Skip to main content

COMPEL Glossary / attack-surface

Attack Surface

The attack surface of an AI system encompasses all the points where an unauthorized actor could attempt to access, manipulate, or extract data from the system, including model API endpoints, training data pipelines, model weight storage, user interfaces, supply chain components, and the inference process itself.

What this means in practice

AI systems have a larger and more complex attack surface than traditional software because they introduce novel vulnerabilities such as model extraction attacks, data poisoning, prompt injection, and adversarial input manipulation. For organizations, understanding and minimizing the attack surface is essential for protecting both the AI system's integrity and the sensitive data it processes. In COMPEL, attack surface analysis is part of the AI Security Architecture framework in Module 3.3, Article 5, conducted during the Calibrate assessment and addressed during Produce.

Why it matters

AI systems have a larger and more complex attack surface than traditional software, introducing novel vulnerabilities like model extraction, data poisoning, prompt injection, and adversarial input manipulation that conventional security approaches do not address. Organizations that fail to understand and minimize their AI attack surface carry hidden security risks that can compromise both system integrity and sensitive data. Proactive attack surface analysis is essential before deploying AI in any production environment.

How COMPEL uses it

Attack surface analysis is conducted during the Calibrate stage as part of the Technology pillar's security assessment and addressed during Produce where security controls are implemented. The Model stage designs defense architectures that minimize the attack surface through input validation, access controls, and model protection mechanisms. The Evaluate stage tests the effectiveness of attack surface reduction through security validation and penetration testing.

Related Terms

Other glossary terms mentioned in this entry's definition and context.