Security Measures at Network Layer of IoT Using Machine Learning Techniques

Abstract— The world is changing dramatically from discrete systems to ubiquitous "things" with Internet connectivity that can share data and produce analytically-derived insights. The highly integrated global network structure known as the Internet of Things. But compared to other kinds of equipment, IoT devices are more susceptible to attacks because of their small size and constrained memory, processing power, battery life, and other resources. Security and privacy pose several challenges, making them the most essential considerations for Internet of Things applications. Through an analysis of security architecture and key technologies, including encryption, cryptographic techniques, and communication security, as well as a description of the challenges they present. Numerous security issues are addressed by machine learning (ML) techniques The benchmark dataset from CICIDS 2017 is used in this investigation. The Friday-WorkingHours[1]Afternoon-DDos.pcap_ISCX.csv file from this dataset has been used to train and test the three[1]machine learning (ML) models: Linear Regression, Random Forest, and Decision Tree. There are 79 columns and 225745 rows in this CSV file. The identical set of data was used to evaluate all three techniques, and the results show that while linear regression is the fastest, it is also the least accurate. Random Forest is the slowest of the three but provides higher accuracy. In the end, decision trees are thought to be the most effective approach for these kinds of problems because they provide a reasonable balance between speed and accuracy. The purpose of hybrid algorithms is to generate more precise results. Fitting and residual graphs, among other important visualizations, are displayed by this machine learning model.Keywords: Internet of Things, Security & Privacy issues, Machine Learning Classifiers, Hybrid Algorithm.