IoT Network Cybersecurity Assessment with the Associated Random Neural Network

This paper proposes a method to assess the security of an n device, or IP address, IoT network by simultaneously identifying all compromised devices and addresses. It uses specific Random Neural Network (RNN) architecture composed two mutually interconnected sub-networks that complement each other in recurrent structure, called Associated RNN (ARNN). For addresses network, distinct neurons ARNN advocate opposite views: not compromised. The fully 2 neuron structure paired learns offline from ground truth data. Thus rather than requiring separate attack detector at node, offers single overall observes incoming traffic about interdependencies between nodes, formulates recommendation for device address network. weight initialization learning algorithm are discussed, performance is evaluated using real data, compared against several testing techniques. Results obtained both off-line with on-line incremental simplified average metric measured packet traffic. Comparisons best state-of-the-art techniques show significantly outperforms previously known approaches.