A Quantum-Inspired Intrusion Detection System for Serverless Computing using QMV4ECX framework

The serverless platform is designed to make it easier to deploy, scale, and manage cloud applications. Unfortunately, it is challenging to directly transfer typical security measures to serverless computing due to its distributed nature. The current serverless solutions mostly use runtime metrics collection, encrypted traffic offloading detection, or post-analysis of modified operating system audit logs to find potential threats. Nevertheless, these techniques frequently fall short of fully identifying cross-functional communication problems. This restriction makes it more difficult to maintain low connection latency and limits real-time log surveillance and verification abilities in dispersed contexts. Therefore, this paper presents an effective deep learning based framework called quantum-inspired MobileNetV4 with Enhanced ConvMiXer (QMV4ECX) for intrusion detection in a serverless computing environment. In this framework, we use Quantum-inspired MobileNetV4 (QMV4) as the feature extractor to gather key characteristics from the network data. The Enhanced ConvMiXer method is then used as the classifier to identify and classify threats. Furthermore, the Wombat Optimization Algorithm (WOA) is utilized to tune the hyperparameters of the classifier. Performance analysis has been conducted using the DoW, MDCD-AINA, and synthetic datasets, and the experimental results show that QMV4ECX can accurately detect a wide range of attacks, enabling thorough function execution tracking and evaluation while drastically reducing performance overhead.