as a condition of being granted access to cor- in the DLP security policy. It turns out that the porate resources. airline’s app did not enforce the transmission to be encrypted through a secure HTTPS One thing that can be easily overlooked is the connection, but rather simply passed it need to protect the integrity and privacy of through clear text HTTP. So sensitive informa- corporate data by isolating that data inside the tion – including his phone number, house ad- firewall from personal data. This can be done dress and flight information – would all have either by “sandboxing” or taking a virtualized been transmitted had the DLP not stepped in approach to data storage. and prevented it. Of course, the key to this matrix of controls is At the enterprise level, it’s critical to under- enforcement of strong security policies that stand which apps are mission-essential and prevent data security breaches. These polices standardize mobile users on those apps. should address encryption, PINs and pass- Those can be published for download only words, auto-lock capabilities, location track- while a user is on the corporate image and ing, remote wipes, disabling non-approved connected to the network. Organizations applications, features and functionality, and should also examine their internal app store policy removal prevention. and focus on setting restrictions on apps that are not business-essential. Once all of these controls are in place, or- ganizations can prioritize and determine how Location. Location. Location. and when users will be provisioned with enterprise-class applications, and address The big problem with mobility is that organiza- ramifications for non-compliance with these tions don’t know where people are going to be controls. Enterprise MDM risk mitigation poli- when they try to access the network with their cies should also be reviewed at least yearly. devices. www.insecuremag.com 35
- 36. Whether they are sitting in a coffee shop or at database be used to help authenticate users a desk in their home or at work, users are at corporate? more and more frequently looking to access their network through their mobile device than Organizations need to constantly keep their through a PC or desktop terminal. guard up when it comes to mobility. Employ- ees will continue to adopt the latest devices, So part of the risk assessment also needs to platforms and applications much faster than examine how users plan to connect to the corporate IT departments can react. network, where they will be using it, and what access points are acceptable. For instance, However, by leveraging an effective security- what will the corporate profile look like if a centric approach to risk mitigation, organiza- user is connecting through a hotspot at the tions today can understand where the security airport, as compared to connecting via a wire- risks lie, whether their operating systems are less modem within the company’s headquar- secure, if the mobile devices being used have ters? It will also be important to decide how to adequate security features, and how to battle authenticate to the access point itself. Will it malware-laden code in applications. And that be through a shared key, or will a third-part will let them – and their mobile users – rest easier. Elvis Gregov and Keith Olsen are Solution Architects with Forsythe Technology Canada, Inc. (www.forsythe.com/na/aboutus/forsythecanada), an IT infrastructure integrator headquartered in Toronto, with offices in Edmonton, Vancouver, Winnipeg and Calgary. www.insecuremag.com 36
- 37. HD is Chief Security Officer at Rapid7 and Chief Architect of Metasploit, the popular open-source penetration testing platform. It's been a long road for Metasploit. What tinued success, both commercially and as an began as a personal project is now a major open source project. name in the security industry. How has the project evolved since it was acquired by Metasploit has always been a platform for Rapid7 and, overall, how has your profes- building security tools, testing out new ideas, sional life changed? and sharing those with a wider audience. Our recent move to GitHub and the increasing size After almost nine years, Metasploit is still an of the community continues to prove that open incredibly fun project to work on. The acquisi- collaboration is the best way to raise the bar tion by Rapid7, the development of commer- within information security. cial editions, and the dedicated development team have increased the project's capacity to Through 2011, the project averaged more grow and provide bigger and better things for than one new module a day, with many of security professionals. those coming straight from the community. A large portion of the team's time is spent work- The "corporate" environment and the expan- ing with contributors, tuning submitted code, sion of our open source user base (nearly and testing that code prior to rolling it into the ~150,000 now) have not changed the soul of master repository. the project or the personal nature of contribu- tions. Rapid7 is a strong supporter of open The Rapid7 team also handles things like source, community collaboration, and just as quality assurance, core library changes, data- importantly, common sense vulnerability dis- base architecture, and maintaining the build closure. These traits are why Rapid7 was a and installer environments for the open source good fit in 2009 and the driver behind our con- code base. www.insecuremag.com 37
- 38. The focus on community submissions has tures that would have been time-prohibitive to changed how we manage the project and do otherwise. This is a somewhat dated ex- where we allocate funds within the Rapid7 ample, but it was a case of a one-off contribu- team. Our biggest revelation was that for tion changing the direction of the project in a some roles, we are better off focusing on the way that we wouldn't have done on our own. community submissions than trying to provide everything to everyone solely on our own. Staying involved with the security community is great for the open source project, but it also This was an ego check in some ways, but it helps us align commercial development with opened the door to faster progress and unex- the challenges our customers may yet to hit. pected innovation. One example is the Rail- gun functionality within the Meterpreter exten- In many cases, new modules or features will sion. land in the Metasploit Framework trunk, only to become critical features to our enterprise Railgun provides a generic API for calling arbi- customers at a later time. A recent example trary Win32 methods and returning the results includes a remote exploit for LifeSize video to the user. This code was dropped, anony- conferencing systems. On its own, this mod- mously, to the framework mailing list without a ule did not appear to be that noteworthy - single follow-up from the original author. Since most of the Rapid7 team (not to mention our then, Railgun has been expanded, improved, customers) had little experience with these and is now a driving force behind many of our systems and were not aware of their deploy- post-exploitation modules and enables fea- ment scale or patch cycle. STAYING INVOLVED WITH THE SECURITY COMMUNITY IS GREAT FOR THE OPEN SOURCE PROJECT Fast forward two months to the introduction of To this day, I stay actively involved in both the the H.323 scanning module that was used to open source and commercial product devel- produce our recent "board room hacking" re- opment, as well as the media and social net- search, and the LifeSize module becomes working aspects of the project. In addition, I immediately applicable. Not only does it in- spend a lot of time on the phone with custom- crease awareness of "system" exploits for ers, handling support cases, working with in- video conferencing equipment, but the H.323 tegrators, and expanding the development survey results allowed us to see exactly how team through new hires. often these devices were patched and what percent of internet-exposed systems may be Having the perspective of a developer as well vulnerable. as sales, marketing, and support part of the business helps keep the project and our The resulting news articles and blog posts re- commercial products on the right path. sulted in many of our customers identifying these devices in their environment, proving The biggest change has been handing off that they were indeed a security risk using the parts of the project to my co-workers and LifeSize exploit module, and being granted the leaders in the community. appropriate resources to fix the problem. We have some amazing contributors (not to This was a great of example of a module mention employees) and it has been liberating submitted by another security company (Se- to share the load with this group of talented cureState) being combined with work from individuals focused on a common goal. Rapid7 to identify and validate a real-world risk that many organizations had ignored to that point. www.insecuremag.com 38
- 39. What major challenges did you face devel- ment in the early aughts had contributed to a oping Metasploit on your own? commercial environment where exploits were not just desired, but actually valuable. The I was the founder of the project, but between huge (at the time, this meant ~150) number of 2003 and 2009 a handful of other developers exploits within the Metasploit Framework and were involved. In the early days, the core the permissive BSD license meant that many team consisted of myself, Matt Miller (skape), organizations took a second look at the pro- and spoonm. By 2007, I was the last man ject and started using the framework for both standing and this lead to the expansion of the internal and external work. team to include James Lee (egypt) and a number of new frequent contributors. Until re- By the beginning of 2009, it was hard to find cently, I was the only developer handling the any organization providing network security release process, packaging, and testing, as services that was not a Metasploit user in well most of the project hosting, server ad- some form. The day I realized we had crossed ministration, and legal paperwork. the tipping point was when I witnessed a sales associate demonstrating an IPS product using Early on, Metasploit faced two major non- the Metasploit Framework on the exposition technical challenges. floor of the RSA conference. The first was convincing the security commu- Throughout all of this, the community contin- nity to use a brand new toolkit that aimed to ued to expand. Releases went from taking just replace one-off tools and exploits that had be- weekends to almost an entire week of free come second nature. We managed to solve time. Myself as well as the other project mem- this through brute force development and bers all had full-time jobs, many of them within time. After years of off-and-on ridicule and fre- startup companies that demanded long hours quent releases, many professionals finally as well. The discussion with Rapid7 and the tried the software and understood the point of eventual acquisition could not have come at a our efforts. It took a couple more years before better time for the project or the open source the work going into Metasploit was recognized community. as driving innovation, not just a replacement for existing solutions. Perseverance and con- What advice can you offer to other open tinuous improvement was the only way to win source security software developers? this battle. There a few things I recommend: The second challenge was convincing the rest of the world that Metasploit was not designed 1. Assign developer copyrights to a legal en- to help under-skilled hackers break into corpo- tity, such as a LLC or other limited partner- rate environments. The project was started at ship. This makes tracking expenses, register- a time when vulnerability disclosure and ex- ing domains, filing for trademarks, and han- ploit release was under attack by government dling copyright violations and other nastiness bodies, security professionals, and the anti- much simpler. In the case of the Metasploit disclosure underground. Framework, the core developers assigned their rights to an LLC, which in turn provided Each of these groups had their reasons for the same developers with an unlimited license opposing an increase in open security infor- to use and repurpose the codebase. mation, and the most telling trait was their shared opposition to projects like Metasploit. 2. Choose an open source license based on Over the next 5 years, I spent a lot of time de- the goals you are trying to solve, not based on fending the policies of the project, risking my peer pressure or unjustified paranoia about own livelihood, responding to complaints, and corporate "abuse". At the end of the day, if generally fighting back against the perception someone wants to steal your code, they will, that exploit tools did not improve security. and in some cases just translate it to another language to build a competing product. This In 2008, the tide had finally turned. The "chill- happens, so plan for it, stick by your goals and ing" of open research and exploit develop- use copyrights, trademarks, and other legal www.insecuremag.com 39
- 40. mechanisms to protect your brand where nec- future efforts. This isn't to say that there is essary. anything inherently wrong with GPL, just that it is not always the best default, especially for 3. Choose an open source license that will not new projects. If you are intensely concerned cause a mountain of future work if you decide with companies using your code without to change it later or build a commercial prod- authorization, your best bet is to hire a lawyer uct. BSD-style licenses are a great choice. If to draft a commercial-style End User License you go with a license like GPLv2 or GPLv3, Agreement. you may need to get copyright assignments from every single individual who contributed This is definitely not open source, but it can code to the project to effectively use that code buy time while you work out exactly how you within a commercial product. want to license the project going forward. Two versions of the Metasploit Framework were The viral nature of GPL can seem like a great released under a EULA-style license (v3.0 defense to commercial abuse, but that sword and v3.1) before the project converted back to cuts both ways, and it can easily hobble your a permissive BSD model. CONTRIBUTORS WHO SUBMIT CODE YOU DON'T LIKE TODAY, CAN GROW INTO CORE DEVELOPERS OVER TIME 4. Identify a small number of goals that set grow into core developers over time. Over the your project apart from what is already avail- history of the Metasploit project, nearly all the able. Stay laser-focused on those goals until major contributors started off with a couple your project is the best fit for solving that type patches and a longer discussion about coding of problem. If you decide to expand the scope guidelines and design goals. If you can find of the project, do so intentionally and commit the time, make public your design philosophy to continuing in that direction for some time. and formatting standards. Users don't like features that work poorly and As the project grows, document the process they definitely don't like to see features disap- for contributing and set expectations about pear due to a change of mind later on. Keep- how long it takes to respond to submissions. ing focused on differentiators also helps drive Stay involved with the communities that use awareness of the project and keep it top-of- your project and recognize contributors who mind for anyone trying to solve this problem. submit code, ideas, or just help answer ques- tions for other users. In the open source If you are interested in merging your code into world, the only real form of compensation is a larger project or in being acquired by a recognition. commercial entity, doing one thing really well makes the economics simple. You contribute a Based on the feedback you get from your working implementation and deep knowledge extensive user base, what are the most re- of one area that would be more expensive for quested Metasploit features yet to be im- them to do on their own. plemented? 5. Stay friendly, stay humble, and appreciate To go by sheer volume, the two most re- the work being done by contributors, even if quested features are the magic "hack every- the code itself makes you gag. There are a thing" and "evade my antivirus" commands. A number of security projects where the most lot of time is spent in the community - whether common response to "do you know about X?" its Twitter, IRC, mailing lists, or discussion fo- is a statement about how awesome one of rums - setting expectations for what the pro- their developers is and a story of how they ject can do and what the scope of our devel- helped them figure something out. Contribu- opment is. tors who submit code you don't like today, can www.insecuremag.com 40
- 41. There are tons of great ideas submitted by the Most of the common requests boil down to community, but we have to stay focused on current design limitations (more consistency what we do best (providing a platform for se- between session types) or automating a chain curity tools and exploits) to continue pushing of actions that would better fit into a plugin, a the project forward. In the case of a request resource script, or outside of the core frame- not matching our acceptance guidelines work. We still see a lot of requests for addi- (mass-automation modules, modules that tional payload capabilities, whether its new don't meet our API requirements, etc.) we APIs for Meterpreter, expanded platform sup- recommend that users simply fork the public port, or stealth and evasion features. repository and maintain it as a separate Strangely enough, what we don't see that of- branch. ten are requests for additional exploits. METASPLOIT HAS EVOLVED FROM AN OPEN SOURCE FRAMEWORK THAT FOCUSED ALMOST ENTIRELY ON EXPLOITS TO A GENERAL-PURPOSE SECURITY PLATFORM WITH A MULTITUDE OF OPEN SOURCE AND COMMERCIAL OPTIONS. THIS TREND WILL CONTINUE What is your vision for Metasploit in the increasing network sizes and exploits. The next five years? soft limit for concurrent sessions (open con- nections to compromised systems) is a bit Metasploit has evolved from an open source over 1,000 today, per process, and this will framework that focused almost entirely on ex- likely need to increase. ploits to a general-purpose security platform with a multitude of open source and commer- One area that we have touched on, but not cial options. This trend will continue. As secu- really dove into, is wireless protocol testing. rity testing continues to move away from tradi- Metasploit includes a number of modules for tional exploitation methods, the open source 802.11-based vulnerabilities as well as DECT core will evolve to support additional types of station scanning and call monitoring, but the attacks, sessions, and data management. expansion of WiMax and new RF protocols will require new security tools to adequately The networking layer will continue to expand assess their deployment. Metasploit may be to support even more protocols and evasion the right tool for the job and we will focus de- methods. As much as we would like to avoid velopment efforts accordingly. it, baking in additional payload-level evasion, specifically anti-virus systems, will become In the end, it really depends on where the big- even more critical as user-assisted code exe- gest risks are and what our open source users cution becomes the predominant vector for and commercial customers need to be suc- remote exploitation. cessful. Metasploit has proven to be incredibly adaptive over the years, supporting everything Over the last two years, the database back- from remote kernel exploits to serial-based end has gone through a number of major wardialing and VoIP audio codecs. changes, and now ships enabled by default. This provides a new level of data persistence, Metasploit encompasses the open source storage, and automation capabilities through framework, the free Community Edition of our the PostgreSQL backend. Database architec- commercial platform, and the flagship product, ture and data management will continue to Metasploit Pro. We would love to apply the play an important role in the design and func- same modular automation and chaining tech- tionality of the framework. niques used in penetration testing to other ar- eas of IT and Operations. How we get there Scalability is another area where we have al- will depend a lot on where we can help and ready made major improvements, but will whether we have the right capabilities to solve need to continue growing to support the ever- the problem at hand. www.insecuremag.com 41
- 42. TARGETED ATTACKS WILL ALWAYS WORK FOR THE SAME REASON THAT CON ARTISTS STILL SUCCEED AT STEALING MONEY We see the most dangerous elements in would have been just fine using WPA2 alone. the threat landscape moving toward highly As technology continues forward, the folks targeted attacks. What type of long-term designing new protocols and products will impact will this have on the security tools make mistakes, and just like before these will we use today? Are we looking at a stronger introduce introduce security flaws that can and artificial intelligence (AI) component in fu- will be exploited by malicious intruders. ture computer security products? Metasploit will stand ready to help our users The more things change, the more they stay and customers identify these risks and dem- the same. Prior to the glut of buffer overflow onstrate their impact. and memory corruption vulnerabilities, hack- ers still hacked, and most of this was focused In August 2011, Rapid7 committed on design flaws, logic issues, weak creden- $100,000 to open source projects. How did tials, and exploiting the human behind the this idea come about? What projects were terminal on the other side. chosen and how are they developing? Targeted attacks will always work for the same The Magnificent7 project is an idea that rose reason that con artists still succeed at stealing from the Rapid7 executive team as a straight- money. New protocols still ship with incredibly forward way to contribute back to open source poor security measures and even mature while driving progress in the area of informa- technology introduces new flaws in the form of tion security. features. This $100,000 budget will be split across 7 A great example of this is the recent 802.11 projects over the course 2012, focusing on WPS flaw (WiFi Protected Setup). WiFi secu- specific milestones that the project creators rity had finally reached the point where WPA2 identified as being the biggest roadblocks to with a strong password was good enough for reaching their goals. The first round of pro- many organizations. The introduction of WPS jects will be announced at the RSA 2012 con- as a simple, secure way to access a network ference in San Francisco. backfired by exposing millions of routers that Mirko Zorz is the Editor in Chief of (IN)SECURE Magazine and Help Net Security. www.insecuremag.com 42
- 43. Defensive search-and-destroy According to The Daily Yomiuri, the virus is not only particularly effective when it comes to "virus" delivered to Japanese identifying the computers participating in government DDoS attacks, but also the computers that control these botnets. Unfortunately, it is supposedly less effective when identifying sources of attacks aimed at stealing information from targeted systems. Upon delivery, the cyber weapon and the monitoring system were tested by the ministry in a "closed network environment", and have It took three years and 178.5 million yen obviously proven to have been worth the (around $2.3 m) to develop a defensive cyber money invested in them. weapon that can track down the sources of cyber attacks and disable them, but Fujitsu Citing client confidentiality as the reason, apparently did it. Fujitsu had so far declined to comment on the program. But even if the "virus" is as good as Contracted in 2008 by the Japanese Defense it seems, the question about whether it can be Ministry's Technical Research and freely used by anyone is open for debate as Development Institute, the company was the Japanese Parliament has recently made charged with producing the aforementioned malware production and distribution a criminal computer "virus" and a separate system offense. capable of monitoring and analyzing cyber attacks. www.insecuremag.com 43
- 44. Recycled cybercrime tactics adapted Software. “Most cyber-attacks at any given time rely on old techniques deployed with a to conceal fraud new disguise. The reason we see them again and again is quite simply because they work, and we anticipate 2012 to bring many fresh takes on old scams.” In a continuing trend highlighted in the last VIPRE Report, bank related phishing is increasingly becoming a common threat. Barclays customers received messages from a free Yahoo email address claiming that their account had been suspended due to incorrect login attempts. GFI Software released its VIPRE Report for December 2011, a collection of the most The phishers employed scare tactics by prevalent threat detections encountered insisting information had to be provided to during the month. reactivate the account within a certain amount of time. Once the victim’s identity was Phishing campaigns once again proved to be submitted, they were redirected to the official among the most significant threats, with Barclays website in order to further mask the scammers targeting Chase and Barclays crime. Chase clients were targeted by a customers, as well as launching malware similar phishing campaign last month as well. attacks against Amazon shoppers expecting holiday packages. Another familiar cybercrime tactic that continued to gain momentum in December “The threats we uncovered last month was scareware—fake antivirus software and illustrate the consistent reuse of tried-and-true system utility programs—that warn infected attack methods slightly modified to target new users of completely false threats to their groups of potential victims,” said Christopher computers. Boyd, senior threat researcher at GFI The anatomy of the Gameover Zeus the ZeuS 1.4/2.0 beta program, but it was dropped from the final 2.0.x release because variant lack of demand among ZeuS customers in the face of significant coding and testing time. It was put back in as a feature during the recent, ongoing 2.2/3.0 beta program. The “Gameover” version of Zeus also supports the use of complex web injections that allow the attacker to perform Man-in-the- Browser (MITB) attacks to bypass multi-factor authentication mechanisms. The ZeuS author has also rolled a Distributed Denial of Service The “Gameover” malware is a relatively new, (DDoS) component into the Gameover "private" version of ZeuS. Support for the bundle. distributed command and control (C2) tools, integrated into the ZeuS botnet, were Gameover has been used in this way. First, implemented at the request of one of the financial institutions were targeted with DDoS "private" clients of the ZeuS author. attacks against their online banking websites. These attacks were timed to coincide shortly Distributed C2 is a feature which was after accounts at the targeted financial originally considered by the malware author in institution had fraud committed against them. www.insecuremag.com 44
- 45. Chinese using malware to attack US readers running ActivClient - the client application of ActivIdentity, whose smart cards DoD smart card security are standardized at the DoD and a number of other US government agencies,” he said. “The smart cards are an important facet of security for the DoD – which manages the three main branches of the military in the US, the Departments of the Army, the Navy and the Air Force – and use the cards as a standard means of identifying active duty AlienVault found evidence of Chinese- military staff, selected reserve personnel, originated attacks against the US government civilian employees, and eligible contractor agencies including the US DoD, which use a staff,” he added. new strain of the Sykipot malware to compromise DoD smart cards. So far, Blasco and his team have seen attacks that compromise smart card readers running One of the original versions of Sykipot was a Windows Native x509 software, which is Trojan horse application that opened a reportedly in commonplace use amongst a backdoor into the infected PCs. According to number of US government and allied Jaime Blasco, AlienVault’s Lab manager, this agencies. latest generation of diversified attacks may have been occurring as far back as March of This new strain, he says, is thought to have last year, if not longer. originated from the same Chinese authors that created a version of Sykipot late last year that “This is the first report of Sykipot being used piped out a variety of spammed messages to compromise smart cards, and this latest with the lure of information on the next- version of the malware has been designed generation unmanned 'drones' developed by specifically to take advantage of smart card the United States Air Force. Identities of likely Koobface gang In 2009, independent researcher Jan Drömer mounted his own investigation. Starting with members revealed crucial information gleaned from one of the Koobface C&C servers and searching for links to it on the Internet - IP addresses, domain registration information, underground and legitimate forum posts, social network accounts and more - he made a beeline to the aforementioned group of individuals. According to him, there is a variety of reasons First, details about a likely member of the "Ali behind the success of the Koobaface gang: Baba & 4" group (as they dubbed themselves) they misused powerful online services to were made public by researcher Dancho spread the worm, didn't overdo on the size of Danchev and, as the story begun to unfold, the botnet, haven't aimed at making the worm security firm Sophos and the NYT revealed perfect but invested just enough revenue to the names of the five individuals thought to be earn more than enough money, and have part of the Koobface gang. operated in countries whose law enforcement agencies haven't a good record when it Facebook started an investigation into the comes to cooperating with their US and gang shortly after the Koobface worm first European counterparts. began to spread on the social network in 2008, and it took them only weeks to link the attacks to the suspects. www.insecuremag.com 45
- 46. "Frankenmalware" active in the wild compromised by a worm, the virus will infect the exe files on that PC - including the worm. When the worm spreads, it will carry the virus with it. Although this happens unintentionally, the combined features from both pieces of malware will inflict a lot more damage than the creators of either piece of malware intended." To explain how the symbiosis works, she shares the example of the Virtob virus/ Rimecud worm "collaboration". The Rimecud worm spreads via file-sharing If you're not careful and you don't use anti- apps, USB devices, Microsoft MSN malware software, you might end up with Messenger and locally mapped network various viruses, Trojans and worms on your drives. Besides that, it also steals passwords computer. But, according to Bitdefender by injecting itself into the explorer.exe researchers, you might even get saddled with process, opens a backdoor that will allow it to a hybrid or two of this different types of download additional malware from a C&C malware. server and - if the computer has remote control software installed - allows cyber The researchers have dubbed these hybrids criminals to access it and control it. "frankenmalware", and out of some 10 million detected and analyzed malicious files, they As it turns out, Bitdefender has recently begun identified over 40,000 of these "malware spotting the Virtob virus attached to the sandwiches". aforementioned worm. The virus - which also opens a backdoor, contacts IRC C&C servers, "A virus infects executable files; and a worm is modifies a host of files - infects executable an executable file," explained Loredana files and, as the worm itself is an executable, Botezatu. "If the virus reaches a PC already it is also likely to be infected. A peek into the Sykipot campaigns "Each campaign is marked with a unique identifier comprised of a few letters followed by a date hard-coded within the Sykipot Trojan itself. In some cases the keyword preceding the numbers is the sub-domain's folder name on the Web server being used," they shared. "These campaign markers allow the attackers to correlate different attacks on different organizations and industries." Symantec researchers have recently The location of the server (Beijing), those of discovered and managed to take a peek into a attackers contacting it (Zhejiang province) and staging server for the Sykipot campaigns, Chinese words contained in path and some which was also occasionally used as a C&C file names seem to validate the theory that server for delivering instructions to the Chinese hackers are behind the attacks. malware installed on the compromised computers. The researchers found over a hundred of of malicious files sent as attachments to the In it they discovered many things that gave targets. They were mostly specially crafted them insight into how the campaigns are PDF files that would drop the Trojan onto the differentiated and waged. targeted system once they were run. www.insecuremag.com 46
- 48. The last decade has seen a major shift in the notion of perimeter exposure. Firewalls are robust and modern operating systems are increasingly hardened by default. The new perimeter is the web application, un- Vega from source can obtain the source code seen by the firewall, universally exposed, rich from our repository, hosted at with complex functionality, often consisting of github.com/subgraph/Vega. Vega can be a mix of custom and third party code. compiled by simply running “ant” (note that the Shorter development cycles for web applica- build script will download dependencies from tions means that codebases change with a a Subgraph server). To build the newest ver- higher frequency. Meanwhile, developers are sion of Vega: commonly inexperienced and lack tools to help. While many open source tools exist, $ git clone most of them can be tricky to use by non- git://github.com/subgraph/Vega.git security professionals. These are some of the $ cd Vega reasons that make managing the security of $ git checkout develop web applications a challenging problem. $ ant Vega is a new open source platform for testing After a successful build, the binaries will be in: the security of web applications developed by $ ls build/stage/I.VegaBuild/ Subgraph (www.subgraph.com) and released VegaBuild-linux.gtk.x86.zip under the Eclipse Public License (EPL) 1.0. VegaBuild-macosx.cocoa.x86_64.zip Vega is written in Java, is GUI-based, and compilelogs/ runs on OS X, Linux, and Windows. VegaBuild-linux.gtk.x86_64.zip VegaBuild-win32.win32.x86.zip The 1.0 beta was included in BT5R1 and later. VegaBuild-macosx.cocoa.x86.zip Users interested in building bleeding-edge VegaBuild-win32.win32.x86_64.zip www.insecuremag.com 48
- 49. Vega includes a crawler for automated vulner- two corresponding "perspectives" (arrange- ability scanning, as well as an intercepting ments) of UI components known as "views". proxy for manual hacking. While Vega in- This terminology may be familiar to users of cludes a set of built-in vulnerability checks, the the Eclipse IDE. In the current version of real power of Vega comes from its extensibil- Vega, there is a scanner perspective and a ity: there is a built-in Javascript interpreter for proxy perspective. Views within each perspec- creating custom modules using a rich API. In tive can be moved around and re-sized. Se- this article we will describe all of the features lecting "reset perspective" from the Window of Vega and walk through simple examples of pull-down menu in the Vega toolbar will reset custom module development for each of the the perspective to its default arrangement, two types. should the user ever want to return to the ini- tial configuration. Vega is based on Equinox OSGi and Eclipse RCP, the modular framework and UI toolkit Vega saves scan/proxy data and configuration underlying the Eclipse IDE. Vega also incorpo- settings in a data store known as a "work- rates the Mozilla Rhino Javascript interpreter, space". The workspace can be cleared by se- Apache HC, jsoup and db4o. Development lecting "Reset Workspace" from the "File" has continued since the 1.0 beta release on toolbar menu. The workspace can be backed July 1, 2011. In this article we will demonstrate up or transferred by locating or moving the some new features in the pre-1.0 version "model.db" file. On Linux systems, this file will available from our repository at github. be in a sub-directory within ~/.vega/workspaces. Basics The scanner UI is the default perspective, The two core modes of operation for Vega are presented when Vega is run for the first time. as an automated scanner and as an intercept- We will therefore describe the scanner first. ing proxy. The Vega user interface is split into www.insecuremag.com 49
- 50. Scanner Just below the website view is the Scan Alerts view. This area is where alerts generated by The Vega automated scanner is a vulnerability modules during scans or proxy usage will be assessment tool that crawls web applications, listed. Each individual scan that has been run, actively and passively probing for known and known as a scan instance, will have its own unknown vulnerabilities using customizable tree of generated alerts listed in order of se- Javascript modules and Java probes. verity, grouped by type. The proxy has its own tree for alerts. Like modules, alerts are entirely The scanner interface has four sections. In the customizable. Alerts rendered by Vega contain top right is the website view, where a tree of static content from XML template files and dy- web paths seen and visited by Vega will be namic content from the modules that generate rendered. The website view presents data in them. Users can edit existing alerts or create hierarchical order: for each website, the arrow their own new ones very easily. icon to the left of the hostname can be clicked to expand it into a list of paths discovered on Vega has a general console for text output. the server. Sites and paths that are grayed out This is where output is printed when the de- indicate that Vega has seen but not accessed bug setting is enabled for the scanner. The them. For example, this can occur when Vega modules also send their debug output to the crawls a website and discovers a link to a host console. or path outside of scan scope. There is a but- ton above the web view to remove these un- The console is accessible by clicking the con- visited paths from the list. sole fastview icon in the bottom left corner of the Vega UI. The fastview icon will blink with a The user can also select a website or path warning indicator when there is pending out- and instruct Vega to begin scanning from that put. point. www.insecuremag.com 50
- 51. In its most basic usage, the Vega automated The base URI field is the starting point of the scanner crawls a website, running vulnerabil- crawler: for example, the user could input ity detection modules written in Javascript. To www.example.com. The identities field is for start such a scan, the user can click the "Start assigning a set of credentials that Vega will New Scan" target icon in the top right corner use during the scan. This is used when the of the scanner perspective. Doing so will application being scanned requires authenti- prompt a pop-up dialog with some scan pa- cation. rameters to be set by the user. Identities Macros Identities are the general facility provided by Vega allows for sequences of requests to be Vega to store sets of credentials. Identities recorded and replayed before the start of a can be created for various authentication scan. These sequences are known as "mac- mechanisms, including basic, digest, and ros". This function is useful for automatically NTLM. replaying login form submissions to establish For form-based authentication, it is possible to an authenticated session for the scanner. A bind an identity to a macro, which instructs macro can be created one of two ways: by Vega to authenticate using a recorded set of clicking "Create Macro" button in the macro requests. Macros can be created before an view, at the bottom of the scanner perspec- identity is created, or during the process. tive, or from within the identity creation dialog. www.insecuremag.com 51
- 52. To create a macro, the user should first per- fault, cookies are preserved. The user may form the requests through the proxy with a also add or modify HTTP header fields in the HTTP client. These requests can then be se- macro requests. The macros are given names lected from a request table of recorded proxy by the user and can be saved. requests within the macro interface. By de- www.insecuremag.com 52
- 53. Below the identities selector is a tree of mod- spective. The progress bar will adjust in size ules that can be selected or deselected for in- as the scanner discovers more of the applica- clusion in the scan. Modules typically repre- tion during its recursive crawl. Vega performs sent individual vulnerability checks and each various tests on each accessed path, trying to module is a single Javascript file in the Vega determine if it is a file or directory. Vega also modules directory (scripts/scanner/modules/ does 404 analysis to fingerprint the server re- injection or response). Vega supports two sponse in cases where a path that does not types of modules: "basic" (active) and "re- exist is accessed. As Vega identifies vulner- sponse processing" (passive). abilities, the summary table in the "Scan Info" central view will be populated and correspond- The basic modules, which are also known as ing alerts will be added to the Alerts view. injection modules, run on each injection point When an alert is selected for review from identified by the crawler: all files, directories, within the Alerts view, it will be rendered in the and parameters. The basic modules do the Scan Info central view. The alert contents will fuzzing: they generate multiple new requests be described in more detail below. and process the responses using a callback Clicking on the top level node of the scan in- function that they register. The response stance in the Alerts view will switch the con- processing modules run on all HTTP re- tents of the central view back to the scan sponses received by Vega, grepping for pat- summary. terns corresponding to security vulnerabilities. Both modules can generate alerts and store/ Scan alerts retrieve data in an internal database. Adding a module to the list is as easy as dropping a file Vega modules generate alerts when they de- in the right directory. Modules can also be ed- tect possible vulnerabilities. It is up to the ited and reloaded without restarting Vega. module developer to decide when and why to generate an alert, and which alert should be To continue the setup of a new scan, the user generated. The alerts are generated when the can click "Next" to continue to the second module invokes a specific method in the mod- step, or skip it. For the purpose of this tutorial ule's context, specifying the XML template to we will proceed by clicking "Next". The second use for the alert. It also passes parameters step allows for the user to input a custom such as a HTTP request and response, some cookie value, as well as any paths that they relevant content, a link to the vulnerable re- do not wish the crawler to access. This is use- source on the target server, and a unique key ful if there are logout links in the application for the alert to prevent duplicate instances. that will clear the authenticated session if they Vega assembles the final alert using static are accessed. Clicking "finish" will start the content from the XML template file and dy- scan. namic content from the module. Once the scan has started, the progress is in- Example - Module "vinfo-1918.js" invoking an dicated in the central view of the scanner per- alert (the first parameter is the XML template): [..] ctx.alert("vinfo-1918", request, response, { // XML file, request object, response object output: result.join(" "), // output included in alert resource: request.requestLine.uri, // vulnerable resource (link) key: "vinfo-1918" + request.requestLine.uri + result.join(" ") // unique key }); [..] Example - XML template "vinfo-1918.xml": <?xml version="1.0" encoding="UTF-8"?> www.insecuremag.com 53
- 54. <alert> <title>Internal Addresses Found</title> <class>Information</class> <severity>Low</severity> <impact>May reveal internal network structure to outside attackers.</ impact> <impact>Internal IP addresses that have been disclosed could be used as targets in otherwise blind attacks.</impact> <discussion> Vega has discovered references to internal hosts or networks in publicly accessible content. These addresses may reveal information to an attacker about the internal network structure, increasing the likelihood of success for blind attacks involving other vulnerabilities. </discussion> [..] These will be discussed in further detail later in this article. Request viewer requests saved by alerts are logged). The user is advised to review the settings in the Most Vega modules save a specific request preferences menu option of the "Window" and response pair for inclusion in an alert. menu bar. This is useful for users who want to verify that the possible vulnerability is present or investi- Proxy gate it further. This request and response pair is made available to the user in a link within The Vega intercepting proxy is meant for use the alert. Clicking on the "Request" link will with a HTTP client, such as a web browser, open up the request viewer fast view, with the and allows for close observation and manipu- request and response in message viewers be- lation of client-server interaction. When the low a request table. The user can then inspect proxy is enabled, Vega opens a listening TCP the full HTTP request and response associ- port on a configurable port number (default is ated with the alert. Right clicking on the re- 8888). HTTP clients can be configured to use quest log entry in the request table above the the proxy on this port. Firefox is a good choice message viewers allows for it to be selected of browser for use with the proxy because it for replay. maintains its own proxy settings, distinct from system-wide proxy settings. If a request is selected for replay, a request editor tab will open in the Scan Info view. The The Vega intercepting proxy can be accessed user can then modify the request and click the by clicking the proxy button at the top right, "play" button above the editor region to trans- which will open the proxy perspective. The mit it to the server. The server response will proxy can be enabled by clicking the "Play" be rendered in the message viewer below. button in the top left corner of the proxy inter- The user may modify and send as many re- face, and can be stopped by clicking the stop quests as they like from this view and then icon. A status indicator in the bottom right cor- close it when they are finished. ner of the Vega UI will indicate that the proxy is listening. The proxy perspective is com- Scanner preferences prised of three major views: the website view, the request table, and the HTTP message The scanner preferences allow for resource viewers. The website view in the proxy is iden- limits to be set. This can constrain the scope tical to the small version embedded in the top of scans. One useful debugging feature during left corner of the scanner perspective. The re- module development is the logging of all quest table is a list of all requests saved by scanner requests (by default, only scanner Vega. www.insecuremag.com 54
- 55. Request table tional request tables to which other filters can be applied by clicking on the "Open New Re- By default, all requests and responses that quest Viewer" icon above the request list. pass through the proxy are stored in Vega's underlying database. The contents of this da- Right-clicking a row in the request list will tabase can be viewed in request tables, an bring up options such as replaying the request arbitrary number of which can be created, and tagging it. Requests can be tagged and each with specific filters applied. The request assigned highlighting colors to distinguish list can be filtered by criteria such as regexp them if they are of some specific interest. matching paths and status code. Clicking the Clicking on replay request will open a request "recycle" icon will reset the filter. If multiple fil- editor tag. The request can then be edited and ters are needed, it is possible to create addi- re-transmitted an arbitrary number of times. HTTP message viewer prove in future versions of Vega. Presently the Vega message viewer supports rendering of HTTP requests and responses are rendered syntax-highlighted markup, binary image con- in a component called the message viewer. tent, and binary data in hexadecimal repre- There is a message viewer for the request sentation. The Vega development team plans and the response. The arrangement of these to improve substantially in this area, adding viewers is configurable - while the default is support for a variety of types of structured tabbed (request, response), the positioning data. Within the message viewer are two sec- can be changed by selecting one of the icons tions: the HTTP headers and the message above the message viewer views. body. The header can be collapsed to make more room for the content. There is also an The message viewer is meant to serve as a icon to hide the request table and fill the UI container for rendering content in HTTP mes- area normally occupied by both with the mes- sages, including headers and message bod- sage viewer, creating more room for inspect- ies. The message viewer supports rendering ing the content of a message pair. of some complex structured data, and will im- www.insecuremag.com 55
- 56. Configuring interceptor rules of the response processing modules selected for use with the proxy. Alerts triggered by The Vega proxy can be configured to intercept these modules during proxy usage are listed HTTP requests and responses passing in their section in the Alerts view in the scan- through it. When a message is intercepted, it ner perspective. is held by the proxy until the user chooses to drop it or forward it. Pending messages can Having explained the scanner and proxy, we be modified before they are forwarded. The will now walk-through extending Vega through interceptor can be set to intercept all mes- the development of custom modules. sages, or only those that match certain crite- ria. Examples of criteria for interception in- Extending Vega clude method type, status code, and regexp matching on hostname or path. For example, Vega modules are written in Javascript and it is possible to configure an interception rule are available to use when placed in the cor- so that all outgoing requests for rect directory - restarting Vega should not be /vulnerable.php are intercepted, while all oth- necessary. Modules can also be modified ers are passed through. without necessitating a restart. On Linux sys- tems, this directory is in scripts/scanner/ An indicator at the bottom of the Vega inter- modules. There are two additional sub- face will notify the user when an intercepted directories, injection/ and response/, used for message is pending. Clicking the button will storing the two respective types of modules. take the user to an interface where the pend- ing request can be edited and then forwarded Response processing module or dropped. When multiple requests are pend- ing, it may be more useful to view all of them Response processing modules run on every in a table. Clicking the "Proxy Status" tab response received by Vega. They process re- brings up such a table. Multiple rows can be sponses to scanner-issued requests as well selected and forwarded or dropped at once. as responses passing through the proxy. In this tutorial, we'll look at the vinfo-email.js SSL module, located in scripts/scanner/modules/ response. For observing/manipulating communication between a HTTPS client and server, Vega The first requirement of any response proc- performs a dynamic man-in-the-middle certifi- essing module is a module object. This object cate injection when SSL is encountered. This supplies the name and category of the module can (and should) cause a certificate error in to Vega. It can also supply a flag to indicate connecting HTTPS clients, as the certificate whether or not this module should be disabled injected by Vega is not issued by a trusted by default. CA. For convenience, it is possible to have Vega generate a CA certificate that can be var module = { imported into a client's certificate store. To name: "E-Mail Finder Module", generate this certificate, visit the magic proxy type: "response-processor", URL http://vega/ca.crt with a browser config- defaultDisabled: false ured to use the Vega proxy. With Firefox, the }; user will be asked if they wish to import the certificate. The certificate may need to be Some of the modules are set to be disabled saved and then manually imported into the by default. This may be desirable for a variety certificate store for other HTTPS clients. of reasons, such as their computational cost or the number of false positives they produce. Response processing modules The entry point of a response processing It is possible to run response processing module is a function called run() that accepts modules during use of the Vega proxy. Most of three parameters: them are set to run by default. The tool icon to function run(request, response, ctx) the right of the proxy stop icon brings up a list www.insecuremag.com 56
- 57. These parameters are: an object representing Response processing modules are invoked the HTTP request, the HTTP response, and when a response is processed. The logic of the context. The context object connects the the module occurs within the run() function. module to Vega and exposes the scanner API to the module developer. These objects are In this example the module is analyzing the documented in detail on the Subgraph docu- body of the responses using Javascript regu- mentation website: lar expressions to try and identify email ad- https://support.subgraph.com/trac/wiki/Respon dresses: seProcessingModules function run(request, response, ctx) { var atDomainRegex = /@(?:[^s.]{1,64}.)+S{2,6}/, mailRegex = /w[^s@]*@(?:[^s.]{1,64}.)+S{2,6}/g, strictMailRegex = /[w!#$%&'*+-/=?^`{|}~.]+@(?:(([a-z0-9]{1}[a-z0-9-]{0,62}[a-z0-9]{1})|[a-z]) .)+(?:aero|arpa|biz|com|coop|edu|gov|info|int|mil|museum|name|net|org|pro|tra vel|mobi|asia|xxx|[a-z][a-z])/i, body = response.bodyAsString, emails = [], r, sr, i, found; // First the module attempts to find a basic match of characters@domain, if it does not, it returns if (!atDomainRegex.test(body)) return; // It then attempts to match a more strict regular expression. Any matches are converted to lowercase and uniquely stored in an array while (r = mailRegex.exec(body)) { sr = strictMailRegex.exec(r[0]); if (sr && emails.indexOf(sr[0]) == -1) { found = 0; for (i = 0; i < emails.length; i++) { if (emails[i] == sr[0].toLowerCase()) { found = 1; } } if (!found) { emails.push(sr[0].toLowerCase()); } } } The processing of the response is complete, and the alert can now be generated if e-mail addresses were identified: if (emails.length) { A unique key is then constructed for this alert. come up with a scheme that prevents too The key in this example is constructed by sort- many duplicate alerts while still providing use- ing all of the discovered e-mail addresses and ful findings. In this case, the key should pre- delimiting them with a space. The key is arbi- vent other alerts from being generated for the trary - it is up to the module developer to same precise type of finding: var key = emails.sort().join(" "); www.insecuremag.com 57
- 58. var uristr = String(request.requestLine.uri); var uripart = uristr.replace(/?.*/, ""); The alert() function is exposed through the context object: ctx.alert("vinfo-emails", request, response, {!! // The XML file, the request, and response objects "output": emails.join(" "),!// The output to be rendered in the alert "resource": uripart, ! ! // The URI for the resource field of the alert key: "vinfo-emails" + uripart + key! // The unique key }); } Regular expressions are not the only tool that from a response body. If it succeeds, a DOM can be used to analyze response content. The object will be returned. Otherwise Vega Javascript API comes with JQuery to response.document will be null. JQuery can analyze content at the DOM level. A DOM ob- then be used on the DOM. An example of this ject can be obtained by accessing module in use is in response.document. When the module does scripts/scanner/modules/response/vautocompl this, Vega will lazily attempt to parse a DOM ete.js: [..] if (response.document) { var form = jQuery("form", response.document); form.children().each(function() { if ((this.getAttribute("type") != null) && (this.getAttribute("type") == "password")) { if ((this.getAttribute("autocomplete") == null) || (this.getAttribute("autocomplete").toLowerCase() != "off")) { found++; } } }); } [..] Writing a basic module The entry point of a basic module is a function named initialize() which accepts a single pa- This guide will explain a very simple example rameter, the "context" object. As with re- of a basic module. The module is located at sponse processing modules, the context ob- scripts/scanner/modules/injection/header-injec ject connects the module to Vega, exposing t.js. the API. The context object for basic modules is distinct from the object of the same name As with response processing modules, every for response processing modules. basic module has a metadata object called "module": function initialize(ctx) { var module = { Vega populates a tree-like data structure name: "HTTP Header Injection known as a path state as it crawls a website. checks", Basic modules run on path state nodes, which category: "Injection Modules" may either be files, directories, or parameters. }; www.insecuremag.com 58
- 59. Vega handles identifying and iterating over the of the responses to try and identify instances parameters while still providing great flexibility of header injection. to the module developer. The first parameter to this fuzzing example is In a very simple example, the following API the callback function, listed in the next code function accepts only fuzzed parameter values snippet, followed by an array of parameter and requires no knowledge of where in the values to be injected. The last parameter is an application the module is. This module at- optional Boolean indicating whether the in- tempts to inject values into the application. jected values are to be appended to a pa- The callback function examines the headers rameter seen by the crawler, in this case it is set to true: ctx.submitMultipleAlteredRequests(process, ["bogusnVega-Inject:bogus", "bogusrVega-Inject:bogus"], true); } There are many functions for generating new https://support.subgraph.com/trac/wiki/PathSt requests exposed through the context object ate for basic modules. The reader is invited to view them all at the Subgraph documentation When a basic module queues requests for the website: crawler (for example, to fuzz parameters), it must register a callback function that Vega will https://support.subgraph.com/trac/wiki/BasicM run for each of the responses. The callback oduleContext function is passed three parameters by Vega: the request object, response object, and con- The path state node of the module is accessi- text object. In this example, the callback func- ble through the context object, if the module tion checks the response headers to see if in- writer wishes to know information about the jected values are present, generating an alert path state node on which it is running. The if it finds them: path state structure is explained here: function process(req, res, ctx) { if (res.hasHeader("Vega-Inject")) { ctx.alert("vinfo-header-inject", request, response, { message: "Injected Vega-Inject header into response", resource: request.requestLine.uri }); } More advanced analysis and shell injection. It is also possible to ana- lyze the timing of responses. Readers inter- The basic module explained in this tutorial is ested in experimenting with Vega modules are among the simplest examples. It is possible advised to read the API documentation for ba- for basic modules to send many requests to sic modules on https://support.subgraph.com. perform more complex logical analysis. It is useful to note that the ctx.debug() function One facility for doing this provided by Vega is can be used to print output to the console dur- page fingerprinting, where Vega distills page ing module development. It is also possible for contents to a simpler representation such that modules to store data in a key-value database two fingerprints can be efficiently compared to share data between modules. The reader for page differences. This is used as the basis should refer to the documentation on the con- for determining positive or negative results in text objects for more information. several injection modules, including blind SQL www.insecuremag.com 59
- 60. Alerts It is possible to create completely customizable alerts. The template structure is very simple. The example XML file "test.xml" is located in xml/alerts/: <?xml version="1.0" encoding="UTF-8"?> <alert> <title>Test vulnerability</title> <class>Example</class> <severity>High</severity> <impact> Could be used to demonstrate partially completed functionality of web application scanner.</impact> <impact> May cause boredom.</impact> <remediation> There is currently no solution for this vulnerability. Contact your vendor. </remediation> <discussion> Discuss it here. </discussion> <external> <url address="http://subgraph.com">Subgraph security.</url> </external> <references> <url address="http://minecraft.net">Minecraft is a good game.</url> <url address="http://en.wikipedia.com">Learn stuff here</url> </references> </alert> Conclusion entirely new features to support more ad- vanced security checks in the future. We invite Vega is a relatively new platform. The primary feedback via Twitter (@subgraph), e-mail objective of the project is to build the most ex- (info@subgraph.com) or on IRC, in #subgraph tensible platform for web security assessment. on freenode. The Vega development team hopes to bring David Mirza Ahmad is the President of Subgraph (www.subgraph.com). David has over 10 years in the infor- mation security business. He started his professional experience as a founding member of Security Focus, which was acquired by Symantec in 2002. David also moderated the Bugtraq mailing list, a historically impor- tant forum for discussion of security vulnerabilities, for over four years. He has spoken at Black Hat, Can Sec West, AusCERT and numerous other security conferences, as well as made contributions to books, magazines and other publications. David also participated in a NIAC working group on behalf of Symantec to develop the first version of the CVSS model and served as editor for the At- tack Trends section of IEEE Security & Privacy for over three years. His current obsession is building Sub- graph, a Montreal-based open source security startup. www.insecuremag.com 60
- 61. This article shows what next-generation security really means, and why it’s critical that organizations understand user and application activity in order to fully protect their networks. The firewall is now over 20 years old. That’s ment, and people began talking about the quite an achievement, considering that some deperimeterization of networks. A few years security industry observers have been predict- later, it was the emergence of SSL VPNs and ing its demise for over half of that time. increased use of smartphones; and today, cloud applications are supposedly the latest Evolving IT infrastructures and increasingly threat signaling the firewall’s demise. sophisticated security threats have brought repeated warnings about the firewall's im- These predictions usually go hand-in-hand pending obsolescence. This started in the late with talk about next generation firewalls, a 1990s, when laptop usage and remote access term that implies we have something new and started to spread in the corporate environ- beyond what came before it. www.insecuremag.com 61
- 62. Certainly, the emerging technology trends Organizations use many different ways to ac- mentioned before have forced business net- cess corporate data, such as client-based and works to handle an ever-increasing number of clientless VPNs from laptops and smart- events and a greater variety of traffic than be- phones, or cloud applications – but the bor- fore. ders are still there. Overall network activity is simply more complex, with more events to Border control control, more crossing points, and a greater variety of traffic than ever before. While it’s true that networks have changed dramatically, from the relative simplicity of a It’s similar to a country controlling its borders. decade ago to far more complex topologies There are many different ways to travel into a today, and perimeters have become more ex- country: by air, by rail, by sea or by road – just tended and even fragmented, but those pe- as there are different ways to access a net- rimeters still exist. work. Yet these don’t make border security controls obsolete. You simply need to imple- There is still a very clear separation and bor- ment different types of controls at airports, der between the internal, trusted infrastruc- ferry terminals and international railway sta- ture, and external untrusted networks. tions, in order to effectively monitor and in- spect the different types of traffic. OVERALL NETWORK ACTIVITY IS SIMPLY MORE COMPLEX, WITH MORE EVENTS TO CONTROL, MORE CROSSING POINTS, AND A GREATER VARIETY OF TRAFFIC THAN EVER BEFORE What do you mean by "next generation"? change IP address, tracking them is relatively simple. This used to mean that applying secu- Similarly, gateways have evolved beyond the rity was relatively simple, too, with security simple monitoring of certain ports, IP ad- policy management defining access based on dresses, or the packet activity streaming to the internet protocol (IP) addresses of the de- and from each address, to be able to scruti- vices in use. However, this approach to policy nize specific user and application activity. management is now dangerously outdated in most organizations, because it is fundamen- While this is an evolution, it’s not really next- tally dependent on how much physical control generation. In fact, firewalls have been able to you have over the devices that connect to identify applications in-use for the last 17 your network. years by analyzing packet data. Of course, there are far more applications in use in most Moving risks companies now than ever before, but the prin- ciple of application identification in itself is The growing demand for smartphones and nothing new. tablet PCs has resulted in employees having multiple devices and, therefore, many IP ad- The key issue today is more about adding dresses. greater capabilities to look deep within the web traffic passing through the gateway and The rise of mobile computing, together with identify precisely which applications are in use new online applications, makes it difficult for and track exactly which users are running businesses to keep up with policy change re- them. quests. If the requests keep coming in based on users and their devices, and organizations This is the area that’s truly new, because are still compiling policies based on static IP companies are no longer just dealing with addresses, the business is already at risk of fixed devices, or static, office-bound users on exposure. their networks. Networks have, until now, been defined by the addresses they use. Pro- Even more worrisome is the fact that many of vided users don’t move around too often or these devices are being brought in from users’ www.insecuremag.com 62
- 63. homes without being validated, secured or the ability to identify applications that are NOT even looked at by the IT department. defined by standards - such as web applica- tions, social media portals and more - is a Users are bleeding their personal devices powerful addition to creating a next generation such as tablets, smartphones and personal policy. laptops all over networks, taking work home and bringing home to work. If you add the ability to detect and manage user access to those applications, businesses As organizations adopt more agile computing can further strengthen application control. By solutions, they are finding that security poli- allowing users to interact with the security sys- cies cannot keep pace with the changes, cre- tem, both to remind them of corporate policy ating all sorts of headaches. As such, what’s on acceptable use of applications and to take needed to help companies manage risk, pro- feedback in real-time on why the user needs tect data, audit network activity and give better access and the intended purpose of their us- control over what users are doing isn’t a “next age, organizations can add a further layer of generation” product or feature set: it’s next security reinforcement and protection. generation policies and policy management. Data – the core element So how should you approach the development of security policies that reflect the way net- The three points covered so far help to ensure works are being used today? And how do you organizations can identify which users are ac- ensure those policies are enforced? cessing the network, from which device or ap- plication. However, the core element of secu- User ID checks rity policy is the ability to analyze the data that is being accessed, sent and manipulated to Knowing who your users are is critical to ensure users are not sharing – or leaking – managing policy; knowing what IP addresses sensitive information. they are using is less so. As such, defining policy based on user access AND type of de- This requires assessing not only what applica- vice is the only logical choice, as it gives a tions employees can use, but what data these smarter means for managing access from applications are allowed to use, and, in turn, fast-growing consumerized estates, where the taking steps to protect sensitive data from in- device may not always be known. appropriate or non-compliant usage. Managing devices In conclusion, the increasing adoption of con- sumerization, virtualization and cloud comput- Understanding what devices employees are ing means that network infrastructure is no using for network access will also help organi- longer static: it’s agile, dynamic and frag- zations make informed decisions about their mented, with data flowing in unexpected and security policies. This allows them to track unpredictable ways. what devices have accessed which data, so if they need to determine where the networks Next-generation security has to include the may have been breached from, there is al- “human factor” – the people using networks, ready a defined limit on the number of people the devices they use, the applications they are and devices with access. Consider just how allowed to run, and the data those applica- more effective a security policy could be with tions can access and modify – to reflect this the addition of this parameter. dynamic network usage. Only then can you create cohesive, next-generation security poli- Application control cies that truly protect what matters to your business. The ability to identify application activity on a firewall or gateway is nothing new. However, Terry Greer-King is the UK managing director at Check Point (www.checkpoint.com). www.insecuremag.com 63