Unmasking the Artificial: Forensic Defense Against Deepfake Imagery
Unmasking the Artificial: Forensic Defense Against Deepfake Imagery
As generative AI technologies like GANs and diffusion models reach unprecedented levels of photorealism, the line between genuine photography and synthetic fabrication has blurred, creating a crisis of trust in digital media. This presentation explores the evolving landscape of digital image forensics, moving beyond simple visual inspection to rigorous scientific detection methods. We will examine the principles of “Camera Ballistics” and Photo Response Non-Uniformity (PRNU), which use unique sensor noise to fingerprint specific devices with court-admissible accuracy. We will contrast these established methods with emerging AI-specific detection techniques, such as frequency domain analysis and model fingerprinting, which can attribute synthetic images to specific generators like DALL-E or Midjourney. Finally, we will discuss the shift toward cryptographic provenance via the C2PA standard, differentiating between hardware-embedded credentials and software-based assertions. Attendees will leave with a practical understanding of the tools and methodologies required to authenticate digital media and maintain the chain of custody in an increasingly synthetic world.
Unmasking the Artificial: Forensic Defense Against Deepfake Imagery - SLIDES (2/26/2026)
Unmasking the Artificial: Forensic Defense Against Deepfake Imagery - NOTES (2/26/2026)
« Back to Presentations