一种基于卷积循环自编码器的物联网网络入侵检测方法
By employing a convolutional recurrent autoencoder-based method for feature selection and encoding/decoding of IoT network traffic data, combined with loss function optimization, this approach addresses the poor classification performance of traditional detection systems under diverse attacks, achieving highly efficient intrusion detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANDONG JIAOTONG UNIV
- Filing Date
- 2023-04-14
- Publication Date
- 2026-07-17
AI Technical Summary
Existing IoT network intrusion detection systems are limited by insufficient data labels and the simple structure of traditional autoencoder models when facing diverse network attacks, resulting in poor classification performance and difficulty in adapting to rapidly changing network intrusion types.
We employ a convolutional recurrent autoencoder-based approach, which optimizes model parameters by combining mean squared error and L2 regularization loss functions through data preprocessing, feature selection, spatial and temporal feature encoding and decoding, and data extraction. This approach enables unsupervised learning and improves the accuracy of feature extraction and classification.
By effectively utilizing network traffic data, the accuracy of IoT network intrusion detection has been improved, adapting to rapidly changing network intrusion types, reducing reliance on data tags, and enhancing the generalization ability of the detection model.
Smart Images
Figure CN116405298B_ABST