Cyberattack Detection Using Unsupervised Learning Techniques

Abstract Cybersecurity threats are becoming increasingly sophisticated and challenging to detect, making it critical to develop effective and robust detection mechanisms. Unsupervised learning techniques offer a promising approach for cyberattack detection, particularly in scenarios where labeled data is scarce or not available. In this research, we propose a novel approach for cyberattack detection using adversarial autoencoders, which combine the traditional autoencoder architecture with an additional adversarial component. The proposed approach involves data collection and preprocessing, training an adversarial autoencoder model using unsupervised learning with adversarial training, detecting anomalies, and evaluating the performance of the approach. The proposed approach offers several benefits, including the ability to capture complex relationships in the data and generate realistic reconstructions of normal data. The effectiveness and potential of the proposed approach will be evaluated through extensive experiments and comparisons with existing unsupervised learning techniques or other benchmark methods.