Preventing Overshadowing Attacks in Self-Jamming Audio Channels

Short-Range Audio Channels Security: Survey of Mechanisms, Applications, and Research Challenges

“…However, the contributions [81] and [4] confuted experimentally the idea that MITM attacks on the audio channel can be always detected because of the human perception of the communication channel.Hence, in [81], the authors demonstrated that systems based on self-jamming cannot fully reliably detect the presence of signals injected by a malicious adversary that is co-located with the victim, i.e., they are not able to face the so-called overshadowing attack. It is worth noting that facing the overshadowing attacks is more challenging than dealing with co-located MITM attacks described in Section III in the context of TFA techniques.…”

“…Further proofs of the effectiveness of the overshadowing attack are discussed in [4], where the authors also propose some countermeasures. The most effective is based on some security symbols, embedded within the transmitted data at a random location, such that the bit-string size of a generic security symbol corresponds to a single symbol in the bitstream.…”

“…• Overshadowing and Active Attacks Vulnerability. As shown by the summary in Table IV, the overshadowing vulnerability discussed in [4] is an issue for all the approaches, independently from the human involvement in the pairing process. Especially in systems based on selfjamming, the overshadowing attack could be deployed by skilled adversaries, leading to the poisoning of the communication channel and a Denial of Service (DoS) on the pairing process.…”

“…Especially in systems based on selfjamming, the overshadowing attack could be deployed by skilled adversaries, leading to the poisoning of the communication channel and a Denial of Service (DoS) on the pairing process. The authors in [4] also described some possible solutions to thwart overshadowing attacks, including physical-layer solutions based on Frequency Hopping Spread Spectrum (FHSS) and Direct Sequence Spread Spectrum (DSSS) techniques, and MAC-layer ones, based on the inclusion of security symbols within the legitimate audio signal. However, the deployability of these solutions within secure device pairing protocols is not straightforward, as all of them require the sharing of secrets between the involved devices (either for the synchronization of the frequency to be extracted among the possible ones, or for the positioning of security symbols within the audio data stream).…”

“…are a more secure communication channel when compared to legacy RF communications. For instance, given that audio channels can be perceived by participating entities, they are usually assumed to be robust against active attacks, since any malicious signal could be heard by the participants as well-and hence the attack could be detected [4]. At the same time, the increasing diffusion and affordability of Machine Learning (ML) techniques have boosted the efficiency of audio signals classification and identification, leading to low-cost and disrupting attacks.…”