Web Technologies Security in the AI Era: A Survey of CDN-Enhanced Defenses

AI-enhanced security at the network edge improves response times: The paper finds that deploying machine learning-driven defenses at Content Delivery Networks (CDNs) and edge computing locations measurably reduces both "time-to-detect" and "time-to-mitigate" for web attacks, meaning threats are caught and stopped faster than traditional centralized approaches.

New security risks emerge from AI defenses themselves: While AI-powered protections offer benefits, the survey identifies significant governance concerns including "model abuse, poisoning, and governance" challenges—meaning the AI systems protecting websites could themselves become targets for attackers or be manipulated through corrupted training data.

Explainability and transparency gaps need urgent attention: The paper's research agenda explicitly prioritizes "XAI" (Explainable AI), highlighting that current AI security systems often operate as "black boxes" where defenders and organizations cannot fully understand why the system blocked a particular user or request—a critical gap for responsible deployment.

Edge-based AI reduces privacy exposure but requires careful governance: By processing security decisions closer to users rather than sending all traffic to central servers, edge AI "reduces data movement and enhances compliance," yet the paper emphasizes that this approach still requires robust governance frameworks to prevent misuse.

Adversarial robustness remains an unsolved challenge: The survey notes that attackers continuously evolve AI-assisted attacks, and defenses must resist "human-mimicry" attacks, indicating that current AI security systems are vulnerable to adversarial manipulation—a fundamental problem requiring ongoing research.