Challenge background
With the rise of digital media consumption in the Philippines, where over 80% of the population has internet access and more than 70% are active mobile internet users, social media plays a crucial role in information dissemination. However, this ease of access also comes with risks. Deepfake technology, which uses artificial intelligence to create hyper-realistic manipulated images and videos, has become a growing concern. Such technology can be weaponized to spread misinformation, manipulate public opinion, damage reputations, and influence political discourse.
During elections, national events, and major news cycles, deepfake content has the potential to mislead the public, creating confusion and distrust in media, institutions, and government bodies. Despite existing fact-checking organizations such as VERA Files, Rappler's Fact Check, and the Philippine Center for Investigative Journalism, deep fake images and videos continue to spread unchecked.
The problem
While numerous fact-checking initiatives focus on verifying textual misinformation, there remains a significant gap in detecting AI-generated deep fake content. There is no dedicated, automated system in the Philippines that can quickly analyze and verify whether an image or video has been digitally altered. The goal of this project is to create an AI-powered platform that can detect deep fake images and videos using computer vision models, verify their authenticity, and report them to relevant authorities and the public.
Types of Misinformation Addressed:
- Deepfakes and manipulated media
- Misinformation in visual content
- Political propaganda using AI-generated visuals
- Hoaxes and false claims using altered videos
- Fake endorsements or fabricated statements using deepfake technology
Goal of the project
The Project Goal
- Collect a dataset of verified deep fake and authentic media from Philippine news sources, government agencies, and fact-checking organizations.
- Conduct data preprocessing and analytics to understand the distribution of manipulated media.
- Apply AI, Machine Learning, and Computer Vision techniques to develop a deep face detection model.
- Develop an API that allows government agencies, media outlets, and the public to verify the authenticity of an image or video.
- Build a user-friendly web platform where individuals can upload and check media content for potential manipulation.
Project timeline
- 1
Week 1
Week 1: Data Collection & Preprocessing
- Gather datasets of known deepfake images and videos from Philippine news sources and fact-checking organizations.
- Collaborate with journalists and government agencies for dataset enrichment.
- Clean and label images/videos as authentic or manipulated.
- Implement face detection and feature extraction techniques.
- 2
Week 2
Week 2: Model Development & Training
- Train deepfake detection models using convolutional neural networks (CNNs) and deep learning techniques.
- Fine-tune detection accuracy using adversarial training and facial mapping.
- Test the model on various real-world deepfake samples.
- Optimize the model for performance and accuracy.
- 3
Week 3
Week 3: API & Platform Development
- Develop a public API that allows users to submit and verify images/videos.
- Integrate with government and media platforms for fact-checking.
- Build a user-friendly web interface for users to upload and check media authenticity.
- Implement user feedback mechanisms for improving detection accuracy.
- 4
Week 4
Week 4: Testing, Deployment & Awareness Campaign
- Conduct final testing and validation of the system.
- Deploy the platform for public and governmental use.
- Launch an awareness campaign to educate citizens on deepfake risks and encourage the use of the platform.
- Monitor and refine model accuracy through real-time testing and user feedback.
- 5
Week 5
Week 5: Platform Development & User Experience Optimization
- Build a user-friendly web interface for users to upload and check media authenticity.
- Implement user feedback mechanisms for improving detection accuracy.
- Conduct initial testing and collect user feedback for improvements.
- 6
Week 6
Week 6: Final Testing, Deployment & Awareness Campaign
- Conduct final testing and validation of the system.
- Deploy the platform for public and governmental use.
- Launch an awareness campaign to educate citizens on deepfake risks and encourage the use of the platform.
- Monitor and refine model accuracy through real-time testing and user feedback.
What you'll learn
The Learning Outcomes
- Data Collection: Scrape and source manipulated and authentic media.
- Data Cleaning and Preprocessing.
- Data Analysis and Model Development.
- Training Deep Face Detection Models using Computer Vision.
- API Development for real-time media verification.
- Front-end Web Development and Deployment.