Supervised Machine Learning Method for Anomaly Detection

Abstract

Cybersecurity is an essential topic, as most of our daily activities are controlled by web applications. These applications become susceptible to various threats that lead to unauthorized access to personal data. Therefore, protecting application data has become essential. Supervised machine learning is widely utilized in various applications, such as spam detection; it functions as a powerful tool for automating decision-making and producing predictions based on historical data. This study employs supervised machine learning to classify anomalies in a network using the NSL-KDD dataset, which is utilized to assess intrusion detection techniques. This dataset contains no repeated items in the training subset, making the approach impartial to any particular items. This research utilizes approaches such as CNN, LSTM, hybrid CNN-LSTM, RBFN, MLP, and SVM. Evaluating multiple algorithms and analyzing their results to select the most efficient option is typically a wise strategy. The results of the implemented models were evaluated and compared based on detection rate, time efficiency, and accuracy. The findings demonstrate that the CNN-LSTM hybrid model exceeded the benchmark methods, with a detection rate of 99.61% and an accuracy of 99.8%.