Harnessing AI for Cybersecurity: Performance Evaluation of Machine Learning Techniques in Intrusion Detection Systems

Due to the current development of networked systems and the emergence of global cyber threats, efficient and fast NIDS are required. The suggested article aims to analyse new and improved AI and Machine Learning (ML) frameworks to design and implement efficient Network Intrusion Detection Systems (NIDSs) that accurately detect and differentiate network intrusions. The article assesses several effective algorithms for intrinsic and practical use: Logistic Regression, Random Forest, Feedforward Neural Networks (FNN), K Nearest Neighbours (KNN), and Long Short-Term Memory (LSTM), in terms of actual positive intrusions without creating high positive false intrusions. A comparative analysis is then made to evaluate the models based on precision, recall rate, accuracy, and the model processing rate. This suggested article presents a comparison of the two techniques regarding their effectiveness and applicability of the techniques for implementation in the next generation of NIDS. Although observations showed good performance and improvement regarding class imbalance, they remain key issues affecting such systems' reliability, including high computational demand and model interpretability. The reported results such as accuracy 99.897%, precision 99.894%, recall 99.898%, and F1 Score 99.892% in case of Logistic Regression and accuracy 99.882%, precision 99.876%, recall 99.882%, and F1 Score 99.892% in case of Random Forest from the work contribute to the advancement of fully intelligent NIDS that will improve the system stability to defend and adapting for new emerging threats, hence improving the advance of technologies in the complex and sophisticated era of cyberstalking.