External memory extraction attack detection on Pokemon
1AUROCControlNet
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ControlNetAttack=RAG-Thief2026.02 | 1 | 100 | 100 | 0.05 | |
| MemPotAttack=RAG-Thief2026.02 | 1 | 98 | 100 | 0 | |
| ControlNetAttack=DGEA2026.02 | 1 | 100 | 100 | 0.05 | |
| MemPotAttack=DGEA2026.02 | 1 | 100 | 100 | 0 | |
| MemPotAttack=IKEA2026.02 | 0.99 | 67 | 98 | 0 | |
| AgentAttack=RAG-Thief2026.02 | 0.94 | 16 | 100 | 0.91 | |
| AgentAttack=DGEA2026.02 | 0.94 | 16 | 100 | 0.93 | |
| ControlNetAttack=IKEA2026.02 | 0.88 | 2 | 56 | 0.04 | |
| AgentAttack=IKEA2026.02 | 0.2 | 0 | 13 | 0.95 |