چكيده لاتين
With the rapid expansion of Internet of Things (IoT) technology as one of the most advanced and practical innovations of our time, numerous capabilities and opportunities have been provided to users. This technology is widely used in various fields, including industry, smart homes, healthcare, traffic management, and environmental protection. Given the continuous development of IoT and the growing number of its users, this domain has increasingly become a target for malicious attackers, emphasizing the critical need for its security. IoT operates using intelligent and dynamic systems that, through sensors and network-connected devices, can monitor and respond to various events. Based on reports received from the environment, these systems can perform specific actions in the real world. However, the process of transmitting and receiving these reports is highly sensitive and vulnerable. One of the primary vulnerabilities in this area lies in weaknesses in the report transmission process, which can be exploited. Specifically, an attacker may send a fake report, leading to a spoofing attack, or prevent the transmission of an actual report, resulting in an event masking attack. Given the computational resource constraints of small IoT devices, employing heavy encryption protocols and algorithms as a practical solution for securing exchanges is not feasible. Therefore, alternative and more suitable methods for ensuring security are essential. In this regard, event verification systems are employed to address these challenges. These systems are responsible for verifying the accuracy of reported events and are generally categorized into rule-based and machine learning-based systems. Current methods in this field are limited to detecting spoofing attacks and cannot simultaneously identify event masking attacks. The only available method capable of detecting event masking attacks suffers from low accuracy and the need for a large number of sensors.
This study introduces a machine learning-based event verification system that can simultaneously detect both spoofing and event masking attacks with higher accuracy and speed compared to existing methods, while requiring fewer sensors in IoT environments. The proposed method achieves improved results by combining a neural network algorithm for selecting appropriate sensors and a 4-class classifier that integrates sensor values and environmental notifications to identify attacks. Compared to existing methods, the proposed approach increases the detection accuracy of spoofing attacks by up to 10% and event masking attacks by up to 38%. Additionally, it is noteworthy that the proposed method achieves an average 75% reduction in the number of sensors required for event verification compared to some existing methods.
