Re: New sensor for Linux deployments
Steve Grubb <[email protected]>
| Newsgroups | gmane.comp.security.ids.prelude.user |
|---|---|
| Message-ID | <[email protected]> |
On Monday 04 February 2008 06:13:23 Sebastien Tricaud wrote: > I can see the auditd sensor offline in the prewikka agent view, maybe > you start prelude with incorrect parameters. You may add: > > prelude_client_set_flags(prelude_client, > prelude_client_get_flags(prelude_client) | > PRELUDE_CLIENT_FLAGS_ASYNC_TIMER); Hmm...that's basically what I'm doing. > To test auditd, I wrote a program to segfault and I got the following > message: > > assessment: > impact: > severity: low (2) > completion: (null) (-1) > type: other (0) > description: An application terminated > abnormally. An attacker may be trying to exploit a weakness in the program. > > Few questions: > - Why do you consider this attack as low ? I guess a program > segfaulting, is (well should be) unusual and get a medium impact severity. I considered it low since the application was in fact terminated. I am willing to make it Medium if that's more appropriate. So, this will be fixed in the next release. > - Is there a reason why the completion is null ? The code "as is" sets the completion to IDMEF_IMPACT_COMPLETION_ERROR and then it looks for evidence in the audit stream that supports a conclusion of succeeded or failed. If it doesn't find support for either conclusion, it leaves it unchanged. I really think that there should be a way to tell admins that the analysis was indeterminate so they can take whatever steps they feel is necessary. Anyways, I just updated it so that its now hardcoded as a failure for this record type. Thanks for the feedback. I'm happy to make any other updates that people feel need changing. Thanks, -Steve _______________________________________________ Prelude-user site list [email protected] http://www.prelude-ids.org/mailman/listinfo/prelude-user