jwt_tool.py is a toolkit for validating, forging, scanning and tampering JWTs (JSON Web Tokens).
Its functionality includes:
- Checking the validity of a token
- Testing for known exploits:
- (CVE-2015-2951) The alg=none signature-bypass vulnerability
- (CVE-2016-10555) The RS/HS256 public key mismatch vulnerability
- (CVE-2018-0114) Key injection vulnerability
- (CVE-2019-20933/CVE-2020-28637) Blank password vulnerability
- (CVE-2020-28042) Null signature vulnerability
- Scanning for misconfigurations or known weaknesses
- Fuzzing claim values to provoke unexpected behaviours
- Testing the validity of a secret/key file/Public Key/JWKS key
- Identifying weak keys via a High-speed Dictionary Attack
- Forging new token header and payload contents and creating a new signature with the key or via another attack method
- Timestamp tampering
- RSA and ECDSA key generation, and reconstruction (from JWKS files)
- ...and lots more!
This tool is written for pentesters, who need to check the strength of the tokens in use, and their susceptibility to known attacks. A range of tampering, signing and verifying options are available to help delve deeper into the potential weaknesses present in some JWT libraries.
It has also been successful for CTF challengers - as CTFs seem keen on JWTs at present.
It may also be useful for developers who are using JWTs in projects, but would like to test for stability and for known vulnerabilities when using forged tokens.
This tool is written natively in Python 3 (version 3.6+) using the common libraries, however various cryptographic funtions (and general prettiness/readability) do require the installation of a few common Python libraries.
(An older Python 2.x version of this tool is available on the legacy branch for those who need it, although this is no longer be supported or updated)
Installation is just a case of downloading the jwt_tool.py file (or git clone the repo).
(chmod the file too if you want to add it to your $PATH and call it from anywhere.)
$ git clone https://github.com/ticarpi/jwt_tool
$ python3 -m pip install termcolor cprint pycryptodomex requests
On first run the tool will generate a config file, some utility files, logfile, and a set of Public and Private keys in various formats.
- To make best use of the scanning options it is strongly advised to copy the custom-generated JWKS file somewhere that can be accessed remotely via a URL. This address should then be stored in
jwtconf.inias the "jwkloc" value. - In order to capture external service interactions - such as DNS lookups and HTTP requests - put your unique address for Burp Collaborator (or other alternative tools such as RequestBin) into the config file as the "httplistener" value.
Review the other options in the config file to customise your experience.
To fix broken colours in Windows cmd/Powershell: uncomment the below two lines in jwt_tool.py (remove the "# " from the beginning of each line)
You will also need to install colorama: python3 -m pip install colorama
# import colorama
# colorama.init()
The first argument should be the JWT itself (unless providing this in a header or cookie value). Providing no additional arguments will show you the decoded token values for review.
$ python3 jwt_tool.py <JWT>
The toolkit will validate the token and list the header and payload values.
The many additional arguments will take you straight to the appropriate function and return you a token ready to use in your tests.
For example, to tamper the existing token run the following:
$ python3 jwt_tool.py eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJsb2dpbiI6InRpY2FycGkifQ.aqNCvShlNT9jBFTPBpHDbt2gBB1MyHiisSDdp8SQvgw -T
Many options need additional values to set options.
For example, to run a particular type of exploit you need to choose the eXploit (-X) option and select the vulnerability (here using "a" for the alg:none exploit):
$ python3 jwt_tool.py eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJsb2dpbiI6InRpY2FycGkifQ.aqNCvShlNT9jBFTPBpHDbt2gBB1MyHiisSDdp8SQvgw -X a
Some options such as Verifying tokens require additional parameters/files to be provided (here providing the Public Key in PEM format):
$ python3 jwt_tool.py eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJsb2dpbiI6InRpY2FycGkifQ.aqNCvShlNT9jBFTPBpHDbt2gBB1MyHiisSDdp8SQvgw -V -pk public.pem
All modes now allow for sending the token directly to an application.
You need to specify:
- target URL (-t)
- a request header (-rh) or request cookies (-rc) that are needed by the application (at least one must contain the token)
- (optional) any POST data (where the request is a POST)
- (optional) any additional jwt_tool options, such as modes or tampering/injection options
- (optional) a canary value (-cv) - a text value you expect to see in a successful use of the token (e.g. "Welcome, ticarpi")
An example request might look like this (using scanning mode for forced-errors):
$ python3 jwt_tool.py -t https://www.ticarpi.com/ -rc "jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJsb2dpbiI6InRpY2FycGkifQ.bsSwqj2c2uI9n7-ajmi3ixVGhPUiY7jO9SUn9dm15Po;anothercookie=test" -rh "Origin: null" -cv "Welcome" -M er
Various responses from the request are displayed:
- Response code
- Response size
- Unique request tracking ID (for use with logging)
- Mode/options used
Here is a quick run-through of a basic assessment of a JWT implementation. If no success with these options then dig deeper into other modes and options to hunt for new vulnerabilities (or zero-days!).
Read the token value to get a feel for the claims/values expected in the application:
$ python3 jwt_tool.py eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJsb2dpbiI6InRpY2FycGkifQ.aqNCvShlNT9jBFTPBpHDbt2gBB1MyHiisSDdp8SQvgw
Run a Playbook Scan using the provided token directly against the application to hunt for common misconfigurations:
$ python3 jwt_tool.py -t https://www.ticarpi.com/ -rc "jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJsb2dpbiI6InRpY2FycGkifQ.bsSwqj2c2uI9n7-ajmi3ixVGhPUiY7jO9SUn9dm15Po;anothercookie=test" -M pb
If any successful vulnerabilities are found change any relevant claims to try to exploit it (here using the Inject JWKS exploit and injecting a new username):
$ python3 jwt_tool.py -t https://www.ticarpi.com/ -rc "jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJsb2dpbiI6InRpY2FycGkifQ.bsSwqj2c2uI9n7-ajmi3ixVGhPUiY7jO9SUn9dm15Po;anothercookie=test" -X i -I -pc name -pv admin
Dig deeper by testing for unexpected values and claims to identify unexpected app behaviours, or run attacks on programming logic or token processing:
$ python3 jwt_tool.py -t https://www.ticarpi.com/ -rc "jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJsb2dpbiI6InRpY2FycGkifQ.bsSwqj2c2uI9n7-ajmi3ixVGhPUiY7jO9SUn9dm15Po;anothercookie=test" -I -hc kid -hv custom_sqli_vectors.txt
Review any successful exploitation by querying the logs to read more data about the request and :
$ python3 jwt_tool.py -t https://www.ticarpi.com/ -rc "jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJsb2dpbiI6InRpY2FycGkifQ.bsSwqj2c2uI9n7-ajmi3ixVGhPUiY7jO9SUn9dm15Po;anothercookie=test" -X i -I -pc name -pv admin
For a list of options call the usage function:
Some options such as Verifying tokens require additional parameters/files to be provided:
$ python3 jwt_tool.py -h
A more detailed user guide can be found on the wiki page.
Head over to the JWT Attack Playbook for a detailed run-though of what JWTs are, what they do, and a full workflow of how to thoroughly test them for vulnerabilities, common weaknesses and unintended coding errors.
- February 2021
- Python 3.x
- [+] Send original token when targeturl present
- [+] Send Query token when targeturl present
- Bugfixes:
- fixed request header split - error on multiple colons
- fixed error in signing syntax:
ec256vses256 - fixed timestamp calculations
- sign with manual private key reinstated
- January 2021
- Python 3.x
- [+] New scan test (re-signing of tokens with common passwords) in 'Playbook' scan mode (
-M pb) - [+] Added new hard-coded secret from CVE-2020-1764 to jwt-common.txt
- December 2020
- Python 3.x
- [+] NEW exploit: blank password in signature (
-X b) - [+] NEW 'bare' mode: return only tokens to stdout - for using with upcoming integrations (
-b) - [+] additional checks in 'Playbook' scan mode (
-M pb) - [+] reordered help options to group similar options
- Bugfixes:
- fixed Playbook scanner glitches
- fixed config file generation issues
- November 2020
- Python 3.x
- [+] NEW exploit: null signature (
-X n) - [+] NEW scanner mode: Inject Common Claims (
-M cc) - [+] additional checks in 'Playbook' scan mode (
-M pb) - [+] multiple custom headers now supported (
-rh) - [+] reflective JWKS URL created automatically in config file - for JKU/Spoof JWKS attacks (
-X s) - [+] checks added for old/incompatible config files
- [+] report on long HTTP response times
- Bugfixes:
- fixed colours not working in Windows cmd/Powershell
- fixed capitalisation issue in config file
- fixed broken null signed kid attacks in ScanModePlaybook()
- October 2020
- Python 3.x
- MAJOR REWRITE: lots more capabilities and new commandline arguments/flags - docs written and guides published
- [+] Send tokens directly to the web application from jwt_tool, and proxy through existing tools (Burp, ZAP, etc.)
- [+] ALL NEW SCANNING MODE!:
- Scan for common vulnerabilities from the JWT Attack Playbook
- Test for error conditions by forcing invalid content-types in claims
- Test for unused valid claims by injection
- [+] Customise your default options in the config file
- [+] Built-in dictionaries and assistive lists to find bugs and misconfigurations
- [+] Logging enabled for all tokens, allowing audit, review and re-tampering of successful requests
- [+] Inject token claims and values on-the-fly across all modes, fuzz values from lists, and bruteforce accepted values
- October 2020
- Python 3.x
- [+] Enabled reading of multiple-level nesting of JSON objects in claims
- Fixed function names and text referencing 'key length' where it should have been 'hash length'
- May 2020
- Python 3.x
- [+] Updated Tamper mode to allow users to input all JSON data types when editing or creating new claims
- To specify a new JSON object just create a new empty object with curly braces: {}
- To create a JSON array add it in directly: ['item1','item2']
- [+] General streamlining and bug squashing
- Fixed missing urlsafe_b64 decoding in validateToken()
- April 2020
- Python 3.x
- [+] Changed Tamper mode to allow users to specify data type when editing or creating new claims
- To specify number, true, false, null just type the relevant value
- To force a string surround the input with double quotes
- e.g. to include a number as a text string enclose in quotes, or leave without if you want it as a number data type
- November 2019
- Python 3.x
- [+] Added ability to provide Private Key for signing in Tamper mode, or via cmdline (
jwt_tool.py [jwt] -S -u URL -pr PRIVKEY.pem) - [+] JWKS exported as a file as well as displayed to screen
- [*] Bonus tip - you can verify the JWKS with JWKS Check option ('jwt_tool.py [jwt] -J -jw JWKSFILE.json')
- November 2019
- Python 3.x
- [+] Fixed tampering when signing with [3] and [4]
- November 2019
- Python 3.x
- [+] Commandline argument processing for automation
- [+] Support for tampering, signing and verifying all JWT documented algorithms:
- RS256/RS384/RS512
- EC256/EC384/EC512
- PS256/PS384/PS512
- [+] EXPLOIT: Injection of self-signed Public Key (CVE-2018-0114)
- [+] Timestamp parsing
- [+] Expiry check
- [+] Timestamp tampering
- [+] Tamper nested JSON in claim values
- [+] Key generation: RSA, ECDSA
- [+] Key reconstruction from JWKS files: RSA, ECDSA
- [+] Key verification from JWKS files
- [+] JWKS file generation from RSA key pairs
- Bugfixes:
- Cleaning up code dead-ends and error conditions
- October 2019
- Python 3.x
- [+] ADD and DELETE keys/claims from head and payload
- [+] New ASCII art(!)
- [+] Added feedback for long dictionaries - every 1 million passwords
- [+] Added advice for using hashcat when dictionary attack fails
- Bugfixes:
- Squashed errors on invalid input
- Patched an issue with dictionary attack, with not UTF-8 words in list.
- October 2019
- Python 3.x
- [+] Fully converted to Python 3
- [+] Improved menu
- [+] Improved workflow
- [+] Groundwork for some new features coming soon...
- October 2019
- Python 2.x
- Bugfixes:
- Corrected the alg=none issue by adding a non-capitalised version to output
- Fixed excessive load times when using a large dictionary file
- Other minor tweaks
- June 2018
- Python 2.x
- [+] Create new header claims
- [+] Sign with key file (kid)
- [+] Check signature against key file (kid)
- [+] Output as standard and URL-safe tokens
- Bugfixes:
- Fix broken base64 decoding when certain ASCII characters are present
- Fix broken signature checking/brute-forcing on URL-safe tokens
- Many other minor tweaks and improvements
- July 2017
- Python 2.x
- [+] Signature recognition
- [+] Support for HS384, HS512
- [+] EXPLOIT: RSA Public Key mismatch vulnerability (key confusion)
- [+] Improved dictionary attack routine
- January 2017
- Initial release
- Python 2.x
- Tamper existing claims
- EXPLOIT: test for alg:none vulnerability
- Check HS256 key
- Crack with HS256 dictionary attack
Regex for finding JWTs in Burp Search
(make sure 'Case sensitive' and 'Regex' options are ticked)
[= ]eyJ[A-Za-z0-9_-]*\.[A-Za-z0-9._-]* - url-safe JWT version
[= ]eyJ[A-Za-z0-9_\/+-]*\.[A-Za-z0-9._\/+-]* - all JWT versions (higher possibility of false positives)
-
JWT Attack Playbook (https://github.com/ticarpi/jwt_tool/wiki) - for a thorough JWT testing methodology
-
A lot of the initial inspiration for this tool comes from the vulnerabilities discovered by Tim McLean.
Check out his blog on JWT weaknesses here: https://auth0.com/blog/critical-vulnerabilities-in-json-web-token-libraries/ -
A whole bunch of exercises for testing JWT vulnerabilities are provided by Pentesterlab (https://www.pentesterlab.com). I'd highly recommend a PRO subscription if you are interested in Web App Pentesting.
PLEASE NOTE: This toolkit will solve most of the Pentesterlab JWT exercises in a few seconds when used correctly, however I'd strongly encourage you to work through these exercises yourself, working out the structure and the weaknesses. After all, it's all about learning...