Abstract
We propose Secure Pick Up (SPU), a convenient, lightweight, in-device, non-intrusive and automatic-learning system for smartphone user authentication. Operating in the background, our system implicitly observes users' phone pick-up movements—the way they bend their arms when they pick up a smartphone to interact with the device—to authenticate the users. SPU automatically learns the user's behavioral pattern without requiring large amounts of training data from other users, making it highly deployable. We propose a weighted multi-dimensional Dynamic Time Warping (DTW) algorithm to effectively quantify similarities between users' pick-up movements. SPU does not rely on a remote server, making it efficient and usable even without network access. Our system can adaptively update a user's authentication model to accommodate behavioral drift over time with negligible overhead. Extensive experiments demonstrate that SPU can achieve authentication accuracy up to 96.3% with a very low latency of 2.4 milliseconds, reducing the frequency of explicit authentication by 32.9%.
Introduction
Smartphone users unlock their devices dozens of times daily. However, the friction of traditional authentication methods like PINs, patterns, or fingerprints leads approximately 64% of users to leave their devices unprotected. While existing behavioral biometrics like gait analysis or keystroke dynamics offer solutions, they often require significant training data, privacy-invasive cloud processing, or specific user actions.
Secure Pick Up (SPU) addresses this "usability-security" gap by leveraging a natural, ubiquitous action: the act of picking up the phone. SPU is designed to be implicit, running in the background; in-device, requiring no cloud connectivity; and adaptive, learning from the user's changing habits.
System Design and Methodology
The core of SPU is the identification of unique behavioral signatures in the way a user moves their arm to bring a phone from a stable surface to their field of vision.
Signal Extraction and Segmentation
SPU monitors the accelerometer and gyroscope in the background. Upon a trigger event, such as pressing the power button, the system backtracks through the sensor data to isolate the "pick-up" segment. This segment is defined as the transition from a stable state to the moment the screen is activated.
Weighted Multi-dimensional Dynamic Time Warping (DTW)
To measure similarity between temporal sequences that vary in speed, SPU utilizes an enhanced DTW algorithm. Unlike standard DTW, SPU applies asymmetric weights to different sensor axes. Empirical results indicated that translational movement (accelerometer) is more informative than rotational movement (gyroscope) for pick-up actions, leading to a 0.6 to 0.4 weight ratio. Notably, the inclusion of a magnetometer was found to degrade accuracy due to environmental noise.
System Workflow
The SPU architecture integrates signal extraction, profile comparison, and adaptive updates. If the implicit check fails, the user is prompted for explicit authentication (e.g., fingerprint). If the explicit check succeeds, the system uses the new movement data to update the user's profile, accounting for "behavioral drift."
Experimental Evaluation
The system was evaluated using three datasets from 24 participants: a controlled Lab Dataset (2,880 samples), a Real-World Dataset (3,115 samples), and a Security Dataset for testing against mimicry attacks.
Key Results
- Accuracy and Latency: SPU achieved a peak authentication accuracy of 96.3% with a decision latency of only 2.4 milliseconds.
- User Experience: In real-world scenarios, SPU reduced manual unlock requirements by 32.9%.
- Security Robustness: Against "Educated" mimicry attacks, SPU maintained a 0% False Acceptance Rate (FAR) at specific thresholds.
- Resource Efficiency: The system's background operation adds only 2% to daily power consumption.
Conclusion and Future Work
SPU provides a robust, low-overhead solution for implicit smartphone authentication. Its primary limitation is the requirement for a "stable start" state (e.g., picking up from a table), which covers roughly 35.6% of unlocks. Future research will focus on extending SPU to handle more complex start states (like pockets or bags), multi-user profile support, and cross-device applications for tablets and wearables.
Summary updated