Dual-Branch Convolutional Framework for Spatial and Frequency-Based Image Forgery Detection

Authors: Naman Tyagi, Riya Jain

Published: 2025-09-05 17:41:57+00:00

Comment: 14 pages, 5 figures

AI Summary

This paper introduces a dual-branch convolutional framework for digital image forgery detection that integrates spatial and frequency-based features. The proposed system employs a Siamese network to fuse these features, generating 64-dimensional embeddings for classification. When benchmarked on the CASIA 2.0 dataset, it achieved an accuracy of 77.9%, demonstrating a balance between computational complexity and detection reliability for practical deployment.

Abstract

With a very rapid increase in deepfakes and digital image forgeries, ensuring the authenticity of images is becoming increasingly challenging. This report introduces a forgery detection framework that combines spatial and frequency-based features for detecting forgeries. We propose a dual branch convolution neural network that operates on features extracted from spatial and frequency domains. Features from both branches are fused and compared within a Siamese network, yielding 64 dimensional embeddings for classification. When benchmarked on CASIA 2.0 dataset, our method achieves an accuracy of 77.9%, outperforming traditional statistical methods. Despite its relatively weaker performance compared to larger, more complex forgery detection pipelines, our approach balances computational complexity and detection reliability, making it ready for practical deployment. It provides a strong methodology for forensic scrutiny of digital images. In a broader sense, it advances the state of the art in visual forensics, addressing an urgent requirement in media verification, law enforcement and digital content reliability.

Key findings

The proposed model achieved an accuracy of 77.9% on the CASIA 2.0 dataset, outperforming traditional statistical methods. It balances computational complexity with detection reliability, making it suitable for practical deployment, despite acknowledging relatively weaker performance compared to larger, more complex forgery detection pipelines.

Approach

The proposed method utilizes a dual-branch convolutional neural network to extract features from both spatial and frequency domains of an image. These extracted features are then fused and fed into a Siamese network, which is trained using a contrastive loss function to learn discriminative 64-dimensional embeddings, minimizing distances for genuine image patches and maximizing for dissimilar (spliced) ones.

Datasets

CASIA 2.0

Model(s)

Dual-branch Convolutional Neural Network (CNN) architecture with a Siamese network. It incorporates learnable high-pass filters for noise extraction, Local Binary Patterns and statistical measures for spatial features, and Discrete Cosine Transform (DCT) coefficients, JPEG artifact patterns, Fourier analysis, and wavelet decomposition for frequency features.

Author countries

India

← Previous