Abstract
The Internet of Things (IoT) is susceptible to threats from natural disasters and human damage, thereby affecting the communication efficiency and security of the network. The scale-free IoT topology can resist the impact of random attacks, but is particularly vulnerable to malicious attacks against critical nodes. To tackle this problem, this paper proposes an intelligent topology optimization strategy combined with convolutional neural network, which evolves the initial IoT topology into a robust onion-like structure against malicious attacks. In addition, we introduce a method to convert topology information into a sequence model suitable for feature extraction. The experimental results show that this strategy can efficiently optimize the network topology into an onion-like structure.