[rohc] Sigcomp: UDVM security issues

Sigcompers,

I've just had a 30-minute chat about the UDVM with Karsten Sohr (if
you don't immediately recognize that name, it's quite googleable).

While I don't believe in 30-minute security analyses, there were two
notable results of this discussion:

1) 6 bytes (48 bits) is too little.  If we assume that a busy proxy
   can hold a million states, it looks like we need 9 bytes (72 bits)
   for a reasonable collision (im)probability.

2) One significant aspect of the DoS invulnerability of the UDVM is
   the assignment of effort values for the instructions.  We have to
   make sure that the complexity of the implementation of each
   instruction is at least in the right class with respect to the
   nominal effort.  While this is trivial for most of the
   instructions, we should have another close look at the table.

Karsten will continue to look at the UDVM and its security issues; he
hopes to have a more serious security analysis done within the next
two weeks or so.

Gruesse, Carsten

_______________________________________________
Rohc mailing list
Rohc@ietf.org
https://www1.ietf.org/mailman/listinfo/rohc