Invisible Shield: Gesture-Based Mobile Authentication

Intelligent mobile devices have become the focus of the electronics industry in recent years. These devices, e.g., smartphones and internet connected handheld devices, enable quick and efficient access of users to both business and personal data, but also allow the same data to be easily accessed by an intruder if the device is lost or stolen. Existing mobile security solutions attempt to solve this problem by forcing a user to authenticate to a device before being granted access to any data. However, such checks are often easily bypassed or hacked due to their simplistic nature. In this work, we demonstrate Invisible Shield, a gesture-based authentication scheme for mobile devices that is far more resilient to attack than existing security solutions and requires neither additional nor visible effort from user perspective. In this work, we design methods that efficiently record and preprocess gesture data. Two classification problems, "one vs. many" and "one vs. all," are then mathematically formulated and examined using the gesture data collected from 20 individuals. Classification algorithms specialized for each case are developed, achieving a classification accuracy as high as 90.7% in the former case, and an equal error rate as low as 7.7% in the latter using real Android systems. Finally, the system resource requirements of different classification algorithms are compared