Machine Learning-Based IoT Cyber Threat Prediction Using Network Traffic Analysis

As the Internet of Things (IoT) grows, the increasing complexity and volume of network traffic generated by these devices create new cybersecurity challenges. Conventional security measures often struggle to keep up with the advanced and constantly changing nature of attacks targeting IoT networks. This study presents a new method for improving the detection of cyber threats in IoT environments by analyzing segmented network traffic. We divide network traffic into time-based segments and group data packets by their source, enabling a thorough examination of traffic patterns and behaviors that can reveal either regular activity or potential security risks. Central to our approach is identifying specific features from these traffic segments, particularly those unique to IoT systems and their communication behaviors. To evaluate our approach, we compare several baseline classification models, including widely used ensemble methods, assessing their performance across various metrics. Also, we employ the sigmoid calibration method to refine the accuracy and reliability of the threat detection process. Our findings demonstrate the effectiveness of this approach in improving the identification of cyber threats, offering a promising direction for securing IoT networks. This research underscores the value of analytical techniques in IoT cybersecurity and lays the groundwork for future innovations in this critical field.