Speaker Impersonation Attack on VoxCeleb1 and CNCeleb subsampled (test-enrollment)
100ASR AccuracyAudio-NES
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Audio-NESTarget Model=T1, Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 5,200 | |
| Audio-NESTarget Model=T2, Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 4,300 | |
| Audio-NESTarget Model=T3, Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 8,700 | |
| Audio-NESTarget Model=T4, Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 9,500 | |
| Audio-NESTarget Model=T5, Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 3,800 | |
| Audio-NESTarget Model=Avg., Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 6,300 | |
| YourTTS-NESTarget Model=T1, Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 2,500 | |
| YourTTS-NESTarget Model=T2, Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 3,400 | |
| Ours-NESTarget Model=T1, Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 200 | |
| Ours-NESTarget Model=T2, Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 300 | |
| Ours-NESTarget Model=T3, Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 400 | |
| Ours-NESTarget Model=T4, Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 500 | |
| Ours-NESTarget Model=T5, Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 300 | |
| Ours-NESTarget Model=Avg., Decision Threshold=tau_E, Query Budget=50k2026.03 | 100 | 300 | |
| Audio-NESTarget Model=T1, Decision Threshold=tau_M, Query Budget=50k2026.03 | 100 | 8,500 | |
| Audio-NESTarget Model=T2, Decision Threshold=tau_M, Query Budget=50k2026.03 | 100 | 7,600 | |
| Audio-NESTarget Model=T3, Decision Threshold=tau_M, Query Budget=50k2026.03 | 100 | 16,300 | |
| Audio-NESTarget Model=T4, Decision Threshold=tau_M, Query Budget=50k2026.03 | 100 | 16,100 | |
| Audio-NESTarget Model=T5, Decision Threshold=tau_M, Query Budget=50k2026.03 | 100 | 19,300 | |
| Audio-NESTarget Model=Avg., Decision Threshold=tau_M, Query Budget=50k2026.03 | 100 | 13,600 | |
| Ours-NESTarget Model=T1, Decision Threshold=tau_M, Query Budget=50k2026.03 | 100 | 500 | |
| Ours-NESTarget Model=T2, Decision Threshold=tau_M, Query Budget=50k2026.03 | 100 | 600 | |
| Ours-NESTarget Model=T3, Decision Threshold=tau_M, Query Budget=50k2026.03 | 100 | 900 | |
| Ours-NESTarget Model=T4, Decision Threshold=tau_M, Query Budget=50k2026.03 | 100 | 1,000 | |
| Ours-NESTarget Model=T5, Decision Threshold=tau_M, Query Budget=50k2026.03 | 100 | 800 | |
| Ours-NESTarget Model=Avg., Decision Threshold=tau_M, Query Budget=50k2026.03 | 100 | 800 | |
| YourTTS-NESTarget Model=Avg., Decision Threshold=tau_E, Query Budget=50k2026.03 | 97 | 4,400 | |
| YourTTS-NESTarget Model=T3, Decision Threshold=tau_E, Query Budget=50k2026.03 | 96 | 4,100 | |
| YourTTS-NESTarget Model=T5, Decision Threshold=tau_E, Query Budget=50k2026.03 | 96 | 7,900 | |
| YourTTS-NESTarget Model=T4, Decision Threshold=tau_E, Query Budget=50k2026.03 | 93 | 4,200 | |
| YourTTS-NESTarget Model=T1, Decision Threshold=tau_M, Query Budget=50k2026.03 | 93 | 8,900 | |
| YourTTS-NESTarget Model=T2, Decision Threshold=tau_M, Query Budget=50k2026.03 | 82 | 11,500 | |
| YourTTS-NESTarget Model=T4, Decision Threshold=tau_M, Query Budget=50k2026.03 | 79 | 11,300 | |
| YourTTS-NESTarget Model=Avg., Decision Threshold=tau_M, Query Budget=50k2026.03 | 72.6 | 11,000 | |
| YourTTS-NESTarget Model=T3, Decision Threshold=tau_M, Query Budget=50k2026.03 | 72 | 8,900 | |
| YourTTS-NESTarget Model=T5, Decision Threshold=tau_M, Query Budget=50k2026.03 | 37 | 18,800 |