In a previous post I was wondering if the whole concept of centralization is obsolete or has major flaws. The problem that is bothering everybody in the last years is that clients can be used ( unwillingly ) as tools in distributed denial of service attacks. Jan Vidar Krey is proposing a hub refferal on c-c connections that can point to a source of CTM attacks via the messages that the client sends on first connection attempt. In this case an attacked entity can see the hub with problems/intentional flooding that is causing the attacks.
As a first step to prevent this kind of abuses in DC++, poy added a static IP protection for the major hublists that were attacked via the client. This kind of measure is just temporary since hublists can change IP anytime and it protects only them, not everybody else that can be attacked ( Also the fun part is that the hublist server is actually running a DC client and wants to download from other users, it can’t ! ) . A second step was to dynamically resolve the hublist ip’s and block them for c-c connections.
The main idea that I considered is to practically check all the users on a specific hub to see if they actually are real. On CTM receive the client should not connect but send another CTM to see if that IP actually connects to them . This will make sure that the user is the actual owner of that specific IP address. Of course the biggest problem is if the user is passive, in which case it can’t send a CTM back. This could be against the protocol principles but it’s a solution to see if the other peer really exists. I don’t know if a RCM would do something good in this situation but it’s a start.
Another thing that should be done ( if not implemented already ) is that on c-c connections if the first attempt was unsuccessful then no further attempts should be done until the user at least reconnects or changes state ( passive/active ). Also the hubs should be trustworthy. In a previous post I suggested a way to make hubs trustful via a CA authority system, but most people were quite reticent about it. Perhaps this could be the only way to make hubs trustful. Warning messages will not help too much ( Strong DC implements such messages ) since most of the users either don’t read them or don’t care. We shouldn’t let users question this problem, but solve it for them. Continuous problems from the Direct Connect network might be a cause to mark DC software ( and DC++ ) as badware, which will definitely take down the network. It’s time to do something about it.
I’m hoping for more ideas how to make DC++ proof against CTM abuses and I’m waiting for opinions from you as well.
