|
此文章由 sweetheart 原创或转贴,不代表本站立场和观点,版权归 oursteps.com.au 和作者 sweetheart 所有!转贴必须注明作者、出处和本声明,并保持内容完整
Don't mix up prevent attack from external with the daily operation security. I'm not saying my software can prevent a hacker to do anything when that hacker has already gained root access on the system.
A few key things required by Audit department are:
1. Who is on the system and at what time period.
2. When did the bad thing happen.
3. How to determine who did that.
So, if a system is using sudo together with tripwire, and root can't directly login, must be through sudo.
How can this meet the above requirements?
Logs could show when did the user logged in to the system, when did he run sudo, but didn't show for how long, and with sudosh (as you said), it would tell what commands were run, might also tell when, and then when did the user log off. So, my question is if the user did run some commands under sudosh, and quit the sudosh, but not sudo, and then wait for 5min, did some bad things, just before quit from the sudo and after log off. So when in investigation, that user could say that when that bad thing happened, I had already quit sudo so had no privilege to do that bad thing. Remember, sudo only logs when did the user run the command, not when that command finished.
With tripwire, not only scan is very costly (that's expected, WZFileGuard full scan also has the same issue, but WZFileGuard has a scan that's only for detecting traps set by internals, which should be run every hour or even less, and is very light-weight), could contain lots of false alarms, but also it can't prevent root to change the database to hide some bad changes he made to the system config. |
|