In the project "Image Forgery Detection," I developed a robust system aimed at detecting Copy-Move forgery, a commonly employed technique in image manipulation.
The purpose of choosing this project is:
- Digital Images Forensics (DIF): Vanguard of security techniques aiming at restoration of lost trust in digital imagery by exposing digital forgery techniques.
- Existing Techniques: Explore active and passive (blind) approaches in image forgery detection.
- Validation: Validate the originality of digital images by recovering information about their history.
- Trust Building: Analyze images under specific conditions to build trust and genuineness.
The proposed system utilizes SVM classifier for forgery detection, employing hashing techniques and RSA key encryption for security. The methodology involves two main phases: training and testing.
-
Training Phase:
- Database Creation: A database of images is created for training purposes. Images are sourced from various online repositories or captured using digital cameras. Images can vary in size and format (jpg, jpeg).
- RSA Key: An RSA key is generated after training images are ingested into the system. During testing, users are prompted to enter a consistent key to ensure authorized access.
- Pre-processing: Images undergo pre-processing steps such as conversion to grayscale from RGB, noise removal using median filtering, and enhancement techniques like histogram equalization and sharpening.
- Feature Extraction: Various image features are extracted including:
- Pixel Analysis: Calculation of mean and standard deviation of pixel values.
- Texture Analysis: GLCM (Gray-Level Co-occurrence Matrix) analysis for texture representation using Haralick functions.
- Hash Values: Hash values are computed for the extracted features to facilitate efficient comparison and identification of duplicated or manipulated regions within images.
- SVM Classifier: Support Vector Machine (SVM) classifier is trained using labeled datasets to establish decision boundaries and identify fraudulent image regions with high precision.
-
Testing Phase:
- Input Query Image: Users provide a query image to be authenticated.
- RSA Key Authentication: Users are prompted to enter the consistent RSA key generated during training for authentication.
- Pre-processing and Feature Extraction: Similar pre-processing and feature extraction steps are performed on the query image.
- Hash Values Calculation: Hash values are computed for the extracted features of the query image.
- SVM Classification: SVM classifier is utilized to classify the query image based on the decision boundaries established during training.
- High Accuracy: Achieved a remarkable 95% accuracy rate in identifying forged images, showcasing the robustness and reliability of the detection algorithms.
- Effective Detection: Successfully detected instances of Copy-Move forgery, a challenging form of image manipulation commonly employed to deceive viewers.
To get started with the project, follow these steps:
- Clone the repository:
git clone https://github.com/AayushiAhlawat/Image-Forgery-Detection.git - Run the main script:
python Implementation.py
The Image Forgery Detection project demonstrates the effectiveness of machine learning techniques, specifically SVM classifiers, in identifying instances of digital image forgery. With a focus on Copy-Move forgery, the system achieves high accuracy rates and provides a reliable solution for image authenticity verification.