Securing IoT Networks for Detection of Cyber Attacks using Automated Machine Learning

Abstract

Cybercriminals are always developing innovative strategies to confound and frustrate their victims. Therefore, maintaining constant vigilance is essential if one wishes to protect the availability, confidentiality, and integrity of digital systems. Machine learning (ML) is becoming an increasingly powerful technique for doing intelligent cyber analysis, which enables proactive defenses. Machine learning (ML) has the potential to thwart future assaults by studying the recurring patterns that have already been successful. Nevertheless, there are two significant drawbacks associated with the utilization of ML in security analysis. To begin, the most advanced machine learning systems have significant problems with their computing overheads. Because of this constraint, firms are unable to completely embrace ML-based cyber strategies. Second, in order for security analysts to make advantage of ML for a wide variety of applications, they will need to develop specialized frameworks. In this study, we aim to put a numerical value on the degree to which a hub can improve the safety of an ecosystem. Typical cyberattacks were carried out on an Internet of Things (IoT) network located within a smart house in order to validate the hub. Further investigation of the intrusion detection system's (IDS) resistance to adversarial machine learning (AML) assaults was carried out. In this method, models can be attacked by supplying adversarial samples that attempt to take advantage of the defects in the detector that are present in the pre-trained model.