Performance and Scalability Analysis of AI-Driven Cyber-Secure Assessment Platforms

The rapid growth of digital learning has put a strain on scalable and secure online exam systems that will be able to perform in a steady fashion when there are large numbers of concurrent users. Proctoring using Artificial Intelligence (AI) can enhance academic integrity by providing continuous monitoring, behavioral analysis and real-time surveillance, but, because AI inferences are persistent, the system is likely to be limited in its scalability and reliability. This paper presents a performance and scalability analysis of an AI-adapted cybersecure online examination system, which is developed based on a modular full-stack design, stateless backend services, asynchronous processing, and container deployments. Standard load-testing and benchmarking methods were used to carry out experimental analysis at the different workloads. The core performance mesures that were evaluated including reaction time, processing capacity, resource use and Ai inference delay. Result shows that the proposed system will have low latency and high throughput with constant AI-based monitoring at low to moderate concurrency and show the predictable performance degradation at the high loads. The result indicates the balance of the inconsistent AI monitoting and expandability and also offers actionable information in the development of efficient, safe and scalable web based assessment platforms.