*=Equal Contributors
Private units have adopted various authentication strategies, together with biometric recognition and passcodes. In distinction, headphones have restricted enter mechanisms, relying solely on the authentication of linked units. We current Moonwalk, a novel technique for passive consumer recognition using the built-in headphone accelerometer. Our method facilities on gait recognition; enabling customers to ascertain their id just by strolling for a short interval, regardless of the sensor’s placement away from the ft. We make use of self-supervised metric studying to coach a mannequin that yields a extremely discriminative illustration of a consumer’s 3D acceleration, with no retraining required. We examined our technique in a examine involving 50 members, attaining a median F1 rating of 92.9% and equal error price of two.3%. We prolong our analysis by assessing efficiency underneath numerous circumstances (e.g. shoe varieties and surfaces). We talk about the alternatives and challenges these variations introduce and suggest new instructions for advancing passive authentication for wearable units.