To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
Lightweight Autoencoder-Based Anomaly Detection for Resource-Constrained IoT Networks |
|
Author |
Salihah Alotaibi |
| Citation |
Vol. 25 No. 12 pp. 28-32 |
|
Abstract |
The Internet of Things (IoT) is growing quickly and is now used in homes, industries, hospitals, transport systems, and many smart-city services. As more devices get connected, the security risks also increase because most IoT devices have very limited processing power and memory. Many of the current anomaly-detection methods depend on heavy machine-learning models, and these models are difficult to run on small devices that work on low energy. This paper introduces a lightweight anomaly-detection method. The idea is to first reduce the number of features, so the data becomes easier to handle and then use a small autoencoder that learns what normal traffic usually looks like. When the traffic behaves in an unusual way, the reconstruction error becomes higher, which helps in spotting abnormal activity. The method is tested on a synthetic IoT dataset that tries to mimic real network traffic, and it is also compared with some basic existing models. Performance evaluation of the method shows that it can be used for IoT applications with low resources and good accuracy. |
|
Keywords |
IoT Security, Anomaly Detection, Lightweight Autoencoder, Feature Reduction |
|
URL |
