Image Forgery detection using deep learning

IMAGE FORGERY DETECTION MODEL USING CNN- a deep learning model to discover forgeries like copy move, splicing, retouching etc;

1. Seamless Digital Image Forgery Detection Model using Deep Learning A Project By: 1SI21AD026 Inzamam Ali 1SI21AD035 Nafees Siddiquie 1SI21AD061 Shreya Arya Under the Guidance of: Ms Shweta A N

2. Agenda Introduction Motivation Literature Survey Problem Statement Objectives High Level Design System Implementation Results and Snapshots Tools and Plateform Expected Outcomes Conclusion

3. Introduction With over 4.5 billion active internet users, the amount of multimedia content being shared every day has surpassed everyone’s imagination. Large-scale and pervading social media platforms, along with easily accessible smartphones, have given rise to huge visual data such as images etc. One of the perils accompanying this huge amount of data is manipulation and malicious intent to remove the authenticity of these media. The availability of various image-editing software and tools such as Photoshop, GIMP, etc., has made it possible to create forgeries with minimal effort. Today, it is quite easy to produce a manipulated media that looks indifferent to the human eyes. Various types of digital image forgeries have evolved, the major ones include copy-move, morphing, watermarking, etc. But there is no reliable way of detecting these changes. We can create a model using Deep Learning to tackle this proble m.

4. Motivation Growing Concerns about Digital Media Authenticity Limitations of Traditional Methods Need for Advanced Deep Learning Algorithms Development of User-Friendly Web Application

5. Literature Survey Name Author Year Key Findings A Deep Learning Model for Splicing Image Detection Hoang Linh Nguyen, Kha Tu Huynh 2022 Image splicing, a poses challenges in maintaining the authenticity of digital content, particularly in social networks and online platforms. Convolutional block attention based network for copy-move image forgery detection M. Sabeena 2023 The algorithm utilizes the Convolutional Block Attention Module (CBAM) for feature extraction, segmentation, and forgery localization. . Estimation and Concealment Deep Fake Detection in Images using Hybrid LSTM Sunil Kumar Sharma Waseem Ahmad Khan 2024 The proposed HLSTM-ELM techniques hold promise for the development of efficient and reliable systems for identifying deepfakes.

6. Problem Statement With the advent of social networking services such as Facebook and Instagram, there has been a huge increase in the volume of image data generated in the last decade. Use of image (and video) processing software like GNU Gimp, Adobe Photoshop to create doctored images and videos is a major concern for internet companies like Facebook. These images are prime sources of fake news and are often used in malevolent ways such as for mob incitement. Before action can be taken on basis of a questionable image, we must verify its authenticity.

7. Objectives T o find effective and efficient techniques to accurately detect and identify manipulated images. These techniques should be able to detect various types of image forgeries such as copy-move, splicing, and retouching. There should be no limitations on image type i.e. png , jpg, jpeg etc. Implement a user-friendly interface for easy interaction and deployment .

8. High Level Design

9. System Implementation 1. Data Collection: Utilized the CASIA 2 dataset, comprising authentic and tampered images. Preprocessed the images for consistency and quality. 2. Model Development: Selected convolutional neural networks (CNN) for feature extraction. Implemented image forgery detection algorithm using TensorFlow. Trained the model on the preprocessed CASIA 2 dataset. 3. Model Evaluation: Split the dataset into training and validation sets. Evaluated the model's performance using accuracy, precision, recall, and F1-score. Fine-tuned hyperparameters to optimize performance.

10. 4. Integration with Flask Web App: Deployed the trained model using Flask framework for web integration. Created API endpoints for image upload and forgery detection. Integrated the model API into the Flask web app. 5. Frontend Design: Developed a user-friendly interface for image upload and result display using HTML, CSS, and JavaScript. Ensured responsive design for seamless user experience across devices. 6. Testing and Deployment: Conducted extensive testing to ensure accuracy and reliability. Deployed the web app on Hugging Face for accessibility.

11. Results and Snapshots Copy-Move forgery:

12. Splicing Forgery Retouching Forgery

13. Tools and Platform 1. Programming Languages: Python HTML/CSS Javascript 2. Libraries and Frameworks: TensorFlow/ Keras Flask Matplotlib Scikit-learn 3. Development Platform: Jupyter Notebook PyCharm Kaggle 4 . Dataset: CASIA2 Dataset

14. Outcomes

15. Conclusion In conclusion, the proposed system represents a comprehensive approach to image forgery detection by implementing a range of sophisticated algorithms. With its ability to accept input images and generate appropriate outputs for identifying forged content, the system addresses the pressing challenge of detecting digital manipulations. Its primary goal is to swiftly and accurately determine the authenticity of any given image, offering valuable support in law enforcement and cybersecurity contexts. By aiding users in distinguishing between authentic and tampered images, the system contributes significantly to enhancing digital image forensics capabilities, ultimately bolstering trust and reliability in digital media.