نوع مقاله : مقاله پژوهشی
نویسندگان
1 دانشجوی دکتری کامپیوتر دانشگاه جامع امام حسین(ع)، تهران، ایران
2 استادیار کامپیوتر دانشگاه جامع امام حسین(ع)، تهران، ایران
چکیده .
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
Cybersecurity in the Internet of Things (IoT) is a critical area of information and communication technology that faces major challenges due to the complexity and dynamic nature of IoT systems. One of the primary issues in this domain is the detection of anomalies and sophisticated cyberattacks, which stem from the structural complexity and unpredictable behaviors of interconnected devices. These security threats can severely affect system integrity, performance, and data confidentiality. Previous research has explored various approaches, including temporal behavior analysis and network communication modeling, to mitigate cyber risks. However, when applied independently, these approaches often fail to provide a comprehensive defense mechanism.To address these limitations, this study proposes a hybrid approach that integrates graph-based modeling with behavioral deep learning methods. Specifically, by representing device interactions as graphs and analyzing temporal variations using Graph Neural Networks (GNNs) and Long Short-Term Memory (LSTM) models, the proposed method enhances anomaly detection and reduces the likelihood of cyberattacks, thereby lowering overall cyber risk. Experimental evaluations conducted on IoT-related datasets demonstrate that the proposed model significantly outperforms conventional methods. The results show superior performance metrics with 0.92 accuracy, 0.91 precision, 0.94 recall, and an F1-score of 0.92, along with a reduced false alarm rate. These findings highlight the effectiveness of the proposed approach in strengthening IoT cybersecurity, representing a significant advancement in risk reduction and system protection.
کلیدواژهها [English]