[Cryptography] Firewall penetration

Jerry Leichter leichter at lrw.com
Wed Feb 1 07:42:29 EST 2017


> Suppose a server talks to two clients, which are connected to the internet by ordinary consumer type connections.  Is their any reliable, practical, generally useful way whereby it can arrange for the two clients to talk directly to each other, or is it more practical for all data to be stored on the server by one client, and then collected by the other client?
> 
> When last I looked at this issue, direct communication was getting harder, and workarounds were like bugs that were likely to be fixed.
I'm not sure what you're referring to.  We are not yet at the point where "ordinary consumer connections" can't listen for incoming traffic, nor where outgoing traffic to such endpoints is blocked.  Yes, there are some special cases (mainly for mail, to block spam) but it's not a general phenomenon.

There are two common issues.  First, "ordinary consumer connections" don't have static IP's, so finding your target requires something special.  Two solutions are common:  Dynamic DNS, which follows the varying IP address around as it changes; and third-party "rendezvous" sites which come down to the same thing, just effectively using a private namespace separate from DNS.  Some of these "rendezvous" sites may act as proxies, allowing both ends to have outbound connections and simply forwarding the traffic onward; others pass along the needed information and then let the endpoints connect.

The second issue is consumer-level firewalls.  But there are commonly-implemented protocols allowing hosts behind the firewall to create openings through it.

                                                        -- Jerry



More information about the cryptography mailing list