No abstract

This paper demonstrates that highly accurate radiometric identification is possible using CAPoNeF feature engineering method. We tested basic ML classification algorithms on experimental data gathered by SDR. The statistical and correlational properties of suggested features were analyzed first with the help of Point Biserial and Pearson Correlation Coefficients and then using P-values. The most relevant features were highlighted. Random Forest provided 99% accuracy. We give LIME description of model behavior. It turns out that even if the dimension of the feature space is reduced to 1, it is still possible to classify devices with 99% accuracy.

No abstract

We propose a new approach to traffic preprocessing called Differentiation of Sliding Rescaled Ranges (DSRR) expanding the ideas laid down by H.E. Hurst. We apply proposed approach on the characterizing encrypted and unencrypted traffic on the well-known ISCXVPN2016 dataset. We deploy DSRR for flow-base features and then solve the task VPN vs nonVPN with basic machine learning models. With DSRR and Random Forest, we obtain 0.971 Precision, 0.969 Recall and improve this result to 0.976 using statistical analysis of features in comparison with Neural Network approach that gives 0.93 Precision via 2D-CNN. The proposed method and the results can be found at https://github.com/AleksandrIvchenko/dsrr_vpn_nonvpn.