Workshop: Build a Deepfake Learn the Defense

Seeing is no longer believing, and hearing isn’t either. In this four-hour, hands-on workshop, you will build a video deepfake yourself, using the same kinds of open-source and commercial-grade tools a real attacker would use. You’ll retrieve source footage of a consenting target and clone their voice from a short sample. You’ll sync that voice to their face and generate brand-new talking video from a single still photo. Each step saves a file that feeds the next part of the process.

By the end, you’ll have built the full fabrication chain: a synthetic face and a synthetic voice, synced together. Then you’ll watch a live, real-time face swap on a video call, the same class of attack that cost one multinational firm roughly $25 million.

Building it is how you learn where it breaks. At each stage, you’ll look for the audio and visual tells that detection depends on, and you’ll see why a detector alone won’t save you. We finish by turning what you built into a scoped, authorized social-engineering test, and into the verification controls that can stop these attacks.

Walk out with your own deepfake artifacts, a working cloud toolchain you can reuse, and first-hand evidence for your leadership, not vendor projections.

Who Should Take This Workshop

Security practitioners who need to answer the question “Could this happen to us?” with evidence and not a vendor slide.

• Red team operators, penetration testers, and social-engineering assessors who want to add deepfake pretexts to authorized engagements
• Security awareness and human risk management leads who need to show, not tell, what employees are up against
• Incident responders, SOC analysts, and fraud or threat-intel teams who need to recognize synthetic media when it shows up in a case
• Security leaders, GRC staff, or business leaders who brief executives and design verification procedures

No AI or machine-learning background is needed. If you work in security and can find your way around a web app and a file system, you should be good to go.

What You’ll Learn

By the end of the workshop, you will be able to:

• Explain how the modern video-deepfake attack chain works, from source retrieval to finished fake, and what each stage really costs an attacker in time and money.
• Retrieve and assess source media, and determine whether a target’s public footage is good enough to fake.
• Clone a voice from a short reference sample and hear where the clone falls apart.
• Lip-sync a cloned voice onto real video footage.
• Generate talking video from a single still image, then sync it to the cloned voice.
• Plan an authorized deepfake social-engineering test: pretext, delivery channel, scope and consent, evidence capture, and safe-abort criteria.
• Recommend out-of-band verification controls that hold up against voice and video impersonation.