Adaptive Recognition System For Keystroke Content Based On Acoustic Signal

Nowadays,with the continuous development of information technology,information security issues are becoming increasingly important.People are demanding more and more information confidentiality,and users' security awareness is gradually increasing.In work and life,keyboard is the main interactive tool between human and computer.Users' behavior data of using keyboard,which is easy to be stolen and analyzed by illegal elements,and becomes a loophole leading to information leakage.Therefore,with the help of information processing technology and wireless identification technology,collecting,analyzing and identifying user's behavior data using keyboard can provide strong support for protecting user's information security and help to better implement anti-monitoring and anti-eavesdropping in the process of human-computer interaction.At present,there are mainly the following aspects in the keystroke content recognition scheme: first,it's the keystrokes recognition based on acoustic signal.This scheme locates the specific position of keystroke,according to the different time when the keystroke acoustic signal arrives at the microphone array,but it requires more hardware equipment and has lower precision;second,it's the keystrokes recognition based on electromagnetic wave signal.It can identify keystrokes by making use of the different effects of finger movements on electromagnetic signals while users were typing different keys.However,these schemes are vulnerable to external electromagnetic interference and need additional equipment support.Thirdly,it's keystrokes recognition based on computer vision,users' motions are captured by image acquisition devices such as cameras,and keystrokes are identified by image processing algorithms.Nevertheless,such schemes need much more hardware equipment and are energy-consuming.They are also restricted by ambient light intensity,and will no longer be applicable in weak light or dark environment.In order to solve various problems of existing technical schemes for keystroke behavior recognition,this paper proposes an adaptive keystroke content recognition method and system using keystroke acoustic signals.Firstly,the built-in microphone of smart phone is used to collect keystroke acoustic signals,and then Butterworth filtering and wavelet denoising areapplied to the signals.Then,an adaptive threshold endpoint detection algorithm is improved.The keystroke events are extracted effectively.Then the non-Gaussian measure of the signal is used to distinguish the single keystroke events and the combined keystroke events.The two types of keystroke events are analyzed separately.The normalized ASD and MFCC features are extracted from the single keystroke events.After the blind deconvolution of the combined keystroke events,the signal features are extracted.Finally,the optimized KNN classifier is used to classify the signal features,so as to achieve the purpose of identifying keystroke content.In the experiment,on the premise of keeping the position of keyboard and smart phone fixed and the external environment unchanged,the average accuracy of identifying single key position by keystroke acoustic signal is 93.32%,and the average accuracy of identifying combined keystroke is 83.05%.Subsequently,cross-validation is carried out from the angles of location,distance,training set size and cross-user,which proves the feasibility,effectiveness and robustness of the keystroke recognition scheme proposed in this paper.