Quantum-Inspired Feature Engineering for Intrusion Detection Systems

Abstract

Server monitoring is a major part of defending networks from new cyber attacks. Choosing the most useful aspects from an enormous volume of network data is one of the main problems in implementing IDS, since it affects both its success and the time it takes to operate. Traditional ways of choosing which features to use normally have problems moving quickly to the optimal solution, catching already existing good points or working well with large datasets. In our study, we draw from quantum mechanics to enhance the selected features that support the IDS detection process. QPSO locates useful parts of the search space using quantum methods, making the search simpler and keeping important differences in the problem. We look at our framework using NSL-KDD and CICIDS2017 data and compare several machine learning models to understand how features respond to quantum optimization. It is clear from our results that using our new method results in a higher accuracy, fewer incorrect positives and a more dependable system. Because of these findings, we suggest that quantum-inspired optimization may play a part in forming new IDS solutions.