Large-scale application security Charlie Eriksen

Agenda Define the problem space Testing and auditing techniques Step 1 – Define target Step 2 – Understand the target Step 3 – Find critical/high-risk code paths to review Step 4 – Review and test Step 5 – Asses your findings and iterate Future challenges ▫Locating the absence of CSRF protection ▫Real-time alerting

Goals The ability to take a large amount of code and audit it for vulnerabilities Prioritize time spent in an audit based on risk Spend as little time as possible to find as many vulnerabilities as possible

Inspiration

Problem space

Vulnerability classes and mitigation Systematic vulnerabilities Framework Code review Manual/Automatic testing Business/design logic Code review Manual testing

What we’re not trying to solve Reviewing business logic Pin-point accuracy of vulnerabilities (Not really) reviewing for the absence of… ▫We’ll look at my attempt at reviewing for the absence of CSRF protection later!

Testing techniques White box Black box The sweet spot

White box Black box Pros ▫Not restricted by your discovery phase ▫Gets you straight to where interesting things happen Cons ▫Some code can be too hard to statically analyze ▫Things may hide in plain sight Pros ▫Some parts can be automated ▫Allows you to observe behavior ▫Will uncover the most obvious flaws Cons ▫Uncovering some bugs can be expensive ▫Special code-paths may never be executed

Testing techniques

Audit ROI This is probably true for most companies ▫Large number of applications ▫Lots of legacy ▫Not enough time to audit all thoroughly Consider: ▫What do you not know/Are there risks you’re not aware of? ▫What is your vulnerability tolerance? ▫What vulnerabilities have the highest impact? ▫What vulnerabilities are most likely to be discovered?

Audit ROI Hard Easy High Low $$$ Good! Are you google? If not, wasted time/money

Candidate point strategy Fastest way to identify a lot of vulnerabilities Starts with identifying: ▫Points with side effects ▫Known vulnerable patterns Then back-tracing

Step 1 – Define target Find a target ▫In-house? ▫Public projects?  Plugins! Obtain source code

Wordpress example

Step 2 - Understand the target Learn the language Internalize OWASP top 10 Observe the framework and language ▫Dangerous functions ▫Mitigation techniques Find commonly vulnerable code patterns

PHP/Wordpress important functions A good list exists here for PHP: ploitable-php-functions ploitable-php-functions Highlights: ▫include/require ▫system/exec/shell_exec ▫eval Wordpress specific: ▫wpdb

Step 3 – Find critical/high-risk code paths to review Higher risk code paths is where you’ll want to spend more time Determine your critical functionality and assets Examples might be: ▫System calls ▫Database access ▫File system access ▫Web-specific things, forms, markup output ▫Encoding(base64) ▫Cryptographic usage(md5!?) These will be our candidate points

Good PHP regexes (include|require).*\$_(POST|GET|REQUEST) file_get_contents.*\$_(POST|GET|REQUEST) (eval|exec|system|shell_exec).*\$_(POST|GET|REQ UEST) SELECT.* FROM.* \$_(POST|GET|REQUEST) (echo|print).*\$_(POST|GET|REQUEST)

Step 4 – Review and test Start by casting the net wide on a project As you learn more about it, start being more specific and reduce noise Learn to review code at a glance High risk vulnerabilities are usually easily seen at a glance

Example 1

Example 1 – Cryptographic md5\s?\(.*\$_(GET|POST|REQUEST)

Example 1 – Cryptographic exploit

Example 2

Example 2 – File inclusion (include|require).*\$_(POST|GET|REQUEST)

Example 2 – File inclusion /wordpress/wp-content/plugins/zingiri-web- example.com/

Example 3

Example 3 – SQL Injection SELECT.* FROM.* \$_(POST|GET|REQUEST)

Example 3 – SQL Injection /wordpress/wp-content/plugins/all-video- gallery/xml/playlist.php?vid=2 UNION SELECT 1, 2, user(), 5, 6, 7, 8, 9, 10, database(), 12, 13, 14, 15, 16, 17, 18

Example 4

Example 4 – File disclosure file_get_contents.*\$_(POST|GET|REQUEST)

Example 4 – File disclosure /wordpress/wp-content/plugins/google-document- embedder/libs/pdf.php?fn=lol.pdf&file=../../../../wp- config.php

Step 5 – Asses your findings and iterate Steps to take ▫Assess risk ▫Do root cause analysis ▫Consider if there is likely to be more vulnerabilities of this type ▫Find out if there are steps that can be taken to mitigate the class of vulnerability at large Find next steps to improve your ROI ▫Are you coming close to your risk tolerance? ▫Are there still unknowns? ▫Are there other higher-risk areas(ROI?) ▫Have you addressed the most discoverable bugs?

Methodology All vulnerabilities found using this method All vulnerabilities submitted to Secunia SVCRP ▫Sweet program, check it out Data based on their advisories ▫Download numbers pulled manually

Research findings 24 plugins ~2 million downloads 66 vulnerabilities

Downloads

Vulnerabilities

Time to patch 6 advisories without patch 34 days to patch on average

Future challenges – Locate the absence of mitigation Simple question: How do you accurately pin- point for the absence of a vulnerability mitigation? Secure by default helps, but can’t be assumed

Locating the absence of CSRF protection

Results 30 plugins downloads

Plugin downloads 455k downloads on average

Future challenges – Vulnerability alerting We know how to look for vulnerabilities You have to keep up with current development Reviewing the same code over and over again stops paying off How do we make a machine look for things to review in real-ish time?

Conclusions Intelligent grep use can uncover vulnerabilities Using your understanding of a language and framework serves as a great starting point for code reviews By reviewing critical code-paths for mechanical vulnerabilities, you can cover a lot of ground in short time Systematic vulnerabilities exist and can be uncovered en masse quite trivially Use root cause analysis appropriately and fix the problem rather than the symptom

Bonus example!

array_merge\(.*\$_(POST|GET|REQUEST)

Press button, receive answers