Behavioral biometric authentication, unlike traditional biometric methods such as facial or fingerprint recognition, relies on unique behaviors such as gait or keyboard typing to verify the identity of individuals. As part of this authentication, it is important to collect behavioral biometric data under real-life conditions to train AI models. This data collection is carried out through sensors embedded in smartphones and watches.
This research work proposes a data collection methodology for behavioral biometric authentication, focusing on real-life conditions of use. Using a survey we have set up, we analyze the usual wearing positions of the phone and watch, and plan data collection based on these usage habits.
The results of the survey on usual phone and watch carrying positions highlighted users’ behaviors and preferences when carrying these devices on the move. The survey revealed that walking is the activity in which users most frequently carry their phone, often in their trouser pocket. The watch can be seen as a secondary collection device, offering an opportunity to collect additional data.
Based on these results, a plan for data collection was developed, incorporating a protocol adapted to actual user preferences. This protocol takes advantage of the phone’s positioning habits. Implementation of this protocol will optimize the data collected by capturing it in real-life conditions, enabling AI models to be developed based on authentic behavior.
| Date | 25 Nov 2024 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Vincent Lévesque (Supervisor) |
|---|
Ntsako Guiegou, P.-P. (Author),
Lévesque (Supervisor),
25 Nov 2024Student thesis: Master's thesis › Master in Engineering: Information Technology Engineering