Minibatch-Enhanced Generative Adversarial Networks for IoT Network Intrusion Detection
Shoujian YU , Jiakang LIU , Hong ZHANG , Shigen SHEN
Journal of Donghua University(English Edition) ›› 2026, Vol. 43 ›› Issue (5) : 90 -98.
With the rapid expansion of the Internet of Things (IoTs), network connectivity has exposed systems to increasingly complex and previously unseen security threats. Conventional intrusion detection systems (IDSs) often struggle to identify subtle anomalies in high-dimensional and dynamically evolving IoT traffic. To address this problem, we propose a generative adversarial network (GAN)-based IDS tailored for IoT networks. The proposed method integrates minibatch discrimination to enhance sensitivity to inter-sample anomalies, replaces the conventional L1 distance with cosine similarity to better characterize relationships among high-dimensional feature vectors, and incorporates L1/L2 regularization to improve training stability and generalization. These enhancements enable more accurate discrimination between benign and malicious traffic while reducing overfitting. Experiments on the TON_ IoT dataset demonstrate that the proposed method achieves strong overall detection performance in terms of accuracy and F1 score while maintaining a low false positive rate.
generative adversarial network / intrusion detection system / minibatch discrimination / cosine similarity / L1/L2 regularization
| [1] |
|
| [2] |
|
| [3] |
|
| [4] |
|
| [5] |
|
| [6] |
|
| [7] |
|
| [8] |
|
| [9] |
|
| [10] |
|
| [11] |
|
| [12] |
|
| [13] |
|
| [14] |
|
| [15] |
|
| [16] |
|
| [17] |
|
| [18] |
Karras T, Aittala M, Laine S, |
| [19] |
|
| [20] |
|
| [21] |
|
| [22] |
|
| [23] |
|
| [24] |
|
| [25] |
|
| [26] |
|
| [27] |
|
| [28] |
|
/
| 〈 |
|
〉 |