An Unsupervised Learning-Based Network Threat Situation Assessment Model for Internet of Things

摘要: With the wide application of network technology, Internet Things (IoT) systems are facing increasingly serious situation threats; threat assessment becomes an important approach to solve these problems. Aiming at traditional methods based on data category tag that has high modeling cost and low efficiency in assessment, this paper proposes a model unsupervised learning for IoT. Firstly, we combine encoder variational autoencoder (VAE) discriminator generative adversarial networks (GAN) form V-G network. Then, obtain reconstruction error each layer by training collection with normal traffic. Besides, conduct 3-layer output calculate abnormal threshold training. Moreover, carry out group testing test dataset containing traffic probability group. Finally, value (TSV) according impact. The simulation results show that, compared other methods, proposed method can evaluate overall security more intuitively stronger characterization ability threats.