Hacker News Reader

Good point about old-and-vulnerable versions. That seems like a strong argument in favor of using HTTPS. I’m curious why the Debian folks haven’t thought this a serious issue either. I’m going to try to dig up the discussions.

But the use of a weak signing scheme is just an argument in favor of switching to a better signing scheme. Though I guess it is also real world evidence in favor of defense-in-depth.

View on HN

Version freezing attack can be done even if the client tries to connect over TLS, simply by dropping SYN requests to the update server, or by DNS manipulation, etc.

There's a difference between "failed to contact update server" and "no new updates found", though. And DNS manipulation wouldn't work if certificate pinning is used.

That’s not quite the same attack. What you describe can only keep you from updating at all, freezing you at whatever your current version is. The attack described above would let the MITM choose the exact version to update you to.

Absolutely not: the installer blocks downgrade attacks.

I’m not talking about a downgrade attack. I’m talking about upgrading to a known vulnerable version. You are at version X, attacker upgrades you to known vulnerable version X+1, even though the real latest version X+2 has a fix.

Reply on news.ycombinator.com