An Inter-Frame Forgery Detection Technique for Surveillance Videos Based on Analysis of Similarities

Authors

  • Anas Abdullahi Department of Computer Science, Faculty of Sciences Sokoto State University, Sokoto – Nigeria.
  • Mustapha Aminu Bagiwa Department of Computer Science, Faculty of Physical Sciences Ahmadu Bello University, Zaria – Nigeria.
  • Abubakar Roko Computer Science Unit, Department of Mathematics Usmanu Danfodiyo University Sokoto, Nigeria
  • Samaila Buda Department of Physics, Faculty of Sciences Usmanu Danfodiyo University Sokoto, Nigeria
  • Zaharadeen Yusuf Yeldu Computer Science Unit, Department of Mathematics Usmanu Danfodiyo University Sokoto, Nigeria
  • Aliyu Muhammad Bello Computer Science Unit, Department of Mathematics Usmanu Danfodiyo University Sokoto, Nigeria
  • Halimatu Abubakar Ainu Department of Computer Science, Faculty of Sciences Sokoto State University, Sokoto – Nigeria.

DOI:

https://doi.org/10.56471/slujst.v4i1&2.265

Keywords:

Digital video, forensics, passive technique, Inter-frame forgery, Surveillance videos, HSV colour histogram

Abstract

Background: In video forgeries, the insertion, duplication and deletion of frames are the most common forgeries that are used by attackers to alter targeted videos for malicious intent. Researchers have proposed the use of active and passive technologies for detecting video forgeries over the years. Active approaches are used to detect the occurrence of alterations in digital video with the use of embedded features such as digital signature and watermarks. However, techniques that are based on active approaches are only applicable to specialized hardware devices. A passive technique, on the other hand, detects forgery using the behavioral cues encoded in a video. In this paper, a passive video forgery detection system based on frame similarity analysis is presented.Inter frame forgeries (Insertion, Deletion, and Duplication) were detected using the proposed technique, which was unaffected by scene changes.The technique has the overall performance of 98.07% precision, 100% recall and 99.01% accuracy.

Published

2022-07-20