Data Poisoning
An attack method in which malicious samples are mixed into a dataset during training or fine-tuning, causing the model to behave incorrectly under specific trigger conditions. It is commonly examined in security threat analyses of AI systems in critical infrastructure.
Timeline (last 90 days)
Model supply chain risks include the origin and integrity of pre-trained models, fine-tuning data poisoning, inference service access control, and privilege abuse from models being used in automated pipelines.
By 2023, academia had demonstrated multiple attack vectors against RAG systems, including knowledge conflict attacks and backdoor trigger attacks.
All Facts (2)
Model supply chain risks include the origin and integrity of pre-trained models, fine-tuning data poisoning, inference service access control, and privilege abuse from models being used in automated pipelines.
50%UnverifiedBy 2023, academia had demonstrated multiple attack vectors against RAG systems, including knowledge conflict attacks and backdoor trigger attacks.
50%