Learn the Lingo: Data Poisoning

Data Poisoning is the act if intentionally adding false or malicious data into an AI model’s training set to corrupt its learning process. These attacks allow attackers to influence or manipulate the model’s behavior, threatening the accuracy, integrity, and safety of AI symptoms.

Cyber risks include:

Creation of security backdoors for potential future attacks, such as allowing phishing and ransomware to bypass email spam filters.

Degraded performance of the AI, including potential for errors, incorrect decision making, and the introduction of biases.

Reduce risks by:

Implement data validation processes, employing anomaly detection, and performing regular audits of AI models.

Still have questions or need help?