Machine Learning for Cyber Threat Detection and Prevention in Critical Infrastructure

Abstract

Cybersecurity has been getting a lot of attention lately due to the proliferation of important applications and the exponential rise of data networks and computers. Cybercrimes that are well-planned and ongoing pose a greater threat to the Internet. Because hackers are smart enough to get around all of the conventional security procedures in place to detect and prevent cyberattacks, these measures are worthless. There are a lot of cybersecurity apps that use machine learning (ML) methods. This study proposes an advanced cyber threat detection framework leveraging machine learning techniques on the UNSW-NB15 dataset. The proposed Inception model is benchmarked against conventional classifiers, including Random Forest (RF), k-Nearest Neighbors (KNN), and Multi-Layer Perceptron (MLP). Experimental results demonstrate that the Inception model outperforms existing approaches, achieving an accuracy of 98.40%, precision of 99%, recall of 97.90%, and an F1-score of 98.50%. Comparative analysis highlights its superior capability in threat detection and classification. Furthermore, visualization techniques, including confusion matrices and performance graphs, validate the model’s effectiveness. These results highlight the promise of models based on deep learning to improve cybersecurity by providing an efficient and scalable way to identify and prevent intrusions in real time.