Deep-Feature-Based Autoencoder Network for Few-Shot Malicious Traffic Detection

With the increase of Internet visits and connections, it is becoming essential arduous to protect networks different devices Things (IoT) from malicious attacks. The intrusion detection systems (IDSs) based on supervised machine learning (ML) methods require a large number labeled samples. However, abnormal behaviors far less than that normal behaviors, let alone shots behavior samples which can be intercepted as training dataset are actually limited. Consequently, key research topic conduct anomaly for small This paper proposes an model with few solve problem in few-shot convolutional neural (CNN) autoencoder (AE). mainly consists CNN-based pretraining module AE-based data reconstruction module. Only utilized pretrain build structure extracting deep features. simply chooses features data. There also exist some effective attention mechanisms Through samples, accuracy improved compared merely AE. simulation results prove this solution above problems occurring network detection. In comparison original AE other clustering methods, proposed advances visible way.