Hybrid Anomaly Detection in OS Kernel Interfaces for Power Monitoring Systems Using Fuzzy Logic and Deep Belief Networks

Abstract

Anomaly detection in operating system (OS) kernels is critical for the stability and security of embedded systems, particularly in power monitoring applications. OS kernel behavior is complicated, and typical anomaly detection algorithms frequently fail to detect smaller anomalies, especially in power-sensitive applications where energy efficiency is critical. The goal of this research is to create an effective anomaly detection framework capable of reliably identifying abnormalities in the Chinese OS kernel's behavior through power monitoring, assuring consistent system performance and security. The framework includes several critical steps: First, gather a dataset of system call sequences and power usage logs from the OS kernel. Data pre-processing is utilized to clean and normalize the dataset, ensuring it is formatted consistently for investigation. Feature data extraction is then carried out via the Kernel Principal Component Analysis (Kernel PCA) method that uses such important kernel interaction characteristics as the frequency of system calls and the power consumption behaviour. A novel technique, Fire Hawk Optimizer Fused Fuzzy Logic-Based Deep Belief Networks (Fire-Fuzzy DBN) is a hybrid approach that combines FHO to optimize system parameters, Fuzzy Logic to handle uncertainty in system behavior, and DBNs to extract complex patterns, resulting in a robust, adaptive, and effective solution for detecting kernel anomalies. The outcomes reveal that the proposed Fire-Fuzzy DBN strategy, which was implemented in Python, significantly improves kernel anomaly detection accuracy by 99% over previous techniques. The research data analytics for energy-cost efficient system operation establishes the efficacy of fuzzy testing technology in detecting anomalies in OS kernel interfaces for power monitoring systems, therefore improving embedded system dependability and security.