NOBLE Research: Tigris’s Guide
Your role on this project is Research Lead. Your work feeds directly into the paper and the policy brief we deliver to NOBLE.
Your research files live in research/tigris/ on your branch (tigris-mccauley).
Your Tracks
1. Literature Review
Build an annotated bibliography (15+ sources) for the Related Work section of our paper.
Focus areas:
- AI generated image detection methods
- Image forensics in law enforcement
- How image quality (compression, blur, noise) affects detection accuracy
- Legal admissibility of digital evidence
Starting papers:
- Wang et al. (2020), “CNN-generated images are surprisingly easy to spot…for now”
- Corvi et al. (2023), “On the detection of synthetic images generated by diffusion models”
- Groh et al. (2022), “Deepfake detection by human crowds, machines, and machine-informed crowds”
- Farid (2022), “Creating, Using, Misusing, and Detecting Deep Fakes”
- Mendones v. Cushman (case law)
Read these first, then follow their citations.
2. Detection Tool Profiles
Research each tool we are evaluating so we can write about them accurately in the paper.
| Tool | Type |
|---|---|
| Hive Moderation | Commercial API |
| Illuminarty | Web tool |
| AI or Not | Web/API |
| HuggingFace AI Image Detector | Open source model |
| SynthID (Google) | Watermarking |
| Optic AI or Not | API |
For each tool: how it works, what it was trained on, known limitations, pricing, and any published accuracy claims.
3. Legal and Policy Research
This feeds the 2 page policy brief we deliver to NOBLE.
Questions to answer:
- How do courts currently authenticate digital evidence? (Daubert, Frye, Federal Rules of Evidence)
- What cases have involved AI generated evidence?
- What guidance exists from NOBLE, IACP, DOJ, FBI, NIST?
- What is the “liar’s dividend” and how does it apply here?
4. Image Quality Control (later)
When the team starts generating synthetic images, you will review samples for quality. Flag off topic images, obvious artifacts, wrong style, and duplicates. This starts when image generation is underway.
Git Workflow
# First time
git clone https://github.com/ashleyscruse/ai-generated-image-detection.git
cd ai-generated-image-detection
git checkout tigris-mccauley
# Before you start working
git pull origin main
# After making changes
git add research/tigris/
git commit -m "Added 3 literature summaries"
git push origin tigris-mccauley
Your files are markdown (.md). You can edit them in VS Code, any text editor, or directly on GitHub.
Getting Help
If you are stuck for more than 15 minutes, message Dr. Scruse.