Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Yet another example of why to both sign release artifacts AND verify them is important.

Also, if you're running the public website for a security lib or core FOSS package, expect more attacks by kiddies trying to build rep... so very conservative tech choices (mostly static website served from a read-only fs) and defensive practices are de rigueur.



What's the use in a static website or a read-only FS when you can overwrite what's in RAM, or just attack routing or DNS? Security is a little more complex.


The point is to minimize attack surfaces. If you're serving static content that's one less path for an attacker to potentially exploit. With only static files exploits are limited to those contained in the web server or the OS network code. With a read only filesystem certain classes of privilege escalation are eliminated.

Attacks on routing or DNS are more difficult to deal with, but at least it isn't your server being compromised, and if you're using HTTPS properly then the certificate should show as invalid at least.

So yeah, security is complex, but his advice was spot on. The fact that you seem to call it into question says that you don't know much about security.


Thanks for also seeing the bigger picture. In the future, I will only submit complete production conf mgmt repos uuencoded in my comments. Readability is overrated.


Are you kidding me?! At the very least suggest Grsec, SELinux, containers! Who gives a shit about "certain classes" of privilege escalation? Are you securing your webserver against 5th graders or actual hackers?

If you want to minimize your attack surface, what he suggested is quite possibly the least effective possible thing anyone could do. I point out just a few of the more important issues to consider first, and you tell ME I don't know about security? I don't know what kind of systems you secure, but mine don't rely on 'mount -o ro,remount /' as a defense strategy.


You're missing the bigger point: enumerating every possible defense is beyond the scope of a comment AND does not exclude any technique by omission. If you'd like to raise technologies in a civil manner, please. Just don't start getting defensive and name calling. [1]

[1] http://ycombinator.com/newsguidelines.html


Ahh yes, the "if you can't turn it into Fort Knox, don't bother at all" guy.

Turning up at every security discussion since 1874.


You can't claim that serving static content from a read-only filesystem is "de rigueur", and that securing your DNS registration is "fort knox". I think the point was that the OP was claiming static content from a read-only filesystem as the solution, and the reply was pointing out that this is hardly the best or first actions.

I would personally expect that the openssl group is suffering some embarrassment, but this sort of hack is a risk of the business. Hopefully we get a good writeup.


The point is that when your attacker has the skill to overwrite your web server's memory to carry out an attack. You never had a chance.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: