Comparative Study of Self-Organizing and Elastic Maps in Comparison with Soft Computing Paradigms in Intrusion Detection Systems

Modern cybersecurity faces increasing challenges from sophisticated cyberattacks that traditional detection systems fail to identify. While machine learning offers solutions, classical Self-Organizing Maps suffer from fixed topology and sensitivity to data noise. This paper proposes a hybrid intrusion detection architecture integrating novel Elastic Maps with soft computing classifiers. The Elastic Map adapts its structure dynamically to data density, overcoming limitations of standard neural networks. Comparative analysis across diverse datasets demonstrates that the proposÒed method achieves faster convergence and superior stability in noisy environments compared to classical approaches. The hybrid model significantly enhances detection accuracy and reduces false positives without relying on specific attack signatures. These findings confirm that adaptive topological mapping provides a robust and efficient framework for securing corporate networks against diverse threats.