What’s Our Software Doing With All That User Input

What are we doing with all the characters that get shoved into our applications? Have we considered every potential execution context?
It’s often interesting and surprising to see what sort of concoction of characters can be executed in different places… and linking multiple attack vectors together which the builders haven’t thought about.
What are we trusting? Why are we trusting it? What, where and how should we be sanitising?

We have a vast collection of libraries, techniques, cheat sheets, tutorials, guides and tools at our disposal.
I often find myself thinking… how can we commoditise the sanitisation of user input and I keep coming up with the same answer.
It’s not easy. Every application has a completely different set of concerns.

In order for our software to be shielded from an attack, the builders must think like attackers.

In this talk I’ll attempt to:

Increase our knowledge and awareness
Discuss practical techniques and approaches that increase our defences
Break some software

Transcript

1. Dealing with User Input Securely Kim Carter – OWASP Day 2013-09-12
2. Demonstrate vulnerabilities Increase knowledge, awareness and desire to test Discuss practical techniques and approaches that increase our defences Agenda
3. Why the hacker always has the advantage Learn to enjoy breaking your own software. It'll make you a better developer. Our builders must think like breakers Developers Day Job Write Code Hackers Day Job Break Code
4. What does Poor Sanitisation look like?
5. OWASP ZAP also has a REST API. Useful for regression test suites If we have time at the end, we'll go over some AJAX XSS
6. Quality What is Quality? Do we as builders care? Why we should care
8. Quality But increasing quality is expensive right?
9. Quality Not necessarily
10. My Philosophy on Quality Everyone on the team needs to be thinking about it. Not just the testers. Reducing faults much earlier in the cycle.
11. User Input Sanitisation Strategies All code should be driven by executable specifications. Especially sanitisation logic Based around my following two blog posts http://blog.binarymist.net/2012/11/04/sanitising-user-input-from-browser-part-1/ http://blog.binarymist.net/2012/11/16/sanitising-user-input-from-browser-part-2/ Main components were a WCF service which dished up XSL'd XML as HTML to an existing web app
12. User Input Sanitisation Strategies Threat modelling Defence in depth Minimising attack surface Field length validation, incl structured data Parametrised Queries / Prepared Statements Least privilege White lists How to escape untrusted data for the different execution contexts File uploads not covered Why bother with client side Leveraging existing libraries
13. Threat modelling Ideally performed at design time Identify the real risks. How? Decomposition Determine entry points, assets, trust levels of users Analyse dependencies Determine & rank threats Determine security controls to prevent threats
14. Defence in depth Multiple layers may seem redundant Think of each layer as the only layer Attempt to stop the attack as soon as possible User Interface (Mark-up, JavaScript, CSS) Client – Server Comms Server side (internet facing) Back end code Data store
15. Minimising attack surface Field length validation (client side)
16. Minimising attack surface Field length validation (server side)
17. Minimising attack surface Constrain fields to well structured data. Dates, post codes, e-mail addresses, check boxes, radio buttons Minimise free-form text input Hard to create small white lists with free-form
18. Parametrised Queries / Prepared Statements Least privilege
19. White lists Decide which characters are essential for each input Can now use the HTML5 pattern attribute on input tag. Doesn't cover textareas
20. Client Side 1.type the characters in 2.[ctrl]+[v] characters in clipboard 3.right click -> Paste
21. Server Side
22. Escaping Escape all characters depending on potential execution contexts they may end up in. Even if they are not in your white lists Get away with the following escaping example only if you deal with untrusted data in HTML elements and you're sure your attributes are all quoted Escaping details for additional contexts here: https://www.owasp.org/index.php/XSS_%28Cross_Site_Scripting%29_Prevention_Cheat_Sheet
23. Client Side
24. Server Side
25. Why bother with client side User Experience Server side sanitisation can be a lot slower When an honest user submits their data, they're not going to get server side exceptions due to validation
26. Leveraging existing libraries Useful • OWASP Encoding Project (Reform library) Supports Perl, Python, PHP, JavaScript, ASP, Java, .NET • OWASP Enterprise Security API Not so Useful • Microsoft Anti-Cross Site Scripting Library A lot more detail on my blog blog.binarymist.net
27. Using: http://google-gruyere.appspot.com/ Stored XSS via AJAX
28. When the user clicks refresh button, response looks like In the mark-up the snippet looks like:
29. Resources Threat Modelling • https://www.owasp.org/index.php/Application_Threat_Modeling • https://www.owasp.org/index.php/Threat_Risk_Modeling Cheat Sheets and Check Lists I found helpful • https://www.owasp.org/index.php/Input_Validation_Cheat_Sheet • https://www.owasp.org/index.php/OWASP_Validation_Regex_Repository • https://www.owasp.org/index.php/XSS_(Cross_Site_Scripting)_Prevention_Cheat _Sheet • https://www.owasp.org/index.php/DOM_based_XSS_Prevention_Cheat_Sheet • https://www.owasp.org/index.php/OWASP_AJAX_Security_Guidelines