Table 1 Non-adaptive attack evaluation. SRoA denotes the success rate of attack
From: Towards the universal defense for query-based audio adversarial attacks on speech recognition system
Attack | Dataset | SRoA(\(\%\)) | Avg.Queries(n) | Detections | DSR(%) | FSNR (dB) |
|---|---|---|---|---|---|---|
CS | Music-sets | 100.00 | \(\sim\)300 | \(\sim\)3.92 | 98.00 | 7.38 |
DW | Music-sets | 98.00 | \(\sim\)150 | \(\sim\)1.7 | 84.74 | 18.41 |
Average | 100.00 | \(\sim\)225 | \(\sim\)2.81 | 91.37 | 12.90 | |
IRTA | Mini-librispeech | 100.00 | \(\sim\)5000 | \(\sim\)56.00 | 84.00 | 40.97 |
DS | Mini-librispeech | 100.00 | \(\sim\)1000 | \(\sim\)11.00 | 82.50 | 13.02 |
Average | 100.00 | \(\sim\)3000 | \(\sim\)34.00 | 83.25 | 27.00 | |