Data poisoning
Data poisoning: Manipulation of training or continuous-learning data to intentionally alter a model. Attacks can reduce general performance or activate behavior with specific triggers. Provenance, validation and dataset access controls are central countermeasures. Examples include Altered labels, Trigger-based backdoor, Malicious documents in a corpus.
In italiano: Data PoisoningLegge V · Overfitting, bias e collasso dei modelliAttacks can reduce general performance or activate behavior with specific triggers. Provenance, validation and dataset access controls are central countermeasures.
Examples
- Altered labels
- Trigger-based backdoor
- Malicious documents in a corpus