Autoencoder-Based Botnet Detection for Enhanced IoT Security

Abstract

The Internet of Things (IoT) has revolutionized various industries by connecting everyday objects to the internet, enabling them to collect and share data. However, the rapid growth of IoT devices has also introduced significant security challenges, including the emergence of botnets that can compromise the integrity and confidentiality of IoT systems. Detecting and mitigating botnet attacks is crucial for ensuring the security and reliability of IoT networks. In this paper, we propose an autoencoder-based approach for botnet detection in IoT environments. Our proposed method leverages the power of autoencoders, which are unsupervised deep learning models capable of learning data representations and detecting anomalies. By training an autoencoder on normal IoT network traffic, we can learn the underlying patterns and features of legitimate device behavior. Subsequently, we use the reconstructed error from the autoencoder to identify deviations from normal traffic and classify them as potential botnet activities. To evaluate the effectiveness of our approach, we conduct extensive experiments using real-world IoT datasets and various botnet attack scenarios. Our results demonstrate that the autoencoder-based botnet detection approach achieves high accuracy and outperforms traditional rule-based and machine learning methods. Moreover, the proposed method exhibits robustness against evolving botnet attack techniques and can adapt to dynamic IoT network environments.