Securing RFID With GNN: A Real-Time Tag Cloning Attack Detection System
In the field of RFID systems, cloning attacks that replicate authentic tags to deceive readers pose a significant threat to corporate security, potentially leading to financial losses and reputational damage. Many existing solutions struggle to mitigate this threat without altering the Medium Access...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Open Journal of the Communications Society |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10758764/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In the field of RFID systems, cloning attacks that replicate authentic tags to deceive readers pose a significant threat to corporate security, potentially leading to financial losses and reputational damage. Many existing solutions struggle to mitigate this threat without altering the Medium Access Control (MAC) layer protocols or integrating additional hardware resources, which are impractical adjustments for commercial off-the-shelf (COTS) RFID devices. This paper introduces an innovative system framework that leverages Graph Neural Network to detect RFID tag cloning attacks without the need to change the MAC protocols or add hardware resources. The system can automatically uncover implicit topological structures from RFID signal data and adaptively capture complex inter-signal relationships. By constructing dynamic graph and employing Graph Attention Network, this framework not only captures deep data correlations that traditional detection methods cannot identify but also demonstrates exceptional accuracy and robustness in experiments. Experimental results have proven that the framework maintains stable performance even when the training and testing data distributions are mismatched, as verified in both static and dynamic tag cloning attack scenarios. Furthermore, the framework effectively identifies anomalous behavior by comprehensively considering precision, recall, and F1 scores, especially when dealing with highly imbalanced datasets. |
|---|---|
| ISSN: | 2644-125X |