Re: [Monotone-devel] Monotone Security

[Daniel Carrera]

Jack Lloyd wrote:
> That could easily happen due to a time change, though:

Yeah, and a malicious attacker could make the bad revisions children of 
a very old revision. So checking that dates are sequential is useless.

Monotone already has a way to deal with the DOS attack that Peter found 
(10 million encumbered revisions) but it requires a custom script. Maybe 
it'd be easier to just ship Monotone with a Lua script that removes the 
bad key. It's not used by default, but if/when Peter's DOS attack 
happens, the developers can run the script:

function recovery_from_compromised_keys(key) {

    foreach (head) {
        if (head is signed with bad key) {
            1. find the first ancestor revision of that head
            that is signed by a good key.
            2. delete every descendant of that revision.
        }
    }
}



Daniel.