Classification of Network Traffic by using Random Forest based Grid Search for Side-Channel Attacks Detection

In recent years, the smartphone-based Side-Channel Attacks (SCA) are common in this developing technological world. It exploits unintended information leakage such as extract sensitive information like PINs, passwords or keys, without hacking the device code directly. However, detecting SCAs was the most challenging task to provide potential security, so this research proposes a Random Forest (RF)-based Grid Search (GS) for SCAs detection. Initially, data is collected from Video Fingerprinting Encrypted Network Traces-SCA (VFENT-SCA) dataset which includes records and various variables related to network traffic during YouTube video streaming. Then, signal data is preprocessed with Butterworth Filter (BF) which removes noise and unwanted frequency components from the signals. Next, the features are extracted by using Discrete Wavelet Transform (DWT), which extracts frequency coefficient vectors that capture the underlying patterns in the network traffic. Finally, frequency coefficient vectors are fed into a RF classifier, which classifies network traffic patterns to detect SCAs and enhance the security. For improving the performance of RF, GS is utilized which is utilized for optimizes the hyperparameters. From the results, the proposed RF model achieved outstanding results when compared with existing K-Nearest Neighbors Deep Graph Neural Network (KNN+DeepGraphNet) method in terms of precision (0.95), Recall (0.96) and F1-score (0.96) respectively.