GaitAuth

Implementation of Authentication Scheme using Gait Signals Captured from Accelerometer. The full paper is available here (free version here).

Prerequistite Library

1. libSVM. If the latest version does not work, please try v.3.15

Usage

The main functions are located in mainAuthentication.m (for authentication scheme) and mainRecognition.m (for recognition scheme)

Dataset

IMPORTANT NOTE: The dataset is only used for academic or research purposes. The authors do not allow anyone to use this dataset for any commercial purposes.

The dataset is located in DATASET folder, which contain gait signals of 38 users and is collected according to the following configurations

• Device: Google Nexus One Mobile Phone
• Platform: Android 2.3
• Physical sensors used: Accelerometer , Magnetometer
• Sampling rate: 27 Hz with the configuration of SENSOR_DELAY_FASTEST in Android SDK
• Number of Participants: 38

Data in each walking session is recorded and saved in separate files which are named according to following pattern: UserID_Gender_SessionID_[DeviceName]DataType_SessionSequence.txt.

For example:

ID26_M_173358_[NexusOne]BMA 150 3-axis Accelerometer_10.txt

contains the raw acceleration data of user #26 who is Male, recorded in the session #173358 with Sequence #10.

There are three types of data in this dataset:

1. BMA 150 3-axis Accelerometer: raw acceleration data which also capture the influence of gravity. Each line consists of 4 components separated by a comma (,) including (i) the timestamp, (ii) the x-dimensional value a_x, (iii) y-dimensional value a_y, and (iv) z-dimensional value a_z. Those values are in g unit (i.e. 1g = 9.8 m/s2).

2. Linear Acceleration Sensor: acceleration data in which the influence of gravity is completely removed. Each line consists of 4 components separated by a comma (,) including (i) the timestamp, (ii) the x-dimensional value a_x, (iii) y-dimensional value a_y, and (iv) z-dimensional value a_z. Those values are in g unit (i.e. 1g = 9.8 m/s2).

3. Rotation Matrix: contains 9 elements in the 3x3 rotation matrix as presented in the paper.

Citing

If the code and the dataset are found useful, we would be appreciated if our paper can be cited with the following bibtex format

@inproceedings{DBLP:conf/secrypt/HoangCN15,
  author    = {Thang Hoang and
               Deokjai Choi and
               Thuc Dinh Nguyen},
  title     = {On the Instability of Sensor Orientation in Gait Verification on Mobile
               Phone},
  booktitle = {{SECRYPT} 2015 - Proceedings of the 12th International Conference
               on Security and Cryptography, Colmar, Alsace, France, 20-22 July,
               2015.},
  pages     = {148--159},
  year      = {2015},
  crossref  = {DBLP:conf/secrypt/2015},
  url       = {https://doi.org/10.5220/0005572001480159},
  doi       = {10.5220/0005572001480159},
  timestamp = {Sun, 21 May 2017 00:21:13 +0200},
  biburl    = {https://dblp.org/rec/bib/conf/secrypt/HoangCN15},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}

Further Information

For any inquiries, bugs, and assistance regarding using the code, please contact me at hoangm@mail.usf.edu.