AI and edge are two modern technologies that have changed the overall landscape of technologies through their integration in smart city, auto-mobile field, healthcare, industrial automation, etc. Edge computing collects and analyzes data as near to the source as possible, making it possible to get data in real-time, take minimal time in the processing cycle, and alleviate pressure on bandwidth. Nevertheless, the distributed and resource scarce paradigm in edges presents a variety of security and privacy issues to address. These are adversarial attacks, inference threats, malware attacks, and supply chain attacks among others. Similarly, data privacy regulations are still another area of interest when it comes to decentralizing an application architecture.
One of the emerging models to fashion out solutions to these problems is what is known as the zero-trust model which adopts the principle of never trusting and always verifying. Zero-trust does away with the chaotic approach of security where a few channels are deemed secure and protected with a strong firewall while the rest of the network operates with weak security measures, which leaves the network open to intrusions and subsequent data loss. This paper provides a comprehensive survey of security and privacy threat in AI Integrated edge computing identifying how zero trusted security models can be applied to mitigate the threats. Exploring important technical innovations including Federated learning for PRIVACY PRESERVING Artificial Intelligence, ENDE-TO-END ENCRYPTION for secure communication, ANOMALY DETECTION for real-time threat DETERRENCE.
To provide context to theoretical findings, this paper utilizes several industry-specific cases to analyze how zero-trust is applied to practical edge computing use cases. The results show that though the zero-trust provides enhanced security and privacy solutions, some barriers like size compatibility, and resource constraints require more development. Finally, the paper discusses the research implications in continuation of this paper and future research recommendations focusing on lightweight security mechanisms, explainability of AI, policy, and compliance and integration of zero-trust principles with global privacy laws. Thus, highlighting the need to protect the next generation AI-driven edge computing systems, this paper has established zero-trust architecture as indispensable.