LOGISTIC REGRESSION AND RANDOM FOREST CLASSIFIER FOR ATTACK DETECTION IN IOT SENSOR DATA

Abstract

The Internet of Things (IoT) connects a vast array of devices, ranging from home appliances to industrial sensors, creating an interconnected network of smart devices. IoT applications generate large volumes of sensor data, which are highly susceptible to security breaches and attacks. Cyber-criminals may exploit vulnerabilities in the IoT ecosystem to manipulate sensor data, leading to disastrous consequences such as unauthorized access, data falsification, and service disruption. In addition, IoTbased attacks can lead to severe consequences such as data manipulation, privacy breaches, and economic losses. One of the major challenges is detecting and preventing attacks on the valuable sensor data collected by IoT devices. Traditional security methods designed for conventional networks may not be suitable for the complex and distributed nature of IoT systems. To address this concern, there is a need for specialized techniques tailored for IoT sensor data to protect these systems and their users. Further, the proposed work aims to contribute to the field of cybersecurity and foster more resilient and secure IoT implementations. This work introduces a comprehensive and practical solution to enhance IoT security. Here, Logistic Regression, and Random Forest classifiers are employed for attack detection from the IoT sensor data, where the first one is a straightforward yet powerful technique for binary classification, enabling the detection of simple intrusion attempts. Meanwhile, the Random  Forest Classifier excels at handling complex patterns and interactions in data, making it effective in identifying sophisticated attacks with multiple variables and dependencies. By leveraging the strengths of these algorithms, the proposed approach provides a robust and advanced system for detecting a wide
range of attacks in IoT sensor data