Description
CVE-2024-33900 and CVE-2024-33901 have been assigned by Mitre to this highly professional report.
IMO being able to dump plaintext creds from the process memory, when the database is unlocked, is a normal behavior and not a bug. The reporter also mentions:
When the database is closed or locked the probability is 1 in 10 respectively 4 in 10.
which indeed sounds weird.
My following reasoning is purely hypothetical due to poor details shared by the reporter and me not having reproduced the bug.
When the database gets locked, I expect there will be some allocated memory being freed. That memory might contain data from when the database was unlocked and should be zeroed before being released. Do you think that non-zeroed freed memory could be the cause?