This post has been de-listed
It is no longer included in search results and normal feeds (front page, hot posts, subreddit posts, etc). It remains visible only via the author's post history.
A program I'm testing has a null dereference bug which transfers control to a segv handler. The handler then does some logging (including stack info from the glibc back trace functions).
The null dereference doesn't by itself seem exploitable but from reading references like to CWE-479 it may be possible to use the logging code to corrupt memory, perhaps if there's a way to use multiple signals? Has anyone got any working examples of exploits that use this approach? There are a few online but they're all old.
Subreddit
Post Details
- Posted
- 4 months ago
- Reddit URL
- View post on reddit.com
- External URL
- reddit.com/r/AskNetsec/c...