Abstract:The rapid evolution of digital image manipulation techniques poses significant challenges for content verification, with models such as stable diffusion and mid-journey producing highly realistic, yet synthetic, images that can deceive human perception. This research develops and evaluates convolutional neural networks (CNNs) specifically tailored for the detection of these manipulated images. The study implements a comparative analysis of three progressively complex CNN architectures, assessing their ability to classify and localize manipulations across various facial image modifications. Regularization and optimization techniques were systematically incorporated to improve feature extraction and performance. The results indicate that the proposed models achieve an accuracy of up to 76\% in distinguishing manipulated images from genuine ones, surpassing traditional approaches. This research not only highlights the potential of CNNs in enhancing the robustness of digital media verification tools, but also provides insights into effective architectural adaptations and training strategies for low-computation environments. Future work will build on these findings by extending the architectures to handle more diverse manipulation techniques and integrating multi-modal data for improved detection capabilities.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is **detecting and classifying facial image manipulation**, especially the challenges faced in digital media verification. With the rapid development of digital image processing technology, it has become increasingly easy to generate highly realistic synthetic images, which can deceive human perception, thus posing a significant challenge to the authenticity verification of content. Specifically, the paper focuses on the following aspects: 1. **Detection of highly - realistic synthetic images**: Current image - generation technologies (such as the Stable Diffusion model and the Midjourney model) can generate very realistic synthetic images that can deceive human perception. Therefore, effective detection methods need to be developed. 2. **Identification of multiple facial image manipulation techniques**: Facial image manipulation techniques (such as face fusion, splicing, face - swapping, and subtle expression adjustment) can distort reality and create misleading images. These techniques make image authenticity verification more complex, and methods that can identify these different types of manipulation need to be developed. 3. **Improving the generalization ability of existing models**: Existing deep - learning models (such as MesoNet) perform poorly when dealing with unseen data, especially in multi - class image - manipulation detection tasks. Therefore, the model architecture needs to be improved to enhance its generalization ability and classification performance. 4. **Efficient detection in an environment with limited computational resources**: To ensure that the model can run in an environment with limited computational resources, research needs to optimize computational efficiency while maintaining high performance. ### Solutions To solve the above problems, the paper proposes and evaluates three increasingly complex convolutional neural network (CNN) architectures, specifically: - **MesoNet+**: Two convolutional layers are added to the original MesoNet to enhance the feature - extraction ability, especially when dealing with unseen data. - **MesoMultiNet**: MesoNet is extended into a multi - class classification system that can identify multiple types of facial image manipulation. - **MesoMultiNet+**: The number of convolutional layers is further increased (from 4 layers to 6 layers), and the detection ability of edges, details, and general features is improved by optimizing the number and distribution of filters. Through these improvements, the paper shows a significant performance improvement of the proposed models in distinguishing real images from different types of manipulated images, with an accuracy rate of up to 76%, exceeding traditional methods. In addition, the research also explores effective architecture adaptation and training strategies to meet the needs of low - computational - resource environments. ### Summary This research not only highlights the potential of CNN in enhancing the robustness of digital media verification tools but also provides valuable insights for future work, including expanding the architecture to handle more diverse manipulation techniques and integrating multi - modal data to improve detection capabilities.

Detecting Facial Image Manipulations with Multi-Layer CNN Models

Multiple Manipulation Detection in Images Using Frequency Domain Features in 3D-CNN

3D CNN Architectures and Attention Mechanisms for Deepfake Detection

Image Manipulation Localization Using Attentional Cross-Domain CNN Features

A Deep Learning Approach to Universal Image Manipulation Detection Using a New Convolutional Layer

Detecting image manipulation with ELA-CNN integration: a powerful framework for authenticity verification

Identify Videos with Facial Manipulations Based on Convolution Neural Network and Dynamic Texture

SDCN2: A Shallow Densely connected CNN for multi-purpose image manipulation detection

Extracting deep local features to detect manipulated images of human faces

Analysis of adversarial attacks against CNN-based image forgery detectors

On the Effectiveness of Image Manipulation Detection in the Age of Social Media

Face Manipulation Detection Via Auxiliary Supervision

Multi-spectral Class Center Network for Face Manipulation Detection and Localization

Image Manipulation Detection with Cascade Hierarchical Graph Representation

Learning Forgery Region-Aware and ID-Independent Features for Face Manipulation Detection

MVSS-Net: Multi-View Multi-Scale Supervised Networks for Image Manipulation Detection

A multi-purpose image forensic method using densely connected convolutional neural networks

Detect and Locate: Exposing Face Manipulation by Semantic- and Noise-level Telltales

Recurrent Convolutional Strategies for Face Manipulation Detection in Videos

Hybrid LSTM and Encoder–Decoder Architecture for Detection of Image Forgeries

Fake face detection via adaptive manipulation traces extraction network