News for hackers. Stuff that matters.
Re:Sidejack Prevention by jgrahamc on Wednesday October 27, 2010 @04:30AM (#1837788)
Not to diss the OP but most of the proposed solutions to FireSheep I've seen display a distinct lack of knowledge of security. And that's one of those fields where enthusiastic amateurs tend to get things badly wrong.

The solution to FireSheep is something akin to SSL (robust, fast encryption with authentication). Could be SSL, or a VPN, or an SSH tunnel doing SOCKS.

Beware the MITM attack if you think all you need is encryption without authentication.

Re:Sidejack Prevention by barrkel on Wednesday October 27, 2010 @04:53AM (#1837820)
I think engineers (and computer scientists, perhaps even more so) have an unduly binary view of security.

If you are specifically targeted, you're probably screwed no matter how good your security is. If the attack isn't getting through remotely, it may escalate into social engineering, or even into physical attack, up to and including physical keyboard bugs and monitor signal sniffing.

If you're not specifically targeted, and are merely part of the crowd, you don't have so much to worry about, because you can play the numbers game. All you need to be is more secure (a couple of standard deviations) than the average. In these circumstances, security through obscurity is statistically measurable: yes, in the binary sense it's wrong to think of obscurity giving you security, but in the practical sense, it measurably decreases your risk of being compromised.

Re:Sidejack Prevention by nikcub on Wednesday October 27, 2010 @05:12AM (#1837855)
Honorable mention to client certificates which I have implemented in two projects now where security and user authentication were important.

Means that the client is not only authenticating the server, but the server is also authenticating the client.

Client cert interfaces in browsers are fairly user friendly, and github already has user keys, so they may as well make it an account option.

Re:Sidejack Prevention by mseebach on Wednesday October 27, 2010 @07:14AM (#1838080)
What's wrong with GitHubs solution? As I understand it, only insensitive stuff is available in the unencrypted session, while writes and sensitive stuff goes via unsidejackable HTTPS. Stealing a read-only Facebook/Twitter session is much less critical, and allows for CDNs etc.
Re:Sidejack Prevention by nopal on Wednesday October 27, 2010 @07:57AM (#1838164)
Can someone still sidejack a session and view a private repo?
Re:Sidejack Prevention by brown9-2 on Wednesday October 27, 2010 @08:05AM (#1838178)
The secure cookie is meant to prevent this.
Re:Sidejack Prevention by nopal on Wednesday October 27, 2010 @08:33AM (#1838247)
Yes, I see this right in the post now. I completely overlooked that SSL was used for browsing private repositories.
Re:Sidejack Prevention by IgorPartola on Wednesday October 27, 2010 @06:58AM (#1838052)
Once again VPN and SSH tunnels will only solve a certain class of problems. End-to-end encryption is a good start. Restricting the session token to only one/few IP addresses could help solve a small subset of issues. Server authentication is a good thing to have as well. What am I missing?