Don't scan, just ask

Don‘t scan, just ask
A new approach of identifying vulnerable web applications

Fabian Mihailowitsch 28th Chaos Communication Congress, 12/28/11 - Berlin

Summary

It’s about identifying web applications and systems…
Classical network reconnaissance techniques mostly rely on technical facts, like net blocks or
URLs. They don’t take the business connections into view and hence don’t identify all potential targets.
Therefore ‘Spider-Pig’ was developed. A tool that searches for web applications connected to a
company based on a scored keyword list and a crawler. That way web application run by satellite
companies or service providers can be identified as well.

TABLE OF CONTENTS
 Overview on classical network reconnaissance
 Restrictions and possible enhancements
 Presentation of ‘Spider-Pig’ and its approach
 Statistics of a real-life scenario
 Conclusion


Classical network reconnaissance

TECHNIQUES
 Query NIC and RIPE databases
 Reverse DNS
 DNS Zone Transfer
 Research in forums, social networks, …
 Google hacking
 Identify routing and hence new systems
 Brute-Force DNS
…

TOOLS
 whois, dig, web browser, Maltego, …


How Business looks like…

Company A.0 THE TARGET :)

SATELLITE COMPANIES
 Branch office
 Partner firm

Company A.1

SERVICE PROVIDER Company B
 Payment Provider
 Separate Webhoster
 Marketing Campaign
 Service Center
…


…what we see with network reconnaissance

Company A.0 ALL (?)
 Net blocks
 IP addresses
 URLs
 DNS information
SOME (!)
 Net blocks
 IP addresses
 URLs
 DNS information
Company A.1

Company B


Restrictions and possible enhancements

Restrictions Enhancements
Scans solely on a technical level: For an attacker it doesn’t count whether he
- IPs attacks company A or B:
- Net blocks - Access to sensitive data
- URLs - Reputational damage
-…
No business relations are taken into account.
Take business relationships into account.
Potentially incomplete view:
You can only search for what you know Build a logical network between systems,
(e.g. RIPE queries). independent of the owner.

Probably just one company in focus. Identify (all) applications / systems linked
to a company based on a business
connection level, independent of provider.

Not all targets are identified! Development of ‘Spider-Pig’.


Approach of Spider-Pig
Spider-Pig identifies web applications connected to a company in an automated manner and hence
allows to identify potential targets (web applications and systems) for further attacks.
Due to German laws the tool will not be released to public. But its methodology will be explained…

1. Model Business 2. Perform Search 3. Calculate Results 4. Exploit

Model the business, Based on the Duplicates are Targets are sent to a
services and keywords, different removed from the custom vulnerability
relationships of the search engines a results. For each scanner that uses the
company in scope. queried in an domain a score is data from the crawling
automated manner. calculated and a sorted phase and attacks the
Based on a scored list with the results is web sites in an
keyword list with Furthermore ‘whois generated. automated manner.
terms, like company information’ are
name, product names, gathered for the Top results are At the end, we
tax numbers. identified domains and reviewed manually and hopefully get a shell! :)
IP addresses. potential targets are
Note: No technical extracted.
details (e.g. IPs) have Identified web sites are
to be provided. crawled for links to
new web pages.


First step: Model Business

Idea
 Web applications connected to a company contain terms like company name, products or imprint.
 There are imprecise terms like products names that lead to many web sites (e.g. Amazon).
 There are precise and very specific terms like tax numbers or imprints that lead to company pages.

Solution
 Manually assemble a list that contains terms specific to the company in focus.
 Assign each entry a score, how unique and specific it is.

Note: Experience showed that ~200 terms are sufficient and already lead to hundreds thousands of
results. Technical details (e.g. URLs or software used) can be used but don’t have to. A score from 1-100
works pretty well, but can be adopted.


Second step: Perform Search

Idea
 Keywords are used to query different search engines (e.g. Google, Bing, Yahoo).
 Results are read and a list is built, where the fully qualified domain name (FQDN) is the index.
 Every time the FQDN is part of a search result, the score assigned to this search item is added.
 For the IP and domain name a ‘whois query’ is executed to get details on the owner.
 If the whois record contains the company name, address or a keyword, the FQDN is as hit.

1. Take 2. Query
keyword search
from list engines
Example: If you focus on a certain
market (e.g. Europe) you can exclude
all other results based on the
‘CountryCode’ received via whois.
3. Check
5.
results
Additional
and calc.
steps
score

Note: Using the FQDN as identifier,
4. Perform
whois already removes duplicates and
query speeds up the process.

FQDN IP URL Title Score Hit

www.a.com 1.2.3.4 http://www.a.com/about Company A 10 N
(9 + 1)
http://www.a.com/product1 View product1



Implementation Notes
 Search component was developed in Perl and is not more than 503 LOC :)
 Google and Bing are used as search engines.
 Results are stored in memory and wrote to disk in regular intervals.

Querying NIC and RIPE databases

Depending on the database, the results do have
different layouts. Consider this when parsing the
‘whois queries’.

If you query too fast, you’re getting blocked.
Hence you have to rotate the queried systems
or throttle the querying speed.

As the search results can easily consist of multiple
thousands of systems, consider that a small delay
per query will delay the whole process by days.



Using search engines
For Perl there are many modules for different search engines. However most of them don’t work.
But ‘curl’ and some regular expressions are fine to implement it yourself.
Google covers pretty much, however Bing leads to some additional results.

Google Bing
 Google Custom Search API can be used  Bing API is available
 1.000 queries / day are free  An AppID is needed that can be requested
 If you want to use the service efficient, you for free at Bing Developer Center
need to buy a search key, which is charged on  Service is free of charge, however the queries
the amount of search queries have to be throttled
 For a ‘normal’ search this should result in ~200$  Consider different markets (!) as they will lead
 Consider different languages (!) as they will to diverse results and disable filter (!)
lead to diverse results and disable filter (!)

Google:
https://www.googleapis.com/customsearch/v1?q=[qry]&num=10&start=[pointer]&lr=[lang]&key=[...]&cx=[cx]&filter=0

Bing:
http://api.search.live.net/xml.aspx?Appid=[appid]&query=[qry]&sources=web&adult=off&market=[market]&web.count=5
0&web.offset=[pointer]


Second step: Crawl

Idea
 Search queries lead to a scored list of potential targets.
 The higher the score, the more probable the web page is connected to the company in focus.
 Company web pages probably link to new company web pages.
 Web pages with a high score (e.g. 200) should be crawled and checked for external links.

Crawling top scored web pages

Select certain systems from the list (see third step) and feed them to a
crawler.

The crawler then identifies new external links on these web pages
and hence receives new web applications that might be of interest.
These can be appended to the results list.

Furthermore the crawler can already perform the first step of the
vulnerability scan.


Second step: Crawl

Credits
The vulnerability management
framework (pita) mentioned during
the talk hasn’t been developed by me.

It is the effort of multiple years of
development by various security
consultants working at diverse
companies and as freelancer!

For ‘Spider –Pig’ just some
modifications were made to base
components.

Crawler
 Written in Perl
 Identifies external links and adds
them to the results list
 Performs first step of VA


Third step: Calculate Results

Idea
 Now a scored list with all identified targets is available.
 If no filter is applied, the list contains hundreds thousands of results and cannot be checked manually.
 Results with a low score (e.g. <10) are probably not relevant.
 Results with a very high score are probably irrelevant, as they refer to Online Shops, Portals, etc.
 Hence just a small portion of the results has to be analyzed manually.

Score

Interesting Results

Web page


Fourth step: Exploit

Idea
 Now real targets have been identified that somehow are connected to the company in focus.
 Information regarding the content, forms, etc. have been saved in the database during the crawling.
 This basis can be used for an automated vulnerability scan.

pita
Note (again): The vulnerability management framework (pita) mentioned during the talk hasn’t been
developed by me. It is the effort of multiple years of development by various security consultants.

Based on the identified web applications and the data collected during the crawling (e.g. forms that can be
submitted), the vulnerability management framework ‘pita’ is started.

‘pita’ runs:
 Fuzzer (XSS, SQL Injection, …)
 Web Application Fingerprinter (wafp)
 Cookie/Session analyzer
 SSL tests
 3rd party tools (sqlmap, nmap, Nikto, …)

At the end we get a report with all vulnerabilities, out of keywords… 

Statistics

In case we would run ‘Spider-Pig’ against a big known company, we might
get the following results:
Keywords used 287
Duration (including VA) 7 days – it could be optimized :)
Costs ~200 Dollar - due to Google, without Google 0 Dollar
Unique FQDNs 150.871
FQDNs with filter 50.784 – after a filter for EU countries was applied
Applications reviewed 300
Confirmed applications 223
 Marketing campaigns, customer service, official web pages, …
 Applications were hosted on external and internal systems
 Official web pages within top 10! It works! :)
Vulnerabilities 


Conclusion

Classical network reconnaissance doesn’t take all aspects into account and
hence does not identify all potential targets of a company in focus.

Development of ‘Spider-Pig’, a tool to model business connections.
 ‘Spider-Pig’ allows to identify web applications linked to a company due to their business connections.
 Based on a list of keywords – without any technical details – ‘Spider-Pig’ attacks companies.
 Besides assembling a list of keywords and a manual review of results, no human interaction is needed.
 ‘Spider-Pig’ collects some interesting statistics/information and offers new perspectives.

Companies should not only focus on protecting their own systems.
 An attacker will (most likely) take the easiest way.
 Companies should not only protect their own systems but choose service providers with caution.
 Sensitive information should not be spread across multiple systems and business partners.


Thanks! Questions?
Contact me ‘fmihailowitsch [at] deloitte.de’

Credits
Richard Sammet, who developed the idea & tool with me!


Don't scan, just ask

More Related Content

Viewers also liked (11)

Similar to Don't scan, just ask (20)

Recently uploaded (20)

Don't scan, just ask