Phonotactic Reconstruction of Encrypted VoIP Conversations [pdf]

cs.unc.edu

6 pointsjs21 comment

Abstract — In this work, we unveil new privacy threats against Voice-over-IP (VoIP) communications. Although prior work has shown that the interaction of variable bit-rate codecs and length-preserving stream ciphers leaks information, we show that the threat is more serious than previously thought. In particular, we derive approximate transcripts of encrypted VoIP conversations by segmenting an observed packet stream into subsequences representing individual phonemes and classifying those subsequences by the phonemes they encode. Drawing on insights from the computational linguistics and speech recognition communities, we apply novel techniques for unmasking parts of the conversation. We believe our ability to do so underscores the importance of designing secure (yet efficient) ways to protect the confidentiality of VoIP conversations.

via https://news.ycombinator.com/item?id=11994286

This is pretty fascinating... the jist is that they were able to extract the conversation from an encrypted VoIP stream from the lengths of the encrypted packets and knowledge of the underlying voice encoding:

The approach we pursue in this paper leverages the correlation between voiced sounds and the size of encrypted packets observed over the wire. Specifically, we show that one can segment a sequence of packet sizes into subsequences corresponding to individual phonemes and then classify these subsequences by the specific phonemes they represent. We then show that one can segment such a phonetic transcript on word boundaries to recover subsequences of phonemes corresponding to individual words and map those subsequences to words, thereby providing a hypothesized transcript of the conversation.

Reply on news.ycombinator.com