Channel State Information Based Anti-Internet Addiction System

In recent years,with the substantial progress in machine learning and wireless sensing such as RF,infrared sensors,etc.,WiFi has already been utilized over its traditional applications due to its accessibility and low expense,which enables varieties of device-free passive(DfP)WiFi-recognition and detection systems.In this paper,we propose AAOG:Anti-Addiction on Online Gaming system,which firstly presents that WiFi signals could also be leveraged to detect the behavior of Playing Online Game(POG)utilizing the fine-grained Channel State Information(CSI).AAOG recognizes that someone is playing online game by studying the motions of the player and exploiting the common features during POG.We implement AAOG on commercial off-the-shelf devices.The main research is as follows:First of all,through the Kernel Density Estimation(KDE)method,the optimal detection threshold is obtained,ie,the detection from the silent state to the action state(intrusion detection)is performed.The experimental results verify the algorithm in a multi-path indoor environment that the beginning of the behavioral state can be well detected and the precision can reach up to 94.2%.Then,after detecting the start of the behavior state,it will be further determined by a certain algorithm whether the behavior is game-playing.Here,we use the Principal Component Analysis(PCA)to reduce the dimension of the original matrix,then use the wavelet transform(WT)to extract the time-frequency information of the signal,after that we can obtain the statistical characteristics of the time-frequency signal and perform Standardized Processing,now the amplitude feature extraction is completed,combined with the phase correlation characteristics of time correlation,all the two features are injected into the classifier for identification.Finally,we select the Support Vector Machine(SVM)and adjust the parameters to achieve the best detection performance.Experimental results demonstrate that AAOG is robust and accurate,it achieves an optimal detection rate of 95.8%,and the False Positive Rate(FPR)is lower than 4.4%.We envisage that this technology can be widely employed in smart home settings or office room.