Re: Compression considered harmful

Kristian Gjøsteen wrote:
Bertrand Mollinier Toublet <byrtrxnd.mollznzyrtoublyt@xxxxxxxxxxxxxxxx> wrote:
if variable-size encoded, encrypted data was communicated in fixed-size frames (as far as the observable traffic is concerned), then there really would be no such attack.

Would there be a point to variable-size encoding, then?

Sure, why not: the way the voice signal is encoded is entirely independent of the way to encoded data is propagated on the network.

I suggest you read up on signal- and channel-coding, what they are meant for and how they interact.

To illustrate, though, in this case, the VBR-coding is the signal coding. The variable-bitrate nature of the encoding is a means to achieve as much compression as possible (that's the goal of signal-coding).

The attack you originally mentioned comes from the analysis of the signal captured from the transmission channel (that is, after channel-coding). In this case, obviously, the channel-coding did not have the side-effect of removing the apparency of variable bitrate. Nonetheless, it certainly could while remaining true to its goal (addition of redundancy to protect against loss introduced by the channel).
--
Bertrand
.

Quantcast