The significance of Machine Learning in Enhancing the Security for IoT networks to Detect Attack Signature

Abstract

The Internet of Things (IoT) has contributed to many risks to defence and societal issues. Notwithstanding the social gains, IoT may jeopardise the protection and privacy of individuals and businesses at different levels. The most popular forms of attacks faced by IoT networks involve denial of service (DoS) and distributed dos (DDoS). Companies can use an accurate classification and identification model, which is not a simple job, to fight such assaults. This paper provides a model for the classification of many algorithms for machine learning, i.e. Random Forest (RF), k-Nearest Neighbors (KNN), and Naïve Bayes. The algorithms of the machine learning are used to identify assaults on the UNSW-NB15 dataset. The UNSW-NB15 involves regular network traffic and malicious traffic. The experimental findings indicate that RF and KNN classificators are the highest performers with a 100% accuracy (without injection), 99% (with 10% noise filter), and the Naïve Bayes classification provides the lowest outputs with 95.35% accuracy and 82.77% noise and ten% noise. Additional evaluation matrices, such as accuracy and reminder, also illustrate the utility of RF and KNN classification over Naïve Bayes.