Older APIs versions tend to be more vulnerable and they lack security mechanisms.
Leverage the predictable nature of REST APIs to find old versions.
Saw a call to api/v3/login
? Check if api/v1/login
exists as well. It might be more vulnerable.
Never assume there’s only one way to authenticate to an API!
Modern apps have many API endpoints for AuthN: /api/mobile/login
| /api/v3/login
| /api/magic_link
; etc..
Find and test all of them for AuthN problems.
Remember how SQL Injections used to be extremely common 5-10 years ago, and you could break into almost every company? BOLA (IDOR) is the new epidemic of API security. As a pentester, if you understand how to exploit it, your glory is guaranteed.
Learn more about BOLA : https://medium.com/@inonst/a-deep-dive-on-the-most-critical-api-vulnerability-bola-1342224ec3f2
Testing a Ruby on Rails App & noticed an HTTP parameter containing a URL? Developers sometimes use "Kernel#open" function to access URLs == Game Over. Just send a pipe as the first character and then a shell command (Command Injection by design)
Learn more about the open function: https://apidock.com/ruby/Kernel/open
Found SSRF? use it for:
- Internal port scanning
- Leverage cloud services(like 169.254.169.254)
- Use http://webhook.site to reveal IP Address & HTTP Library
- Download a very large file (Layer 7 DoS)
- Reflective SSRF? disclose local mgmt consoles
Mass Assignment is a real thing. Modern frameworks encourage developers to use MA without understanding the security implications. During exploitation, don't guess object's properties names, simply find a GET endpoint that returns all of them.
A company exposes an API for developers? This is not the same API which is used by mobile / web application. Always test them separately. Don't assume they implement the same security mechanisms.
Pentest for REST API? Give it a chance and check if the API supports SOAP also. Change the content-type to "application/xml", add a simple XML in the request body, and see how the API handles it.
Sometimes the authentication is done in a different component that is shared between REST & SOAP APIs == SOAP API may support JWT
If the API returns stack trace with a DUMPling, it's probably vulnerable**
Pentest for APIs? Trying to find BOLA (IDOR) vulnerabilities? IDs in the HTTP bodies/headers tend to be more vulnerable than IDs in URLs. Try to focus on them first.
Exploiting BFLA (Broken Function Level Authorization)?
Leverage the predictable nature of REST to find admin API endpoints!
E.g: you saw the following API call GET /api/v1/users/<id>
Give it a chance and change to DELETE / POST to create/delete users.
The API uses Authorization header? Forget about CSRF! If the authentication mechanism doesn't support cookies, the API is protected against CSRF by design.
Testing for BOLA (IDOR)?
Even if the ID is GUID or non-numeric, try to send a numeric value.
For example: /?user_id=111
instead of user_id=inon@traceable.ai
Sometimes the AuthZ mechanism supports both and it's easier the brute force numbers.
*Use Mass Assignment to bypass security mechanisms. E.g., "enter password" mechanism:
POST /api/rest_pass
requires old password.PUT /api/update_user
is vulnerable to MA == can be used to update pass without sending the old one (For CSRF)*
Got stuck during an API pentest? Expand your attack surface! Find sub/sibling domains using http://Virustotal.com & http://Censys.io. Some of these domains might expose the same APIs with different configurations/versions.
Static resource==photo,video,.. Web Servers(IIS, Apache) treat static resources differently when it comes to authorization. Even if developers implemented decent authorization, there's a good chance you can access static resources of other users.
Even if you use another web proxy, always use Burp in the background. The guys at @PortSwigger are doing a really good job at helping you manage your pentest. Use the “tree view” (free version) feature to see all API endpoints you’ve accessed.
Mobile Certificate Pinning? Before you start reverse engineering & patching the client app, check for both iOS & Android clients and older versions of them. There's a decent chance that the pinning isn't enabled in one of them. Save time.
Companies & developers tend to put more resources (including security) into the main APIs.
Always look for the most niche features that nobody uses to find interesting vulnerabilities.
POST /api/profile/upload_christmas_voice_greeting
Which features do you find tend to be more vulnerable? I'll start:
- Organization's user management
- Export to CSV/HTML/PDF
- Custom views of dashboards
- Sub user creation&management
- Object sharing (photos, posts,etc)
Testing AuthN APIs? If you test in production, there's a good chance that AuthN endpoints have anti brute-force protection. Anyhow, DevOps engineers tend to disable rate limiting in non-production environments. Don't forget to test them :)
A good example of this issue: Facebook Breach (Found by @sehacure) http://www.anandpraka.sh/2016/03/how-i-could-have-hacked-your-facebook.html
Got stuck during an API pentest? Expand the attack surface! Use http://archive.com, find old versions of the web-app and explore new API endpoints. Can't use the client? scan the .js files for URLs. Some of them are API endpoints.
APIs tend to leak PII by design.
BE engineers return raw JSON objects and rely on FE engineers to filter out sensitive data.
Found a sensitive resource (e.g, receipt
)? Find all the EPs that return it: /download_receipt
,/export_receipt
, etc..
Some of the endpoints might leak excessive data that should not be accessible by the user.
This is an example for OWASP Top 10 For APIs - #3 - Excessive Data Exposure
Found a way to download arbitrary files from a web server? Shift the test from black-box to white-box. Download the source code of the app (DLL files: use IL-spy; Compiled Java - use Luyten) Read the code and find new issues!
Got stuck during an API pentest? Expand your attack surface! Remember: developers often disable security mechanisms in non-production environments (qa/staging/etc); Leverage this fact to bypass AuthZ, AuthN, rate limiting & input validation.
Found an "export to PDF" feature? There's a good chance the developers use an external library to convert HTML --> PDF behind the scenes. Try to inject HTML elements and cause "Export Injection".
Learn more about Export Injection: https://medium.com/@inonst/export-injection-2eebc4f17117
Looking for BOLA (IDOR) in APIs? got 401/403 errors? AuthZ bypass tricks:
- Wrap ID with an array
{“id”:111}
-->{“id”:[111]}
- JSON wrap
{“id”:111}
-->{“id”:{“id”:111}}
- Send ID twice
URL?id=<LEGIT>&id=<VICTIM>
- Send wildcard
{"user_id":"*"}
In some cases, the AuthZ mechanism expects a plain string (an ID in this case), and if it receives a JSON instead it won't perform the AuthZ checks. Then, when the input goes to the data fetching component, it might be okay with a JSON instead of string(e.g: it flattens the JSON)
BE Servers no longer responsible for protecting against XSS.
APIs don't return HTML, but JSON instead.
If API returns XSS payload? -
E.g: {"name":"In<script>alert(21)</script>on}
That's fine! The protection always needs to be on the client side
Pentest for .NET apps? Found a param containing file path/name? Developers sometimes use "Path.Combine(path_1,path_2)" to create full path. Path.Combine has weird behavior: if param#2 is absolute path, then param#1 is ignored.
Learn more: https://www.praetorian.com/blog/pathcombine-security-issues-in-aspnet-applications
APIs expose the underlying implementation of the app. Pentesters should leverage this fact to better understand users, roles, resources & correlations between them and find cool vulnerabilities & exploits. Always be curious about the API responses.
Got stuck during an API pentest? Expand your attack surface! If the API has mobile clients, download old versions of the APK file to explore old/legacy functionality and discover new API endpoints.
Remember: companies don’t always implement security mechanisms from day one && DevOps engineers don’t often deprecate old APIs. Leverage these facts to find shadow API endpoints that don’t implement security mechanism (authorization, input filtering & rate limiting)
Download old APK versions of android apps: https://apkpure.com
Found a limit
/ page
param? (e.g: /api/news?limit=100
) It might be vulnerable to Layer 7 DoS. Try to send a long value (e.g: limit=999999999
) and see what happens :)