An Adaptive Cyber Threat Detection via Scaled SVM Kernels: A Data-Driven Perspective

The escalating threat of cybercrime in today's globalized world, where digital technologies are ingrained in every aspect of daily life. With digital technologies permeating every aspect of society in today's networked environment. Numerous cybercrimes, including identity theft, phishing, hacking, online fraud, and cyberbullying, put people, companies, and governments at serious risk. Motivations behind cybercrime can range from financial gain and political agendas to personal grudges or simply seeking thrills. To combat cybercrime effectively, prioritizing cybersecurity is imperative across all levels, from individuals to organizations and governments. This involves raising awareness about cybersecurity threats, educating users on safe online practices, and implementing robust security measures. Machine learning also plays a crucial role in bolstering cybersecurity by enabling advanced threat detection and mitigation techniques. This study explores the effectiveness of different scaling methods combined with SVM classifiers using Radial Basis Function, Polynomial and Linear, Polynomial kernels to enhance cybersecurity. NSL-KDD dataset serves as the basis for evaluating these techniques. Results indicate that MaxAbs Scaler paired with SVM-Polynomial and SVM-RBF achieves the highest accuracy at 97.9%. Noteworthy accuracies were also observed with SVM using linear kernel and Standard Scaler, SVM with polynomial kernel and MinMax Scaler, and SVM with RBF kernel and MaxAbs Scaler. These findings underscore the importance of appropriate scaling methods and SVM kernels in optimizing cybersecurity models. The study highlights the potential of MaxAbs Scaler in improving threat detection and mitigation in cybersecurity applications.