Anti-Backdoor Learning: Training Clean Models on Poisoned Data
–Neural Information Processing Systems
ABL introduces a two-stage gradient ascent mechanism for standard training to 1) help isolate backdoor examples at an early training stage, and 2) break the correlation between backdoor examples and the target class at a later training stage.
Neural Information Processing Systems
Nov-14-2025, 16:40:06 GMT
- Country:
- Asia
- Europe > Denmark
- Capital Region > Copenhagen (0.04)
- North America > United States
- Illinois (0.04)
- Massachusetts > Middlesex County
- Cambridge (0.04)
- Genre:
- Research Report (0.46)
- Industry:
- Information Technology > Security & Privacy (0.53)
- Technology: