Improved Face Spoof Detection using GLCM with Stride and Color Space Variations using Machine Learning Algorithm

  • Sudeep D. Thepade, Shalakha V. Bang, Piyush R. Chaudhari, Mayuresh R. Dindorkar

Abstract

A one-word answer to the question ‘Who are you?’ in today’s world is indeed ‘Biometrics’. The two forms of biometrics are Physiological and Behavioral. Due to their ability to uniquely identify and verify an individual based on characteristics which are specific to them, they are widely used in security systems. Among multiple modes of biometrics, use of face recognition systems is rampant. These systems are reliable and promising, but they have a serious inadequacy. The shortcoming is that they are subjected to Identity theft. The attacks are often called as Face spoofing or Presentation attacks. To ward off this fraud, it is essential to advance current face recognition systems by refining their ability to ascertain the fraud.  This paper introduces an easier yet effective method for face anti-spoofing which is solely based on extraction of Haralick features from Gray Level Co-occurrence Matrix (GLCM) of an image. The image is represented in various color spaces: RGB, grayscale, YCrCb, YUV, Kekre-LUV. The proposed method extracts various texture parameters from the gray level co-occurrence matrix of an image using varied color spaces. In order to obtain a feature vector for particular color space, features of its individual channels are cascaded. The extracted features are classified using RandomForest classifier. Experiments done on two at hand datasets namely NUAA and REPLAY-ATTACK shows that proposed methodology surpasses other state-of-the-art methods in terms of accuracy.

Published
2020-03-30
How to Cite
Sudeep D. Thepade, Shalakha V. Bang, Piyush R. Chaudhari, Mayuresh R. Dindorkar. (2020). Improved Face Spoof Detection using GLCM with Stride and Color Space Variations using Machine Learning Algorithm. International Journal of Advanced Science and Technology, 29(3), 10247 - 10261. Retrieved from https://sersc.org/journals/index.php/IJAST/article/view/27086
Section
Articles