부산대학교 도서관

Fast progress in both smart devices with limited resources and state-of-the-art communication methods have helped establish the Internet of Things (IoT) as the gold standard for low-power lossy networks. However, because to the low computing, storage, and networking capabilities of the endpoint devices, these networks are susceptible to cyber assaults. IoT network security solutions are often rendered ineffective since many attacks are produced by slightly modifying previously known assaults. With their superior performance, deep learning algorithms have seen rapid adoption in security applications in recent years. In this paper, we introduce the IoT-IDCS-CNN, a brand new, unsupervised deep learning-based method for detecting and categorising cyberattacks in IoT communication networks. The system uses high-performance computing on dependable Nvidia GPUs, and parallel processing on fast Intel CPUs. It is structured into three parts: a traffic classification module, a feature learning module, and a refinement module for engineering features. Each part is developed independently and then tested and combined to create the complete system.