Open Set Wireless Transmitter Authorization: Deep Learning Approaches and Dataset Considerations

Due to imperfections in transmitters' hardware, wireless signals can be used verify their identity an authorization system. While deep learning was proposed for transmitter identification, existing work has mainly focused on classification among a closed set of transmitters. Malicious transmitters outside this will misclassified, jeopardizing the In article, we formulate problem recognizing authorized and rejecting new as open recognition anomaly detection. We consider approaches based one several binary classifiers, multiclass signal reconstruction. study how these scale with required number propose using known unauthorized assist training its impact. The evaluation procedure takes into consideration that some might more similar than others nuances effects. authorization's robustness against temporal changes fingerprints is also evaluated function approach dataset structure. When 10 50 WiFi from publicly accessible testbed, were able achieve outlier detection accuracy 98% same day test 80% different set.