Signal's TLS Proxy Failed to Be Probing Resistant

While the authors of this definitely didn't handle this well, I'd argue it's a pretty severe weakness and the tool shouldn't have been released in this state. Active probing has been observed in the wild [1] and pretty much all tooling in the space handles it in their threat model [2], so its naive to not consider it.

I get why the signal team wanted something to use HTTPS, even networks with completely insane firewalls accept it and they get to reuse existing domain fronting code, but existing tools continues to viable in Iran and would have made much more sense in the circumstances.

[1] https://blog.torproject.org/learning-more-about-gfws-active-...

[2] https://github.com/Yawning/obfs4/blob/master/doc/obfs4-spec....

Moxie deserves an ACM award for his contributions to crypto but he shouldn't be leading the project. Maybe posting on the Discourse forum was the right thing to do here. I just see a lot of hostility between Signal employees and those wishing to make the project a little bit better.

How long did he wait for the signal forum to approve his account? The guy, or girl, seems rather aggressive. It's mentioned they've not slept in a while...

My experience with the Signal team in issues and their community has been about the same.

They're generally dismissive, especially so about design problems that cause a big amount of bugs that are strewn throughout Signal, like their handling of message timestamps/sync and dismissal of the IME concerns.

Man I feel bad for the signal devs. Keep fighting the good fight, and thank you!

Wow this guy seems like he's an asshole.

Although i do wonder why signal didn't reuse the work tor did with obfuscated bridges.

For people curious, it looks like they are discussing it in the community but only the mod is involved and I don't think they (Herohtar) are an employee. While the block was unintentional this doesn't seem like the right way to handle the situation and Moxie should have been clearer and that would have avoided the issue. It is easy to interpret the response given as being brushed off.

TLDR: DuckSoft got autobanned because they typed too fast (copy pasted their post into the forum) and had a false positive from spam detection. No comment yet from devs on issue.

Edit: removed my personal preference that seems to be being confused as saying Signal should use GitHub. Signal can do what they want, that's fine. But be clear.

https://community.signalusers.org/t/tls-proxy-server-unable-...

What is that PoC in the issue doing? They check e which is not set after first line in the function:

    func send(addr, server, sni string) int {
     c0, e := net.Dial("tcp", addr)
     if e != nil {
      log.Fatal(e)
     }
    
     c1 := tls.Client(c0, &tls.Config{
      ServerName:         server,
      InsecureSkipVerify: true,
     })
    
     c2 := tls.Client(c1, &tls.Config{
      ServerName:         sni,
      InsecureSkipVerify: true,
     })
     c2.SetDeadline(time.Now().Add(2 * time.Minute))
     s := fmt.Sprintf("GET / HTTP/1.1\r\nHost: %s\r\nUser-Agent: curl/7.68.0\r\n\r\n", sni)
     //b := make([]byte, 4096)
     l, _ := c2.Write([]byte(s))
     log.Println(l)
     if e != nil {
      return 0
     }
     log.Printf("%s->%s->%s\n", addr, server, sni)
     return l
    }

I still think they should have used layer 4 CDN endpoints that use generic names on several of the CDN providers that support L4. It would be endless whack-a-mole to block that. Not perfect, but not perfect is probably useful enough for those impacted by blocks as it would mean periodic latency vs. being locked out entirely. Moxie, if you are reading this, to mitigate some of the probing or fingerprinting, consider borrowing some of the code from sslh [1] and I acknowledge this would be an endless arms race.

[1] - https://github.com/yrutschle/sslh

This seems like a relatively easy issue to fix. If they included a "password" to the proxy (and stuck it on the share URL) then the proxy can reject requests unless the password authentication passed. This way it would look like any other HTTPS site that was password protected. Only if you know the password would you get proof that the other end was connected to Signal.

Things has been updated. Signal banned @studentmain and @ducksoft from their GitHub organzation.