Research On Wireless Sensing-Based Person Identification And Authentication Methods

The advent of the digital era has placed higher demands on personal privacy and data protection,making identification and authentication technologies more crucial than ever.Meanwhile,the emergence of smart space has propelled the advancement of person identification and authentication technologies towards intelligence,automation,and convenience.Person identification and authentication solutions can be divided into two categories,which are credentialbased and biometric-based.Compared with credential-based methods,biometric-based methods are more reliable,but they face many problems in terms of privacy,device cost,user experience and generalizability.Wireless sensing-based identification and authentication technologies can overcome the shortcomings of existing identification and authentication technologies and have attracted the attention of more and more researchers.As an emerging technology in the field of ubiquitous computing and Internet of Things(Io T),wireless sensing has the advantages of no privacy,wide application range,low cost,and is widely used in smart home,smart medical,industrial automation,etc.Wi-Fi sensing and millimeter wave(mm Wave)sensing,as two important branches of wireless sensing,have a wide range of applications in person identification and authentication.However,existing solutions still have some problems and limitations.Specifically,existing Wi-Fi sensing-based person identification and authentication methods require stringent restrictions on user behavior to meet authentication requirements.In contrast,there are relatively few studies on mm Wave sensing-based identification and authentication,and research gaps still exist in certain aspects.This thesis conducts a comprehensive study on wireless sensing-based identification and authentication technologies,with a focus on both Wi-Fi sensing-based identification and mm Wave sensing-based authentication.The main contents of this thesis are summarized as follows:1.This thesis proposes WiID,a Wi-Fi identification system based on bio-electromagnetic information.Compared with existing schemes,WiID can automatically extract bioelectromagnetic information of users for identification without requiring them to remain stationary during the identification process.In this work,a series of signal processing algorithms are proposed to eliminate environmental noise and motion noise from the raw signal,and to obtain the motion sensitivity vector reflecting the user’s state by introducing an adaptive threshold-based signal conversion algorithm,so as to automatically segment the raw Wi-Fi Channel State Information(CSI)data to obtain valid bio-electromagnetic information.Meanwhile,a Long ShortTerm Memory(LSTM)-based neural network is designed in this work to perform person identification.The evaluation results of WiID under a variety of different environments and different experimental configurations show that the proposed system has good identification accuracy and robustness.2.This thesis proposes mmSign,the first mm Wave-based non-intrusive user handwritten signature authentication system,which is applicable to any writing surface,has no privacy issues,and is data efficient and user-friendly.This work proposes a series of signal processing methods to obtain time-velocity feature maps from the raw mm Wave signals reflecting the dynamic characteristics of the user’s handwritten signature,and proposes three mm Wave data augmentation methods based on the changes of the mm Wave signal during the signature execution process.In addition,a Transformer-based authentication model is designed in this thesis to verify the authenticity of the signature.Finally,the handwritten signature authentication task is formulated as a few-shot learning problem and a meta-learning framework is designed to ensure that new users can quickly adapt to the system with only a few samples.The evaluation results conducted in multiple environments using various signature tools and writing surfaces show that mmSign has good adaptability and security for new users.