Boffins devise voice-altering tech to jam 'vishing' schemes

Researchers introduced ASRJam, a proactive defense against AI-driven vishing that uses the EchoGuard algorithm to subtly modify live audio (via reverberation, microphone oscillation, and transient attenuation) so humans can understand it but automated speech recognition systems are disrupted. Tested across multiple datasets and ASR models (including DeepSpeech, Wav2Vec2, Vosk, Whisper, SpeechBrain, and IBM Watson), EchoGuard generally outperforms prior jammers, though Whisper remains more resilient; nevertheless, it still causes meaningful transcription errors that can derail scam interactions. With vishing rising sharply and regulators targeting AI voice misuse, the team plans to improve performance—especially against Whisper—and pursue commercialization.