An Approach to Early Computer Network Intrusion Detection Based on the Wavelet Transform Energy Spectra Analysis

The current stage of development of cybersecurity has led to the need to create new and improve old methods of data analysis. The paper proposes a novel approach to early intrusion detection based on the analysis of the spectral plane of the signal and the detail coefficients obtained by wavelet transform. The approach makes it possible to consider data not only in the frequency domain, but also in the time domain, which greatly simplifies the localization of anomalies. Wavelet analysis can effectively extract information from a signal and is suitable for anomaly detection, while energy spectrum analysis allows you to determine the physical nature of this signal and implement its suppression or filtering. The approach represents the signal at different frequency values. Different wavelets have several decomposition levels, and each level has a different center frequency. The energy spectrum of the signal was reconstructed from the wavelet coefficients. For a given energy spectrum, the energy cumulate at high, medium and low frequencies was calculated. Experimental results have shown that this approach is well suited for detecting anomalies in network traffic and can be applied to detect new attacks.