CVE-2019-6690: Improper Input Validation in python-gnupg
While playing the excellent Insomni’hack teaser CTF we got sidetracked with a real security vulnerability that wasn’t part of the competition.
We discovered a way to inject data through the passphrase property of the gnupg.GPG.encrypt() and gnupg.GPG.decrypt() methods when symmetric encryption is used.
The supplied passphrase is not validated for newlines, and the library passes
--passphrase-fd=0 to the gpg executable, which expects the passphrase on the
first line of stdin, and the ciphertext to be decrypted or plaintext to be
encrypted on subsequent lines.
By supplying a passphrase containing a newline an attacker can control/modify the ciphertext/plaintext being decrypted/encrypted.
python-gnupg 0.4.3, and maybe earlier versions.
Users should upgrade to 0.4.4.
- 2019-01-19: Vulnerability discovered during Insomni’hack teaser 2019
- 2019-01-20: PoC created
- 2019-01-22: Applied for CVE, vendor notified
- 2019-01-23: CVE-2019-6690 assigned
- 2019-01-23: Vendor responded, fix committed
- 2019-01-24: Vendor released 0.4.4
Proof of Concept
Hypothetical application using successful decryption of data to authenticate a user, and a way to exploit it is available here:
Vulnerability discovered by Alexander Kjäll and Stig Palmquist.
Thanks to @dewaelethom who wrote the CTF challenge.