4 ms·
I am wondering if someone who understands statistics better than I do could explain conceptually how encrypted data is distinguishable from compressed data. I
by ilyagr 12y ago
I am wondering if someone who understands statistics better than I do could explain conceptually how encrypted data is distinguishable from compressed data. I always assumed that Shannon's paper says that perfectly compressed data should be indistinguishable from random data (which is indistinguishable from encrypted data). Is mp3 compression not sufficiently perfect? Is my understanding wrong?
Thank you!
P.S. This is a really clever trick!
- MichaelGG 12y agoPretty sure you explained it: it's not perfectly compressed data. Even a small bias is enough to give it away.
- pgeorgi 12y agomedia compression usually isn't perfect. For starters, data is framed with synchronization bits to recover from bad transmissions and the like. Also, audio codecs usually use prepared huffman codebooks for the final step. For most codecs they're defined as part of the format, few (such as vorbis) prepend the codebooks to the data stream so compressors can use different codebooks if they choose to do so. For vorbis there used to be an experimental tool that optimized huffman coding of an existing stream, giving another 4%-or-so compression with otherwise identical data (ie. no quality change). These 4% plus framing are probably the entropy differential they're exploiting here.