Image Hash Minimization for Tamper Detection

Abstract

Tamper detection using image hash is a very common problem of modern days. Several research and advancements have already been done to address this problem. However, most of the existing methods lack the accuracy of tamper detection when the tampered area is low, as well as requiring long image hashes. In this paper, we propose a novel method objectively to minimize the hash length while enhancing the performance at low tampered area.

Methodology

The methodology described in the paper uses k-means clustering on SURF features from the image to generate a robust hash. The same hash is used to determine if the image has been tampered or not.

Results

A sample of detected SURF features in original (left) and tamprerd (right) images.

Plot of L₂ distances between k-means centroids of tampered images with tampered area more than 30% and its corresponding original iamge.

Plot of L₂ distances between k-means centroids of tampered images with tampered area less than 5% and its corresponding original iamge.

Variation in mean L₂ distances with respect to the number of clusters, k for k-means clustering.

Determining the Threshold for L₂ distance for Hash Mactching: The threshold is determined as an optimal point so that the strategy can distinguish tampered images robustly without getting affected by the image content preserving operations.

BibTeX

@inproceedings{maity2017image,
title={Image Hash Minimization for Tamper Detection},
author={Maity, Subhajit and Karsh, Ram Kumar},
booktitle={Ninth International Conference on Advances in Pattern Recognition (ICAPR)},
year={2017}}

Image Hash Minimization

for Tamper Detection

A basic idea of what the paper is about. The goal is to detect if an image has been tampered or not using efficient hashing method. Here is a depiction of successful detection of tampered image.