Dissertations, Theses, and Capstone Projects
Date of Degree
9-2025
Document Type
Doctoral Dissertation
Degree Name
Doctor of Philosophy
Program
Computer Science
Advisor
Shweta Jain
Committee Members
Liang Zhao
Ashwin Satyanarayana
Nasir Memon
Subject Categories
Computer and Systems Architecture | Other Computer Engineering | Signal Processing
Keywords
Perceptual Hashing, Image Manipulation Detection, Digital Forensics, Robustness, Generalizability, Statistical Framework
Abstract
This thesis contributes to research in adversarial image manipulation detection. The primary motivation is the increasing need to verify digital images, especially for legal evidence, journalistic proof, or social media content—where manipulated or fabricated images can mislead, defame, or distort reality. A key application and contribution of this work is the development of eWitness, a blockchain application that generates and registers image provenance at capture time to enable independent verification of authenticity. The secret sauce behind the system is SmartHash, a novel and efficient perceptual hashing algorithm designed for real-world deployment in systems like eWitness. Unlike existing algorithms, SmartHash targets both adversarial image manipulation detection and traditional uses of perceptual hashing such as duplicate detection, reverse image search, identifying child sexual abuse material, and tagging or removing inappropriate content. In current literature, threshold-based classification remains the dominant strategy for such image classification tasks. However, our analysis shows that identifying a universally optimal threshold is impractical—especially for subtle modifications where threshold-dependent algorithms fail. This insight led to the development of a statistical framework that classifies image pairs using class-specific probability density functions and maximum-likelihood estimation. The framework performs categorical classification, provides confidence scores for explainable results, supports any perceptual hashing algorithm including pHash, NMF-based hashing, Microsoft’s PhotoDNA, and Apple’s NeuralHash, and enables robust, dataset-independent decision-making grounded in statistical reasoning. When paired with SmartHash, the system delivers strong performance and interpretability. Together, eWitness, SmartHash, and the generalized framework offer holistic contributions—practical, algorithmic, and statistical—that advance the state of the art in image manipulation detection.
Recommended Citation
Samanta, Priyanka, "Toward a Generalizable Perceptual Hashing
Framework for Image Manipulation Detection" (2025). CUNY Academic Works.
https://academicworks.cuny.edu/gc_etds/6448
Included in
Computer and Systems Architecture Commons, Other Computer Engineering Commons, Signal Processing Commons
