No abstract

Computers are widely used for business and entertainment purposes throughout our modern lives. Computer kits provide a variety of services such as text processing, programming, shopping, and gaming. Computers have greatly enhanced the quality of our lives; however, we discover an often-overlooked fact that engaging in computer-related activities may be eavesdropped upon by an attacker by sniffing the emitted acoustic signals from keyboard and mouse. The activity of eavesdropping via acoustic side channel has lower requirements in terms of hardware instrumentation and is easier to implement in real-world applications than other side channel attacks that have been presented in previous work. In this paper, we design and implement a system, namely, Behavicker, to validate the feasibility of this kind of attack. Unlike conventional activity recognition, Behavicker infers high-level computer-usage activities with a semantics-preserving multiscale learning scheme, based on the recognition of basic keyboard and mouse events including left click, right click, middle click, scrolling up, and scrolling down. Real-world experiments show that Behavicker can recognize six interaction events with an accuracy of 88.3% and infer computer-usage activities with an accuracy of 82.7% in an indoor environment.

Handwritten signatures are widely used for identity authorization. However, verifying handwritten signatures is cumbersome in practice due to the dependency on extra drawing tools such as a digitizer, and because the false acceptance of a forged signature can cause damage to property. Therefore, exploring a way to balance the security and user experiment of handwritten signatures is critical. In this paper, we propose a handheld signature verification scheme called SilentSign, which leverages acoustic sensors (i.e., microphone and speaker) in mobile devices. Compared to the previous online signature verification system, it provides handy and safe paper-based signature verification services. The prime notion is to utilize the acoustic signals that are bounced back via a pen tip to depict a user’s signing pattern. We designed the signal modulation stratagem carefully to guarantee high performance, developed a distance measurement algorithm based on phase shift, and trained a verification model. In comparison with the traditional signature verification scheme, SilentSign allows users to sign more conveniently as well as invisibly. To evaluate SilentSign in various settings, we conducted comprehensive experiments with 35 participants. Our results reveal that SilentSign can attain 98.2% AUC and 1.25% EER. We note that a shorter conference version of this paper was presented in Percom (2019). Our initial conference paper did not finish the complete experiment. This manuscript has been revised and provided additional experiments to the conference proceedings; for example, by including System Robustness, Computational Overhead, etc.

Contactless authentication is crucial to keep social distance and prevent bacterial infection. However, existing authentication approaches, such as fingerprinting and face recognition, leverage sensors to verify static biometric features. They either increase the probability of indirect infection or inconvenience the users wearing masks. To tackle these problems, we propose a contactless behavioral biometric authentication mechanism that makes use of heterogeneous sensors. We conduct a preliminary study to demonstrate the feasibility of finger snapping as a natural biometric behavior. A prototype-SnapUnlock system was designed and implemented for further real-world evaluation in various environments. SnapUnlock adopts the principle of contrastive-based representation learning to effectively encode the features of heterogeneous readings. With the representations learned, enrolled samples trained with the classifier can achieve superior performances. We extensively evaluate SnapUnlock involving 50 participants in different experimental settings. The results show that SnapUnlock can achieve a 4.2% average false reject rate and 0.73% average false accept rate. Even in a noisy environment, our system performs similar results.