Cyber-Physical Threat Intelligence for IoT Using Machine Learning
Sunil Sonawane, Reshma R. Gulwani · Auerbach Publications eBooks · 2024
IoT represents a significant technological advancement that facilitates the seamless exchange of digital information through interconnected devices. This transformative innovation enables us to interact with and harness the potential of various objects and commodities in our surroundings, leveraging affordable and interconnected devices. This study delves into the realm of potential network vulnerabilities and corresponding defensive measures within the context of the IoT. The insights derived from this research are of substantial value to cybersecurity professionals and experts specializing in IoT network security. The research equips individuals with the knowledge required to proactively navigate real-world connectivity challenges by not only modeling potential issues but also devising proactive solutions. The research identifies discrepancies within the IoT ecosystem by conducting a comprehensive replication of the MQTT (Message Queuing Telemetry Transport) protocol across an online network. This chapter offers a novel two-branch deep neural network to address the constraints of commonly used machine learning algorithms, which frequently suffer from overfitting. This neural network makes use of a custom-designed frequency filter bank as well as a multi-scale cross-attention fusion module to extract more generic recapture artifacts. To combat potential threats such as Distributed Denial of Service (DDoS) attacks, a diverse set of classification algorithms such as Multilayer Perceptron (MLP), K-Nearest Neighbours (KNN), Nave Bayes (NB), Logistic Regression (LR), Artificial Neural Network (ANN), Support Vector Machine (SVM), Random Forest (RF), and Decision Tree (DT) are used. The suggested method makes use of a dataset with 34 unique features and 4998 items divided into eight groups based on routing bandwidth. Notably, the categorization performance produced outstanding results, with an accuracy rate of 99.95 percent. Furthermore, the study includes the usage of Snort to build an intrusion detection system (IDS), which contributes to the practical application of the research findings. Overall, the findings of this study provide theoretical insights as well as practical applications, improving our understanding of IoT security and its real-world ramifications.