Interpolique

INTERPOLIQUE
(OR, THE ONLY GOOD DEFENSE IS
THROUGH A BRUTAL OFFENSE)
Dan Kaminsky, Chief Scientist
Recursion Ventures
dan@recursion.com

ANNOUNCEMENT
 This is my new company. Woot.
 Recursion productizes significant research
 It’s time to do things a little differently
 This talk isn’t a sales pitch for Recursion, but it’s an idea
regarding its philosophy

A STORY
 Design flaw in SSL
 The server thought it was resuming, the client thought it
was connecting
 Project Mogul spawned to fix it
 Several months in deep secrecy
 Thousands of hours spent on IETF fix
 The fix broke <1% of servers
 No big deal, right?

REALITY
 “Note that to benefit from the fix for CVE-2009-3555
added in nss-3.12.6, Firefox 3.6 users will need to
set their security.ssl.require_safe_negotiation
preference to true. In Mandriva the default setting is
false due to problems with some common sites.” –
Mandriva Patch Notes
 They thought knocking out a few sites was acceptable
for a remediation
 They were wrong

THE BAD NEWS
 We give bad advice
 Pen testers are very good at breaking things
 Our “remediation” advice tends towards myopia
 We consider only our own engineering requirements
 We assume tools are static, and bash the craftsman

THE GOOD NEWS
 We are the keys to there actually being good advice
 We are the one community that actually knows how
things break
 We hold the knowledge to end the bugs we keep seeing

A SIMPLE QUESTION
 When I log into two SSH servers, do I need to worry
about one accessing the other?
 No
 When I log into two web sites, do I need to worry
about one accessing the other?
 Yes
 Why?
 Because SSH does not have totally broken session
management

SIMPLE THINGS, SIMPLY BROKEN
 The web was never designed to have authenticated
resources
 Auth was bolted on (because Basic/Digest never got
fixed)
 Normal Mechanism For Managing Credentials
 Password causes Set-Cookie
 Cookie sent with each query to target domain
 Cookie is sent even with requests caused by third
party domains
 User’s credentials are mixed with attacker’s URL
 This is why most XSS/XSRF attacks are dangerous
 Cross Site Scripting and Cross Site Request Forgery
wouldn’t be nearly the big deal they are if they didn’t work
cross site

THE PEN TESTER REACTION:
DEV, DO MORE WORK
 XSRF Tokens
 Manually add a token to every authenticated URL
 Requires touching everything in a web app that
generates a URL
 How’s that working out for us?
 This seems to be a lot of work
 If/when we come back six months later, it’s not usually
done, is it?

A MODEST PROPOSAL
 Couldn’t the tools be better?
 The big debate: Should SVGs animate?
 Unsaid: Shouldn’t it be possible to easily log into a
web site without other sites being able to use your
creds?

AN ATTEMPT
 A fix that requires no change to the browser is
better
 So I tried to find one
 Server Side Referrer Checking
 Client Side Referrer Checking
 Window.Name Checking
 Window.SessionStorage Checking
 It says SessionStorage! Surely it’s perfect for Session
Management!
 They all failed
 Thank you Cstone, Kuza55, Amit Klein, David Ross,
SirDarckcat

WHEN FAILURE IS SUCCESS:
OUR PROBLEM WITH LATENCY
 My suggested defenses were defeated early in
development
 We, as a community, have a latency problem
 We don’t break during development
 We don’t break at release
 We don’t break when early adopters are deploying
 We break only when it gets really popular
 By then, it’s in customer hands, and the best we can do is
give the customers really expensive advice on how to fix it
 We need to close the feedback loop

AT MINIMUM
 Whatever’s going on with other defenses, I want
mine to be thoroughly, even brutally audited as
soon as possible
 Life is too short to back broken code!
 Session Management will require modifications to
the browser
 Something else might not…

ON LANGUAGES
 "The bottom-line is that there just isn't a large
measurable difference in the security postures from
language to language or framework to framework --
specifically Microsoft ASP Classic, Microsoft .NET,
Java, Cold Fusion, PHP, and Perl. Sure in theory
one might be significantly more secure than the
others, but when deployed on the Web it's just not
the case.”
--Jeremiah Grossman, CTO, White Hat
Security (a guy who has audited a lot of web
applications)
 Question: Why aren’t the type safe languages
safer against web attack than the type unsafe
languages?

WE AREN’T ACTUALLY USING THEM
 Reality of web development
 HTML and JavaScript and CSS and XML and SQL and
PHP and C# and…
 “On the web, every time you sneeze, you’re writing in a
new language”
 How do we communicate across all these
languages?
 Strings
 And how type safe are strings?
 Not at all

ALL INJECTIONS ARE TYPE BUGS
 select count(*) from foo where x=‘x' or '1'='1';
 The C#/PHP/Java/Ruby sender thinks there’s a string
there.
 The SQL receiver thinks there’s a string, a concatenator,
another string, and comparator, and another string
there.
 The challenge: Maintaining type safety across
language boundaries

ISN’T THIS A SOLVED PROBLEM?
 Escaping?
 Parameterized Queries?

NO ESCAPE
 $conn->query(“select * from foo where x=“$foo”;”);
 Is this secure or not?
 Who knows, depends on whether $foo has been
escaped between when it first came in on the wire, and
when it’s being passed into the DB
 This simple line of code is expensive to debug!
 If somebody removes the escape(), the code still
works
 “Fails open”

ACCIDENTAL ESCAPE
 What does it mean to escape?
 “Block Evil Characters”
 Was very easy to determine evil characters when we just had
ASCII
 Only 256 possible bytes
 Unicode changes that
 Millions of characters
 All of which could mutate (“best fit match”) into one another
 All of which have multiple possible encodings, and
representations within encodings
 Escaping works by accident, without a solid contract
 Keeps getting updated
 escape(), escapeURI(), escapeURIComponent()

WHAT ABOUT PARAMETERIZED QUERIES?
 Which would you rather write?
 $r = $m->query(“SELECT * from foo where
fname=‘$fname’ and lname=‘$lname’ and
address=‘$address’ and city=‘$city’”);
 $p->prepare(“SELECT * from foo where
fname=‘$fname’ and lname=‘$lname’ and
address=‘$address’ and city=‘$city’”);
$p->set(1, $fname);
$p->set(2, $lname);
$p->set(3, $address);
$p->set(4, $city);
$r = $m->queryPrepared($p);

REALITY OF PARAMETERIZED QUERIES
 No developer has ever written a parameterized
query without a gun to his head
 We should know
 We hold the gun

POSITIONAL GENERATION ISN’T ANY BETTER
(C/O MIKE SAMUEL)

O(N) UI WORK FAILS
(BEST CASE EYE TRACKING)

HOW INJECTIONS HAPPEN /
HOW DEVS LIKE TO WRITE CODE
 String Interpolation:
select count(*) from foo where x=‘$_GET[“foo”]';
 String Concatenation:
“select count(*) from foo where x=”“ +
$_GET[“foo”] + “”;”;
 Why they write code this way
 Devs are thinking inline
 They want to be writing inline
 See: Fitts’ Law

IS IT POSSIBLE…
 …to let devs write inline code, without exposing the
resultant strings to injections?
 Yes – by making String Interpolation smarter
 RETAIN: The language still sees the boundary between the
environment(“select * from…”) and the variable ($_GET…).
 TRANSLATE: Given that metadata, the language can do
smarter things than just slap unprocessed strings together
 (This overlaps with, and extends, Mike Samuel’s
excellent “Secure String Interpolation” work, seen at
http://tinyurl.com/2lbrdy.)
 Working with Mike

INTERPOLIQUE DEMO[5]
 Submit
if($_POST[action] == "add"){
$conn->query(eval(b(
'insert into posts values(^^_POST[author] ,
^^_POST[content] );‘
)));}
 Return
$r = $conn->query("select * from posts");
while($row = $r->fetch_assoc()) { echo eval(sb(
'data: ^^row[author] ^^row[content]<br>n‘
)); }

WHAT’S GOING ON
 Language interpolators are blind – they just push
strings into strings
 So we write custom interpolators – the dev puts in what
he wants, the compiler sees what it needs

WHAT TO INTERPOLATE INTO
 Parameterized Queries are an obvious target
 Programmer writes:
select * from table where fname=^^fname and
country=^^country and x=^^x;
 Interpolique expands:
$statement = $conn->prepare("select * from table where
fname=? and country=? and x=? ");
$statement->bind_param("s", $fname);
$statement->bind_param("s", $country);
$statement->bind_param("s", $x);

COULD DO ESCAPES…
 …but no faith they actually work correctly

BASE64: ESCAPING DONE RIGHT
 Programmer writes:
select * from table where fname=^^fname and
country=^^country and x=^^x;
 Interpolique expands:
select * from table where
fname=b64d("VEhJUyBJUyBUSEUgU1RPUlkgQUx
MIEFCT1VUIEhPVyBNWSBMSUZFIEdPVCBUVVJ
ORUQgVVBTSURFIERPV04=") and
country=b64d("d2Fzc3Nzc3Nzc3Nzc3Nzc3NzdXA=
") and x=b64d("eXl5eXk=") ;

WHY THIS WORKS
 Type safe going into b64d() function
 That’s never getting interpreted as anything but a string
 Type safe coming out of b64d() function
 B64d() cast to return a string
 Not a subquery, not a conditional, not anything other
than a string
 B64d() a MySQL UDF that’s already written, has no
apparent time penalty, will be released with Interpolique
 Most other databases already have B64 support
 In a pinch, could use MySQL hex/unhex

TWO MODES OF BASE64
 Late binding
 Interpolation inserts the Base64 handler
 Text is plain until right before it crosses the frontend/backend
layer
 SQL looks like this:
select * from foo where x=^^foo;
 Early Binding
 Base64 the variable as soon as it comes in off the HTTP
request
 SQL looks like this:
select * from foo where x=b64d($foo);
 Pen testers: If somebody fails to escape $foo,
everything still works. If somebody fails to Base64
Encode $foo, everything breaks immediately

STATIC ANALYSIS
 You know what’s better than having a static
analyzer?

AHEM
 Not needing a static analyzer

BASE64 IN THE OTHER DIRECTION
 <span id=3520750
b64text="Zm9v">___</span><script>do_decode(35
20750)</script>
 Create a SPAN with a random ID and a dynamic
attribute that contains its base64’d content
 Call do_decode with that ID, which can now look up the
element in O(1) time
 Use this construction to retain streamability
 Thank/Blame CP for this

DOM INTERACTION: SIMPLE
 Push to textContent
 ob = document.getElementById(id); ob.textContent =
Base64.decode(ob.getAttribute("b64text"));
 We never go through the browser HTML parser

DOM INTERACTION: COMPLEX
 Push to appropriate createElements
 ob = document.getElementById(id);
raw = Base64.decode(ob.getAttribute("b64text")); safeParse(raw,
ob);
 HTMLParser(src, {
start: function( tag, attrs, unary ) {
…
if(tag == "i" || tag == "b" || tag == "img" || tag == "a"){
el = document.createElement(tag);
…
Basic idea is to have a simple HTML parser that extracts what it can,
creates elements according to whitelisted rules, and importantly,
never goes through the browser HTML parser
 See also: “Blueprint”, a system that moves all DOM
generation to JS
 http://www.cs.uic.edu/~venkat/research/papers/blueprint-
oakland09.pdf

IMPORTANT NOTE
 Security Is Quantized
 There’s a set of elements that can be safely exposed
 There’s a set that can’t
 The game is to expose only those tags and attributes
that don’t expand to arbitrary JS
 Either you have prevented wishing for more wishes, or
you have not
 (We see this from the webmail attack surface)

HOW THIS WORKS
 Primary Mechanism: Eval
 Yes, there’s risk here, and yes we’re going to talk about that
risk – we need this for scoping reasons
 Programmer written query: select * from table where
fname=^^fname and country=^^country and x=^^x;.
 To Eval: return ("select * from table where fname=b64d("" .
base64_encode($fname) . "") and country=b64d("" .
base64_encode($country) . "") and x=b64d("" .
base64_encode($x) . "") ;");
 Eval Out: select * from table where
fname=b64d("VEhJUyBJUyBUSEUgU1RPUlkgQUxMIEFCT1
VUIEhPVyBNWSBMSUZFIEdPVCBUVVJORUQgVVBTSURFI
ERPV04=") and
country=b64d("d2Fzc3Nzc3Nzc3Nzc3Nzc3NzdXA=") and
x=b64d("eXl5eXk=")

CAN WE OPERATE WITHOUT EVAL?
 No Eval in Java or C#
 One approach: Combine variable argument functions
with string subclass tagging
 public bwrap w = new bwrap();
w.s(w.c("select * from foo where x="), argument1, w.c("and
y="), argument2);
 If you forget to mark the safe code, it breaks
 Another approach:
 w.code(“select * from foo where
x=“).data(argument1).code(“and
y=“).data(argument2).toString()
 Similar to LINQ etc. but actually works for arbitary grammars
 If you mismark code as data, or vice versa, it breaks
 Both actually implemented! (Tiny HOPE Announce)

THE STATUS QUO
 We see this doesn’t work:
 String s = “select * from foo where x = ”“ + escape(s) +
“”;”;
 By doesn’t work: It is too similar to this:
String s = “select * from foo where x = ”“ + s + “”;”;
 Devs mess this up, but the code works anyway
 As a matter of principle, devs will do enough
work to make the code function
 If it works, it should work securely
 If it isn’t working securely, it shouldn’t be working at all
 The trick is to not make it easier to get around the security,
than it is to do things right

WHY CUSTOM INTERPOLATORS ARE HARD:
THE ANCIENT SCOPE WAR
 Lexical Scope: Scope Known At Compile Time
 Variables are “pushed” into child scopes
 Dynamic Scope: Scope Determined At Run Time
 Variables are “pulled” by child scopes
 Lexical scope has won, and has systematically
removed methods that allow any code to access
variables not explicitly pushed in
 This makes it rather difficult to write a function that sees
^^variable and thus deferences that variable
 There are silly “superclass” or “parent” modifiers in some
languages, but they’re all special case
 In Java and C#, they went so far as to leave local variables
unnamed on the stack, so you couldn’t just hop into previous
stack frames and dereference from there!

TO BE CLEAR
 Yes, there is risk to eval, and we’ll be talking about it
 Yes, there are very nice and very good reasons for lexical
scope to be the default state
 The fact that the vast majority of programming
languages, type safe or not, are repeatedly found to
expose injection flaws is a direct sign that something
is wrong
 Put simply, language design needs to be informed by the
bloody findings of pen testers
 It is informed by performance engineers
 It is informed by usability engineers
 Memory safety didn’t come from security engineers, it came from
reliability engineers
 I think we need a way to write functions that execute in present scope

YES, THIS MEANS
 (LISP) (WAS) (RIGHT)
 (((NOT ABOUT EVERYTHING)))
 (((THEY ( HAD A POINT ( HERE ))))
 Crazy theory
 JavaScript has been successful because it’s been able
to mutate to absorb almost any language construct
 “More dialects of JavaScript than Chinese”

RISKS
 There are three things that can go wrong with any
defensive technology
 It doesn’t work
 None of this mealy mouthed, “well, it depends on what your threat
model is”
 Either it does what it says, or it doesn’t!
 It doesn’t work in the field
 Security: It is too easy to screw up
 It has side effects
 Fails other first class engineering requirements (too slow,
unstable, hard to deploy, etc)
 I am looking for destructive analysis on these
techniques, and will accept criticism on any of the
above fronts
 Here is what I know so far

THE HANDLERS APPEAR RELATIVELY SOLID
 No known SQL Injection bypasses for Base64 into a
b64d() function
 Using a fast base64 decode – could be flaws here
 Could be databases that don’t type-lock return values
 No known flaws when putting arbitrary text into a
span.textContent field
 Well, except it doesn’t work in IE  Will port to its wonky
DOM
 Most testing is in Firefox -- Could be problems in
Chrome/Safari, Opera, etc.
 No known flaws when creating arbitrary DOM elements
and populating them, rather than pushing HTML
 IE6 is apparently slow at this
 Need to enumerate the full set of tags which are safe to put
into HTML

EVAL ADDS SOME RISK
 Don’t buy that a PHP server is safer if it isn’t
running eval
 Month of PHP Bugs = PHP not safe against any
arbitrary PHP, eval or not
 Eval in this context can make programmer errors
more severe
 Correct: eval(b(“select * from foo where x=‘^^x’”));
 Incorrect: eval(b(“select * from foo where x = ‘$x’;”));
 Before we had SQLi. Now we potentially have front end
code execution!
 This is why it’s now ^^foo instead of $!foo

MANAGING RISK OF EVAL
 b() can be smarter
 It can be aware of strings that break out of string-
returner
 It can be aware of SQL grammar, to the point that in
order to write a right hand variable, it must be ^^’d
 Select * from foo where x=^^x and y=safe(1);
 It can even be self-auditing – in PHP, it can use
debug_backtrace() to find the line that called it, and
validate that that line doesn’t have an unsafe language
deref

WHAT ONLY SORT OF WORKS
 “Requiring” Single Quotes
 In some languages, ‘$foo’ doesn’t interpolate, while
“$foo” does
 So, the thinking is, require eval(b(‘$foo’))
 This is a policy that cannot be enforced by present
compilers or languages (both ‘$foo’ and “$foo” turn into
a string in the parse tree)
 Could be enforced by a preprocessor
 At large shops, significant improvements in security are won
by blocking otherwise legal expressions as a coding policy
 Not convinced that smaller shops can/should absorb

PERFORMANCE
 Eval is slower than compiled code
 Translating strings could be a major pain point in some
languages
 Easy to cache the translation (because we retain the
boundary, accessing the normalized query form is trivial)
 Could potentially parameterize/accelerate more,
because it’s suddenly easy for the framework to
autorecognize repeated queries
 Base64 is fast
 Slight bandwidth increase, but nothing compared to
URLEncoding

ANYTHING ELSE?
 I don’t know.
 Hope: There’s about two months till Black Hat.
Lets find out!
 This isn’t a recommendation yet
 Clearly what we are doing right now is not working
 Lets find out the best things we can do with the present
languages
 Lets find out what we’d want from future languages
 It’s time we got involved in the discussion of what
software looks like

Interpolique

More Related Content

What's hot (20)

Similar to Interpolique (20)

More from Dan Kaminsky (19)

Interpolique