Machine Learning-Based Optimization of Network Security Protocols for Enhanced Threat Detection

Current cyber threats are relatively sophisticated, and this has put pressure on traditional network security protocols. It proposes an extended machine learning methodology using a supervised Random Forest algorithm for classification and an unsupervised Autoencoder for anomaly detection to improve network security measures. The consumes network traffic through a powerful preprocessing pipeline of feature selection, scaling, and encoding to increase the performance of a model. The proposed method was also tested on benchmark datasets like CICIDS2017, UNSW-NB15, and KDDCup99 with an accuracy of 98.76%, precision of 96.45%, false positive rate under 1.89%, and a latency of only 4.23ms better than other methods from 2021, 2022 and 2023. These results show that the ability of the developed framework is higher in terms of detection and more adaptable to real-world threats. It offers a transportable and secure solution for next-generation networks, which overcomes the gaps of the previous methods to make way for further developments.