CYBER THREAT DETECTION BASED ON ARTIFICIAL NEURAL NETWORKS USING EVENT PROFILES

Abstract

In this study, we present an artificial intelligence (AI) strategy for identifying cyber hazards that is based on artificial neural networks. We developed an AI-SIEM solution for this project that integrates event profiling for data pre-treatment with a variety of artificial neural network approaches, including LSTM, CNN. The method makes it very easy for security experts to distinguish real false positive signals from positive signals, allowing them to respond to cyber-attacks quickly. The CICIDS2017 and NSLKDD benchmark datasets are used as well as two real-world datasets for each experiment in this research. To compare the performance of the five traditional machine-learning algorithms (DT, NB, RF, k-NN, and SVM), we conducted tests using them. The experimental findings in this research support the usage of the suggested techniques as learning-based network models for intrusion demonstration and detection, showing that they outperform traditional machine learning techniques when used in practical settings