Abstract:Network attack identification effectively secures Internet of Things (IoT) application scenarios. However, dynamic scene changes, attack feature reliance, high data dimensions, and challenges with spatio-temporal feature fusion frequently pose limitations to attack traffic identification in IoT contexts. Definitive intelligent IoT attack identification enables intelligent algorithms to extract attack features for application scenarios with fixed topological environments but cannot construct the intricate changes of IoT application scenarios. Through the dynamic acquisition, feature awareness, and deep learning, intelligent digital twin-based attack detection can address these issues and enhance attack identification for IoT threats. Thus, this paper proposed an intelligent digital twin method based on spatio-temporal feature fusion for IoT attack behavior identification. Firstly, feature subsets are selected based on information gain to reduce the dimensionality of IoT data with high traffic; Secondly, a parallel spatio-temporal feature extraction model is designed unlike the existing tandem model, which uses a simplified Convolutional Neural Networks (CNN) model to learn the spatial features of the attack, a Bi-directional Long Short-Term Memory (BiLSTM) model to learn the temporal features of the attack, an attention mechanism to fuse the temporal and spatial features, and the (Deep Neural Networks) DNN to learn the combined features; Finally, the virtual instance space and topology of the attack scenario are simulated using digital twin (DT) to build a digital version of the complex system for IoT applications and tested in a simulation environment. Based on experimental results using the UNSW-NB15 and CICIDS2017 datasets, this paper shows that the proposed method can extract spatio-temporal features from network attack traffic and has a 5% improvement in test accuracy.

IoT Device Identification Base on Inter-flow Correlation Analysis Using Graph Neural Network

IoT Device Identification Based on Network Traffic Characteristics

Automatic IoT device identification: a deep learning based approach using graphic traffic characteristics

Flow Topology-Based Graph Convolutional Network for Intrusion Detection in Label-Limited IoT Networks

GraphIoT: Accurate IoT Identification Based on Heterogeneous Graph.

Efficient IoT Device Identification via Network Behavior Analysis Based on Time Series Dictionary

IoT Device Identification Based on Network Communication Analysis Using Deep Learning

Network Traffic Analysis based IoT Device Identification

Inter-Flow Spatio-Temporal Correlation Analysis Based Website Fingerprinting Using Graph Neural Network

IoT Device Identification Using Deep Learning

Transformer-Based Device-Type Identification in Heterogeneous IoT Traffic

Modeling IoT Equipment with Graph Neural Networks

Automatic Device Classification from Network Traffic Streams of Internet of Things

Anomaly traffic detection in IoT security using graph neural networks

Type classification and identification of IoT devices by using traffic characteristics

FlowCorrGCN: Enhancing Flow Correlation Through Graph Convolutional Networks and Triplet Networks

Graph Neural Networks in IoT: A Survey

Your Smart Home Can't Keep a Secret: Towards Automated Fingerprinting of IoT Traffic with Neural Networks

N-STGAT：Spatio-Temporal Graph Neural Network Based Network Intrusion Detection for Near-earth Remote Sensing

Leveraging Deep Learning for IoT Transceiver Identification

An Intelligent Digital Twin Method based on Spatio-Temporal Feature Fusion for IoT Attack Behavior Identification