Schedule Your Free Security Assessment

UnHacked Episode 52 Summary

Guest Expert: Dr. Ilkay Demir

  • Expert in trusted media and deepfake detection at Intel
  • Creator of FakeCatcher, a groundbreaking deepfake detection tool
  • PhD in Computer Science from Purdue University with experience at Facebook, Pixar, and Intel
  • ACM Distinguished Speaker who advocates for content authenticity and creator rights

What Are Deepfakes?

Deepfakes are AI-generated fake content (images, videos, audio) where actors or actions aren't real. Unlike simple "shallow fakes" created with basic editing, deepfakes use sophisticated neural networks and deep learning algorithms to create highly convincing forgeries.

Business Risks from Deepfakes:

  1. Financial Fraud: Cases where executives' voices were deepfaked to authorize fraudulent money transfers
  2. Reputation Damage: Defamation campaigns against businesses or executives
  3. Political Manipulation: Misinformation campaigns affecting public opinion
  4. Adult Content: The most prevalent use of deepfakes (97% in 2019) with harmful implications
  5. Identity Theft: Impersonation of business leaders to damage reputation or manipulate stakeholders

Detection Technologies:

  • FakeCatcher: Uses blood flow signals (photoplethysmography/PPG) to detect fakes with 96% accuracy
  • Other Methods: Eye gaze tracking, motion analysis, and multimodal detection (correlating movement and voice)
  • These biological signals are difficult for AI to replicate, making detection more future-proof

Protection Strategies for Businesses:

  1. Content Authentication:
    • Implement C2PA (Coalition for Content Provenance and Authenticity) standards
    • Add provenance information to your content so consumers know its origin
    • Major platforms (Adobe, Facebook, LinkedIn) are already supporting content credentials
  2. Technological Protection:
    • My Art My Choice: Protects content from being stolen by generative models
    • My Face My Choice: Prevents facial identity theft
    • My Voice My Choice: Prevents voice cloning
    • These protection methods embed "landmines" in content that break AI's ability to replicate it
  3. Verification Methods:
    • For live calls: Ask participants to occlude their face, change expressions, or alter lighting
    • Use multiple detection algorithms for important content
    • Stay informed about the latest protection technologies

Key Takeaways:

  1. We are not doomed - there's significant work happening on the "good side" of AI
  2. Add provenance information to your content through C2PA standards
  3. Don't believe everything you see or hear online - use context and visual inspection
  4. Proactively protect your content with available protection technologies
  5. Small business owners must stay aware of evolving threats rather than assuming "it won't happen to me"

The podcast emphasizes that businesses of all sizes need to be proactive in protecting not just their networks and hardware, but also their digital identity, online presence, and content authenticity in the age of deepfakes.

Schedule Your Free Security Assessment