부산대학교 도서관

This research explores the pervasive threat of injection attacks targeting Internet of Things devices. We leverage a generative modeling technique to effectively detect and prevent SQL injection attacks, mitigating potential cybersecurity risks to IoT devices. By learning the normal behavior of data, our model can identify deviations indicative of injection attacks. Our approach allows for unsupervised learning, making it possible to detect zero-day or previously unseen attack patterns without the need for labeled data, which is often scarce in cybersecurity contexts. In this sense, our model can process data in real-time, enabling immediate detection and response to injection attacks. In our study, we use publicly available datasets and develop a comprehensive approach to extract valuable insights from SQL injection data. Our study aims to proactively identify and prevent such attacks on IoT devices. The findings of this study are poised to make significant contributions to the IoT industry, recognizing the intensified vulnerability of IoT devices compared to traditional networks. By enabling real-time detection, our approach seeks to monitor and safeguard IoT devices, preventing unauthorized access, where hackers can gain root-level control, compromising security and privacy of users’ data.