BlowPrint
Blow-based multi-factor biometric authentication for smartphones, combining blow acoustics with face recognition (ESORICS 2025). 99.82% accuracy for the combined factors.
- Android
- DTW
- FaceNet
- kNN
- Python
- R
An Android app that records a short blow through the microphone and a face image through the front camera. The blow audio is reduced to a smoothed intensity signal and compared using dynamic time warping (DTW), while faces are compared using embeddings from a pre-trained FaceNet model. The two similarity scores are normalized and combined by a weighted sum, and a k-nearest-neighbour rule makes the decision. No neural network was trained for this paper. Data was collected from 50 participants, each giving 10 sessions (5 sitting and 5 standing). Reported accuracy: 99.35% for blow acoustics alone, 99.96% for the face alone, and 99.82% for the combined approach. Paper: "BlowPrint: Blow-Based Multi-factor Biometrics for Smartphone User Authentication" (ESORICS 2025).