A Bio-Inspired Hybrid Artificial Intelligence Framework for Cyber Security

This article introduces a Bio-Inspired Hybrid Artificial Intelligence Framework for Cyber Security, focusing on enhancing the security of military information systems (IS) through high protection level measures. It emphasizes the critical role of secrecy and confidentiality in maintaining a strategic advantage in military operations. The framework combines three advanced artificial intelligence methods to preemptively detect and mitigate potential cyber threats, including zero-day attacks.

The first subsystem, Hybrid Evolving Spiking Anomaly Detection Model (HESADM), identifies network anomalies and classifies intrusions with a high accuracy of up to 99%, leveraging evolving spiking neural networks (eSNN) and multi-layer feed-forward artificial neural networks (MLFF ANN). The second, Evolving Computational Intelligence System for Malware Detection (ECISMD), focuses on scanning and identifying malicious code in packed executable files, utilizing eSNN for initial classification and an evolving classification function (ECF) for detecting malware, achieving a detection accuracy of 99.87%. The third subsystem, Evolutionary Prevention System from SQL Injection (ePSSQLI), employs an optimized multi-layer feed-forward neural network (MLFF ANN) with genetic algorithms (GA) to detect SQL injection attacks with an accuracy of 99.6%.

The integration of these subsystems within the proposed framework allows for dynamic control and preemptive threat detection in military networks. The article demonstrates the framework’s effectiveness through extensive testing and comparative analysis, showing superior performance compared to existing approaches with high accuracy rates and reduced false alarms. This innovative framework offers a promising solution for military cybersecurity, combining bio-inspired AI techniques to ensure the integrity and confidentiality of information systems. Future directions include exploring more biologically realistic ANN models, improving online learning for ECISMD, and comparing ePSSQLI’s optimization schemes with other methods like particle swarm optimization.