Towards Adaptive Zero Trust Model for Secure AI
Kexin Zhang, Shengjie Xu, Bongsik Shin · 2023
The attack surface of AI is expanding and posing numerous security threats. Threat actors have created new and unknown vulnerabilities against intelligent information systems and critical infrastructure. Secure and adaptive countermeasures are needed when detecting threats and mitigating risks. Zero Trust Model is a cybersecurity paradigm that assumes no user or system should be inherently trusted. It applies the security principle of Least Privilege to network access and Continuous Verification to monitoring and auditing. In this paper, we study the landscape of Secure AI and envision applying the Zero Trust Model to defend against AI threats.