Journal Article
Computer Vision

A New Method to Detect Splicing Image Forgery Using Convolutional Neural Network

Khalid M. Hosny(Department of Information Technology, Zagazig University, Zagazig 44519, Egypt), Akram M. Mortda(Department of Information Technology, Faculty of Information Technology and Computer Science, Sinai University, Arish 16020, Egypt), Nabil A. Lashin(Department of Information Technology, Zagazig University, Zagazig 44519, Egypt), Mostafa M. Fouda(Department of Electrical and Computer Engineering, Idaho State University, Pocatello, ID 83209, USA)
January 18, 2023Applied Sciences34 citations
DOI

34

Citations

1

Influential Citations

Applied Sciences

Venue

2023

Year

Abstract

Recently, digital images have been considered the primary key for many applications, such as forensics, medical diagnosis, and social networks. Image forgery detection is considered one of the most complex digital image applications. More profoundly, image splicing was investigated as one of the common types of image forgery. As a result, we proposed a convolutional neural network (CNN) model for detecting splicing forged images in real-time and with high accuracy, with a small number of parameters as compared with the recently published approaches. The presented model is a lightweight model with only four convolutional layers and four max-pooling layers, which is suitable for most environments that have limitations in their resources. A detailed comparison was conducted between the proposed model and the other investigated models. The sensitivity and specificity of the proposed model over CASIA 1.0, CASIA 2.0, and CUISDE datasets are determined. The proposed model achieved an accuracy of 99.1% in detecting forgery on the CASIA 1.0 dataset, 99.3% in detecting forgery on the CASIA 2.0 dataset, and 100% in detecting forgery on the CUISDE dataset. The proposed model achieved high accuracy, with a small number of parameters. Therefore, specialists can use the proposed approach as an automated tool for real-time forged image detection.

Analysis

Why This Paper Matters

Image forgery, particularly splicing, poses a significant threat to the credibility of digital media in forensics, medical imaging, and social networks. Existing deep learning solutions often require substantial computational resources, limiting their deployment in real-time or edge environments. This paper addresses that gap by introducing a lightweight CNN that achieves near-perfect accuracy while using only four convolutional layers. The ability to run on resource-constrained hardware without sacrificing performance is a critical step toward practical, widespread adoption of forgery detection tools.

Furthermore, the paper benchmarks against multiple standard datasets (CASIA 1.0, CASIA 2.0, CUISDE), providing a comprehensive evaluation that demonstrates robustness across different forgery scenarios. The high accuracy (up to 100%) on CUISDE suggests the model generalizes well to unseen splicing patterns, which is essential for real-world deployment.

Technical Contributions

  • Lightweight Architecture: The model uses only four convolutional layers and four max-pooling layers, drastically reducing parameter count compared to deeper networks like VGG or ResNet.
  • Real-Time Capability: The small model size enables inference on devices with limited memory and processing power, such as smartphones or embedded systems.
  • High Accuracy: Despite its simplicity, the model achieves 99.1% on CASIA 1.0, 99.3% on CASIA 2.0, and 100% on CUISDE, outperforming many heavier models.
  • Comprehensive Evaluation: Sensitivity and specificity metrics are reported, providing a fuller picture of model performance beyond just accuracy.

Results

The proposed CNN achieves 99.1% accuracy on CASIA 1.0, 99.3% on CASIA 2.0, and 100% on CUISDE. These results are competitive with or superior to state-of-the-art methods that often use deeper architectures. The model's small parameter count ensures fast inference, making it suitable for real-time applications. The paper also reports high sensitivity and specificity, indicating low false positive and false negative rates.

Significance

This work demonstrates that high-performance forgery detection does not require massive models. By prioritizing efficiency, the authors open the door for deploying AI-based forensic tools in environments where computational resources are scarce, such as mobile devices, IoT cameras, or field forensic kits. The approach could be extended to other types of image manipulation (e.g., copy-move, retouching) and integrated into automated verification pipelines for social media or legal evidence. However, future work should test robustness against adversarial attacks and more diverse forgery types to ensure reliability in adversarial settings.

image-forgerysplicing-detectionconvolutional-neural-networklightweight-modelforensicsdeep-learning
Neural Networks

Related Papers