Data Poisoning

Also known as: Poisoning

Caesar AI Atlas Definition

Data poisoning is an adversarial attack in which a malicious actor interferes with the data or training process used to build a model. By inserting, modifying, or corrupting training examples, the attacker seeks to manipulate model behavior, degrade performance, or create targeted failures.

Other Definitions

Data Poisoning Source

A type of cyberattack in which a malicious actor intentionally compromises a training dataset used by an AI model to influence or manipulate the operation of that model.

Data Poisoning Source

‘Adversarial attacks in which an adversary interferes with a model during its training stage, such as by inserting malicious training data (data poisoning) or modifying the training process itself (model poisoning).’

Concept Comparisons

Attack / FailureLlm Security

Prompt Injection vs Data Poisoning

A side-by-side comparison of Prompt Injection and Data Poisoning. Understand how runtime instruction manipulation differs from attacks on training data or the training process.