Research On Footsteps Acoustic-based Near-Field Authentication Technology

With the deeper fusion of AI and the Io T(Internet of Things),smart devices embedded abundant sensors gradually come into our lives and industries.On one hand,smart devices have the ability to sense the changes of environment and interact with each other remotely,which makes it possible to provide highly efficient and convenient services.On the other hand,this kind of devices involve to much privacy information or sensitive data,which makes reliable authentication technology obtain much more attention.The paper proposed a footstep-based short-distance continuous authentication technique as an assistant method to enhance the security of smart devices which are easy to be attacked by replay attack or shoulder surfing when using numbers,graphs,pictures,voices as passwords.Taking the advantage of footsteps acoustic to authenticate users' identities faces two challenges.One is to eliminate environmental noise from the acoustic sequences including footsteps.The other is the consistency and robustness of the extracted footstep features.According to these challenges,the main works of this paper can be summarized as follows:Firstly,proposes an optimized spectral subtraction based environmental noise elimination method.After processing original signals through band-pass filter,Wavelet transformation and First Difference,obtain the real-time noise signals by using the designed automatic noise detection algorithm,and then use the power spectrum of original signals to subtract it of the detected noise signals to amplify footstep signals.Secondly,proposes a four features-based footsteps calculation model.Make use of the location relationship and the body difference of human beings to get footstep interval,footstep duration time,footstep size and the rate of two adjacent footsteps to execute continuous authentication.Finally,deploys the methods proposed in the paper in real scenarios.The results show the methods can continuously authenticate users with a False Rejection Rate under 12%.Every time adding additional 100 attacks,the False Accept Rate increases by 1.6% averagely.