Detection of cyber-attacks on Wi-Fi networks by classification of spectral data

In many areas, communications or computer networks include both wired and wireless sections. In this research, we are interested in the wireless network sections. This part of the networks can targeted by denial of service attacks, affecting the reception quality of communication signals, or by "man-in-the-loop" attacks aiming to intercept information. This paper presents a work based on the analysis of wireless electromagnetic activity to detect such attacks against an IEEE 802.11n Wi-Fi network. The approach is based on the analysis of spectral occupation by classification technics. Experimentations were performed in anechoic chamber in applying jamming attacks and de authentication attacks. In a first step, in performing the Principal component analysis of the spectra measured for the different tested situations, we analyse if the different classes can be separated. In a second step, we assess the ability of a Self Adaptative Kernel Machine to classify the different attacks without a preliminary learning phase of the attack situations.