Adversarial Example Detection on TinyML MNIST, CIFAR-10, Speech Commands
99.984Detection AccuracyAdvScan
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AdvScanEdge Deployment=Yes, Malicious data Requirements for Training=No, Hardware Agnostic=Yes, Attack Agnostic=Yes, Runtime Detection=Yes, On chip Detection Capability=Yes2026.06 | 99.984 | 2.623 | |
| [53]Edge Deployment=No, Malicious data Requirements for Training=No, Hardware Agnostic=No, Attack Agnostic=Yes, Runtime Detection=Yes, On chip Detection Capability=No2026.06 | 98.74 | — | |
| [52]Edge Deployment=No, Malicious data Requirements for Training=Yes, Hardware Agnostic=Yes, Attack Agnostic=Yes, Runtime Detection=Yes, On chip Detection Capability=No2026.06 | 98.41 | 0.26 | |
| [23]Edge Deployment=No, Malicious data Requirements for Training=Yes, Hardware Agnostic=Yes, Attack Agnostic=No, Runtime Detection=Yes, On chip Detection Capability=No2026.06 | 92.6 | — | |
| [39]Edge Deployment=No, Malicious data Requirements for Training=No, Hardware Agnostic=Yes, Attack Agnostic=Yes, Runtime Detection=Yes, On chip Detection Capability=No2026.06 | 90 | 169 |