Cybersecurity definition
What is Model Poisoning?
By Cyber Tool Stack Editorial TeamUpdated
Definition
Deliberate manipulation of training data, model weights, fine-tuning inputs, or related artifacts so a model learns hidden, biased, or malicious behavior.
Where Model Poisoning fits in the security landscape
These beginner lessons use this term while explaining the surrounding security control.
Related cybersecurity terms
Model ExtractionAn attack that uses repeated queries, stolen artifacts, or other access to reproduce a model's behavior or recover valuable information about its parameters and intellectual property.AI SecurityThe discipline of identifying, testing, governing, and reducing security risks in machine-learning models, generative-AI applications, and autonomous agents across their development and operating lifecycle.