[messaging] Test Data for the Usability Study

Tom Ritter tom at ritter.vg
Mon Jul 7 14:06:49 PDT 2014

On 7 July 2014 04:17, Joseph Bonneau <jbonneau at gmail.com> wrote:
> So am I correct in reading that your main concern is that an attacker able
> to do 2^80 work can't always find an 80-bit match (by which we mean any
> desirable type of match that has a probability of 2^-80 of occurring by
> chance)?

No, it's around what _type_ of match we generate for an attacker's
fingerprint.  Do we flip bits at random, or do we optimize for the
encoded fingerprint type.  The latter introduces a certain amount of
subjectiveness.  (It's quite objective if we do a fuzzy match based on
phonetic differences, it's much-less-subjective-but-still-a-little-bit
if we do a 'leading and trailing n bits' match).

I argued strongly for full subjectivity, then went to random bit
flips, and now I'm back to saying let's do something in-between.


More information about the Messaging mailing list