VoipLoc: VoIP call provenance using acoustic side-channels

We develop a novel technique to determine call provenance in anonymous VoIP communications using acoustic side-channels. The technique exploits location-attributable information embedded within audio speech data. The victim’s speech is exploited as an excitation signal, which is modulated (acted upon) by the acoustic reflection characteristics of the victim’s location. We show that leading VoIP communication channels faithfully transfer this information between sender-receiver pairs, enabling passive receivers to extract a location fingerprint, to establish call provenance. To establish provenance, a fingerprint is compared against a database of labelled fingerprints to identify a match. The technique is fully passive and does not depend on any characteristic background sounds, is speaker independent, and is robust to lossy network conditions. Evaluation using a corpus of recordings of VoIP conversations, over the Tor network, confirms that recording locations can be fingerprinted and detected remotely with low false-positive rate.