L1
·
Quiz
·
Lab
L2
·
Quiz
·
Lab
L3
·
Quiz
·
Lab
L4
·
Quiz
·
Lab
Module Test
Lesson 1 · AI & Media — Module 2

What Are Deepfakes?

From face-swapping experiments to election interference — understanding the technology that synthesizes reality.
How does a machine learn to convincingly replace one person's face with another's — and what does that mean for truth?

President Ali Bongo Ondimba of Gabon had not appeared in public for months. Rumours of his death or incapacitation were destabilising the country. On New Year's Day 2019, state television broadcast a video of Bongo delivering a short address — stiff, wooden, oddly lit. Within days, a military faction cited the video as evidence of a deepfake and attempted a coup. Researchers who later analysed the footage disagreed: the video appeared genuine, merely poorly produced. The coup failed. But the episode revealed something new and dangerous: the mere possibility of a deepfake had become a weapon.

The Technology: GANs and Diffusion Models

The term deepfake was coined in late 2017 by a Reddit user who posted AI-generated face-swap videos of celebrities. The name fused deep learning with fake. Within months the technique had spread from research labs to consumer software.

The foundational architecture is the Generative Adversarial Network (GAN), introduced by Ian Goodfellow in 2014. A GAN pits two neural networks against each other: a generator that tries to create convincing synthetic images, and a discriminator that tries to detect fakes. They train in tandem — the generator improving until the discriminator can no longer tell real from fake.

By 2022, diffusion models — the architecture behind Stable Diffusion, DALL·E 2, and Midjourney — had largely superseded GANs for image generation. Diffusion models work by learning to reverse a process of adding noise to images: they are trained to denoise, and at inference time they begin from pure noise and iteratively refine it into a coherent image guided by a text prompt or reference image.

For video deepfakes, systems like DeepFaceLab and commercial successors combine face detection, landmark tracking, and neural re-rendering to transplant one person's facial movements and expressions onto another's head in every frame of a video.

GANGenerative Adversarial Network — two competing neural networks (generator + discriminator) trained together, producing synthetic outputs indistinguishable from real ones.
Diffusion ModelA generative model trained to reverse noise addition; capable of producing photorealistic images and video from text or image prompts.
Face ReenactmentA subset of deepfake technology that transfers the facial expressions and head movements of a source person onto a target's video, in real time or post-production.

A Brief History of Synthetic Media

Synthetic media did not begin with AI. Hollywood has used CGI faces since the 1990s — Forrest Gump shook hands with President Kennedy in 1994 using early compositing. But those techniques required millions of dollars and teams of specialists. What changed after 2017 was democratisation: deepfake creation moved from film studios to anyone with a consumer GPU and a few hours of source footage.

In 2018, the Belgian political party sp.a released a video of President Trump announcing US withdrawal from the Paris Climate Accord. The video was clearly labelled as synthetic — a piece of political commentary. But it demonstrated that convincing political deepfakes were now producible on a small budget. The same year, BuzzFeed and director Jordan Peele released a public-service deepfake of Barack Obama — voiced by Peele, wearing Obama's face — warning viewers not to believe everything they see online.

By 2023, real-time deepfake tools allowed a person on a video call to appear as a completely different individual, with facial movements tracked and re-rendered at 30 frames per second on commodity hardware.

Scale Marker

Sensity AI (a deepfake detection company) estimated that as of 2023, over 95% of all deepfake videos online depicted non-consensual intimate imagery — primarily women. Political and financial fraud deepfakes, while rarer, receive disproportionate media coverage relative to the larger crisis of image-based abuse.

Not Just Video: The Synthetic Media Ecosystem

Voice cloning has become equally accessible. Tools like ElevenLabs can clone a voice from as little as a one-minute audio sample. In 2023, a robocall using a cloned voice of President Biden urged New Hampshire Democrats not to vote in the primary — the audio was forensically identified as AI-generated but had already reached thousands of voters.

Synthetic text from large language models can produce fake news articles, fake academic citations, and fake social media personas at industrial scale. Image generation produces fake photographs of events that never occurred. Combined, these tools constitute what researchers now call the synthetic media ecosystem — an environment where every sensory channel of human communication can be fabricated.

Key Insight

The most dangerous aspect of deepfakes may not be the fakes themselves but the liar's dividend — the ability for any real video to be dismissed as AI-generated. Once audiences doubt the authenticity of all media, bad actors can deny genuine evidence of their wrongdoing.

Lesson 1 Quiz

What Are Deepfakes? — Check your understanding before the lab.
The term "deepfake" was coined in what year, and by whom?
Correct. The username "deepfakes" appeared on Reddit in late 2017, posting face-swap videos of celebrities using GAN technology.
Not quite. The term originated on Reddit in 2017. Ian Goodfellow invented GANs in 2014 but did not coin the term "deepfake."
In a Generative Adversarial Network, what role does the discriminator play?
Correct. The discriminator's job is to detect fakes — its failure drives the generator to produce more convincing outputs.
Not quite. The discriminator is the "judge" in a GAN — it learns to tell real from fake, forcing the generator to improve.
What is the "liar's dividend" as it relates to deepfakes?
Correct. Once synthetic media is widespread, real evidence can be denied by simply claiming it is a deepfake — a dangerous epistemic escape hatch.
The liar's dividend is the strategic benefit of living in a world where everything could be fake: real footage can now be credibly denied.
According to Sensity AI's 2023 estimate, what percentage of deepfake videos online depicted non-consensual intimate imagery?
Correct. The vast majority of deepfake content targets private individuals — primarily women — with non-consensual intimate imagery, not political figures.
The figure is over 95%. Political deepfakes dominate news coverage but are a small fraction of the actual deepfake ecosystem.

Lab 1: Spot the Fake

Deepfake Detection — AI-assisted analysis practice

Your Task

You are a junior analyst at a media verification desk. A video clip has been flagged as potentially synthetic. Your AI assistant has pre-analysed several frames and is ready to walk you through the forensic indicators. Ask it about specific visual or audio tells, discuss the Gabon 2019 case, or explore what current detection tools can and cannot do.

Starter prompts: "What are the most reliable visual tells in a deepfake video?" · "Why was the Gabon video hard to verify?" · "Can AI detectors be fooled?"
Verification Desk — AI Assistant
Deepfake Analysis
Footage flagged. I've run a preliminary frame analysis. There are three artefact zones worth examining — unnatural eye reflections, inconsistent ear geometry on turns, and a subtle flicker in the neck-jaw boundary. What would you like to dig into first?
Lesson 2 · AI & Media — Module 2

How Deepfakes Are Made

The technical pipeline — from data collection to synthesis — and why understanding it is essential for detection.
What does it actually take to fabricate a convincing video of a real person, and where in that process are the seams most likely to show?

A finance worker at a multinational firm received a video call invitation from someone claiming to be the company's UK-based CFO. On the call were several colleagues — all familiar faces. The CFO instructed him to transfer HK$200 million (approximately US$25 million) as part of a confidential transaction. He complied. Every face on that call was a deepfake. Hong Kong police later confirmed that attackers had used publicly available footage of the real employees to train face-synthesis models, then ran them live on the video call. It was the largest known deepfake financial fraud at the time.

Stage 1 — Data Collection

Every deepfake begins with source footage of the target — the person whose face will be rendered. The more footage available, and the greater its variety (different angles, lighting conditions, expressions), the more convincing the result. Public figures are especially vulnerable because hours of broadcast footage exist.

For the Hong Kong attack, attackers scraped video conference recordings, public interviews, and corporate videos. The US$25 million transfer required perhaps weeks of model training on commodity hardware.

The second requirement is footage of the driver — the person whose live expressions and movements will animate the target face. In real-time deepfakes (as used in the Hong Kong case), the attacker is the driver: their face is tracked by a webcam and mapped in real time onto the synthesised target face.

Stage 2 — Model Training

Using tools like DeepFaceLab (open source, continuously maintained) or commercial successors, an operator feeds aligned face images from source footage into a neural network. The network learns a compressed latent representation of the target's face — essentially, a mathematical description of how that face transforms across angles and expressions.

Training on a modern consumer GPU (e.g. an NVIDIA RTX 4090) takes between 12 and 72 hours depending on quality targets and dataset size. By 2024, cloud-based deepfake-as-a-service platforms had reduced this to minutes by using pre-trained base models fine-tuned on a small set of target images.

Latent SpaceA compressed mathematical representation learned by a neural network — in face synthesis, the "coordinates" that describe a face's geometry, texture, and expression state.
Fine-tuningAdapting a pre-trained model to a specific target with minimal new data — enabling rapid, cheap personalised deepfake creation.

Stage 3 — Synthesis and Blending

The trained model generates synthetic face frames that must be blended back into the original video. This is where most detectable artefacts arise. Blending requires:

Mask generation — defining exactly which pixels belong to the face region versus hair, ears, and background. Imperfect masks create the "face in a face" shimmer visible at edges in lower-quality deepfakes.

Colour correction — matching the synthesised face's skin tone and lighting to the background plate. Mismatches produce flat, plastic-looking skin.

Temporal consistency — ensuring that frame-to-frame, the synthesised face does not flicker or jitter. This is computationally expensive and is the primary limitation of real-time systems.

Technical Tell

Eye reflections (specular highlights) are among the most reliable forensic indicators. A real face in a real environment shows consistent reflections of light sources in both eyes. Synthesis models often produce incoherent or physically impossible reflections — a tell still reliable as of 2024 for non-commercial tools, though rapidly improving.

Stage 4 — Audio Synthesis

A convincing video deepfake requires matching synthetic audio. Voice cloning systems — including ElevenLabs, VALL-E (Microsoft, 2023), and open-source alternatives — can reproduce a target's voice from a few minutes of audio. VALL-E was demonstrated reproducing a speaker's voice (including their emotional tone and acoustic environment) from a 3-second clip.

In the Hong Kong fraud, the fake CFO's voice was indistinguishable to the victim. The attack succeeded partly because the company's verification protocols relied entirely on visual and auditory recognition — both of which the deepfake had defeated.

Defence Implication

The Hong Kong case led several multinational corporations to implement out-of-band verification for large transfers — requiring a separate, pre-established communication channel (e.g. a confirmed phone call to a known number) before any significant action is taken based on video instruction.

Lesson 2 Quiz

How Deepfakes Are Made — Technical pipeline comprehension.
In the February 2024 Hong Kong deepfake fraud, approximately how much money was transferred?
Correct. The HK$200 million (≈ US$25 million) transfer made this the largest known deepfake financial fraud at the time.
The amount was HK$200 million, approximately US$25 million — a landmark case in deepfake-enabled financial crime.
What is "fine-tuning" in the context of deepfake production?
Correct. Fine-tuning leverages a large pre-trained base model — requiring only a small dataset of the target — dramatically reducing the time and compute needed.
Fine-tuning refers to adapting a large pre-trained model using small amounts of target-specific data, enabling rapid and cheap personalised deepfake creation.
Which of the following is described as a reliable forensic indicator in deepfake detection as of 2024?
Correct. Eye specular highlights (reflections of light sources) are difficult for synthesis models to render consistently and physically plausibly.
Eye reflections are a key tell — synthesis models struggle to produce physically coherent specular highlights in both eyes simultaneously.
What is "temporal consistency" in the context of deepfake synthesis?
Correct. Frame-to-frame stability is computationally expensive and is the primary quality limitation of real-time deepfake systems.
Temporal consistency means keeping the synthesised face stable across frames — flickering is the most visible artefact in low-quality deepfakes.

Lab 2: The Fraud Call

Corporate deepfake scenarios — designing verification protocols

Scenario

You are a security consultant advising a multinational firm in the wake of the Hong Kong deepfake fraud. Your AI assistant has been briefed on the technical pipeline and the specific failure modes of the attack. Work through what verification protocols should have existed, and how they should be designed going forward.

Starter prompts: "What verification steps would have prevented the Hong Kong transfer?" · "How should companies authenticate video calls for high-stakes decisions?" · "What are the limits of deepfake detection software in real-time calls?"
Security Consultant — AI Briefing
Fraud Prevention
The Hong Kong case had three clear failure points: no out-of-band verification, no challenge-response protocol during the call, and complete reliance on visual/audio recognition. Where would you like to start building your recommendations?
Lesson 3 · AI & Media — Module 2

Deepfakes in Politics & Elections

When synthetic media targets democratic processes — documented cases, information warfare, and the limits of legal response.
If a convincing deepfake of a candidate can be released hours before polls open, what institutional defences remain?

Days before the New Hampshire Democratic primary, voters received robocalls featuring a voice that sounded unmistakably like President Biden. The voice told them: "Don't vote this Tuesday." Election officials moved quickly; the audio was identified as AI-generated within 24 hours. A Democratic political consultant named Steve Kramer later claimed responsibility, describing it as a "wake-up call" about AI vulnerability — though critics noted this framing conveniently minimised accountability. The FCC subsequently ruled that AI-generated voices in robocalls require explicit consent and disclosure.

The Political Deepfake Landscape

Political deepfakes and synthetic audio have appeared in documented elections across multiple continents in a short window. In Slovakia (September 2023), two days before a parliamentary election, an audio recording circulated on Facebook appearing to show liberal candidate Michal Šimečka discussing how to buy votes. The recording was assessed by fact-checkers as likely AI-generated, but it had spread to tens of thousands of listeners before labelling. Šimečka lost. Whether the audio changed the result cannot be established, but Meta's response — citing election-silence rules and not removing the audio — became a case study in platform failure.

In Taiwan (January 2024), AI-generated videos and audio of candidates circulated in the weeks before the presidential election. The Taiwanese government had enacted legislation requiring AI disclosure on political ads, but enforcement of viral social media content proved difficult. Taiwan's election integrity commission documented over 40 distinct synthetic media incidents in the campaign period.

In India (April–May 2024), during the world's largest election, deepfake videos of deceased political figures endorsing candidates were distributed via WhatsApp. Several used voice cloning to recreate leaders who had died years earlier. The Election Commission of India issued guidance but had no rapid-removal authority over encrypted messaging platforms.

Pattern

Across documented cases, synthetic political media tends to be released in the 48–72 hours before an election — a window specifically chosen because verification takes time, corrections spread slowly, and there is no opportunity for the targeted campaign to mount a credible response before polls close.

Information Warfare Applications

Beyond elections, deepfakes have been deployed in active conflict. In March 2022, days after Russia's full-scale invasion of Ukraine, a deepfake video circulated showing Ukrainian President Volodymyr Zelensky apparently calling on Ukrainian soldiers to surrender. The video was quickly identified as fake — the facial proportions were off, the head movements were unnatural, and Zelensky immediately appeared live to deny it. But the operation demonstrated that even a poor-quality deepfake can achieve objectives if it reaches audiences before debunking.

Ukrainian officials noted that the fake Zelensky video was distributed via hacked Ukrainian news websites — the synthetic content was credible partly because of the credible delivery channel. This illustrates that deepfakes operate within information ecosystems; the surrounding infrastructure of trust matters as much as the video itself.

Pre-bunkingInoculation against misinformation by explaining manipulation techniques before exposure — shown to reduce susceptibility to deepfakes more effectively than post-hoc debunking.
Election Silence WindowThe period immediately before voting closes, during which many jurisdictions restrict new campaigning — but which bad actors target specifically for synthetic media releases.

Legal and Platform Responses

As of 2024, legal frameworks lag significantly. In the US, no comprehensive federal deepfake law exists. Several states — including California, Texas, and Virginia — have enacted narrow laws targeting non-consensual intimate deepfakes or political deepfakes in the 60 days before an election. But enforcement has been limited and penalties modest.

The European Union's AI Act (adopted 2024) requires that AI-generated content be labelled when it could mislead users about its synthetic nature. Implementation will take years. China has enacted the most prescriptive deepfake regulations globally — the Provisions on the Management of Deep Synthesis Internet Information Services (2022) require watermarking and operator liability — though enforcement reflects domestic political priorities.

Major platforms have adopted content credentials — cryptographic metadata standards (C2PA protocol) that record a media file's provenance and any AI-generation steps. Adobe, Microsoft, Sony, and the BBC are among the signatories. But content credentials only work if they are not stripped, and most platforms do strip metadata on upload.

Critical Gap

The fastest-spreading political deepfakes travel primarily through encrypted messaging apps (WhatsApp, Telegram) and short-video platforms (TikTok) where platform moderation is least effective and metadata is routinely stripped. Legal frameworks designed around broadcast media or open social platforms have minimal purchase in these environments.

Lesson 3 Quiz

Deepfakes in Politics & Elections — Documented cases and legal responses.
What did the FCC rule in response to the New Hampshire AI robocall incident involving a cloned Biden voice?
Correct. The FCC issued a ruling requiring consent and disclosure for AI-generated voices in robocalls — an incremental but significant regulatory step.
The FCC ruled that AI-generated voices in robocalls require explicit consent and disclosure — it did not ban them outright or suspend all robocalls.
In the Slovakia election deepfake incident (2023), what made the audio particularly dangerous from a spread perspective?
Correct. The audio reached tens of thousands of listeners before fact-checkers could assess it, and Meta declined to remove it under election-silence rules.
The danger was its rapid spread on Facebook — tens of thousands heard it before fact-checking caught up, and platform policy decisions compounded the problem.
Why was the fake Zelensky surrender video (March 2022) distributed via hacked Ukrainian news websites?
Correct. Deepfakes operate within information ecosystems — a poor-quality fake on a trusted news site can be more effective than a high-quality fake on an unknown channel.
The goal was credibility by association — a fake arriving via a trusted news source inherits that source's authority, at least initially.
What is "pre-bunking" in the context of deepfake resilience?
Correct. Pre-bunking (inoculation theory applied to media) has research support showing it reduces susceptibility more effectively than correcting beliefs after the fact.
Pre-bunking is about inoculation — teaching people about manipulation techniques before they encounter manipulative content, not responding after the fact.

Lab 3: Election Integrity Analyst

Political deepfakes — rapid response strategy

Scenario

You work for an election integrity organisation. It is 60 hours before polls open and a viral audio clip is circulating that appears to show a candidate making inflammatory remarks. Your AI research assistant has access to the documented case library. Develop your rapid-response strategy.

Starter prompts: "What are the first steps in assessing whether a viral political audio clip is synthetic?" · "How do we communicate uncertainty to the public without amplifying the fake?" · "What did Taiwan's election commission do well in 2024?"
Election Integrity — Research Assistant
Rapid Response
60-hour window is tight. The Slovakia case shows that once synthetic political audio hits tens of thousands of shares, corrections rarely reach the same audience. What's your first priority — verification, platform escalation, or public communication?
Lesson 4 · AI & Media — Module 2

Detection, Provenance & Resistance

The arms race between synthesis and detection — and why technical tools alone cannot solve a social problem.
If detection tools can never be fully reliable, what other layers of defence does a media ecosystem need?

In 2021, a coalition including Adobe, Microsoft, the BBC, Sony, and Intel published the Coalition for Content Provenance and Authenticity (C2PA) specification — a cryptographic standard for embedding tamper-evident metadata into media files at the moment of capture or creation. A C2PA-compliant camera or editing tool records who created the file, what software was used, and whether any AI tools were applied. The metadata travels with the file. By 2024, Leica had shipped the first C2PA-compliant camera, and Nikon and Canon had announced roadmaps. But the system has a critical weakness: every major social media platform strips metadata on upload. Content credentials can only work end-to-end if every link in the chain preserves them.

Automated Detection: State of the Art

Deepfake detection is technically a binary classification problem: given a media file, output a probability that it is synthetic. In practice, it is vastly more difficult. The central challenge is generalisation: detectors trained on known deepfake architectures fail on new ones. When a new synthesis method is released, existing detectors perform near chance until retrained.

As of 2024, best-in-class research detectors (e.g. from the FaceForensics++ benchmark) achieve high accuracy on held-out data from known generation methods but drop significantly on unseen generators. In competitions run by DARPA (the Media Forensics programme, MediFor) and Facebook (the Deepfake Detection Challenge, 2020), winning models achieved around 65–82% accuracy on novel deepfakes — useful but far from reliable enough for high-stakes single-video decisions.

Detection tools that analyse biological signals — subtle patterns in heart rate (rPPG, remote photoplethysmography) or micro-expression timing that synthesis models fail to reproduce — have shown promise in research settings. But they require high-quality, uncompressed video and fail with heavy compression.

rPPG DetectionRemote photoplethysmography — detecting the subtle colour changes in skin caused by heartbeat, which deepfake faces typically fail to reproduce authentically.
C2PACoalition for Content Provenance and Authenticity — an open technical standard for cryptographically recording media provenance, enabling verification of origin and editing history.
Watermarking (AI)Embedding imperceptible signals into AI-generated content at creation time, enabling later identification — used by Google's SynthID and Meta's research systems.

Watermarking and Provenance

An alternative to detection is provenance — instead of trying to identify fakes, ensure all authentic media carries verifiable proof of origin. Two approaches have emerged:

C2PA content credentials (described above) work by attaching signed metadata at capture. Watermarking works by embedding imperceptible signals into the pixels or audio waveform itself — signals that survive moderate compression and editing. Google's SynthID, launched in 2023, embeds invisible watermarks in images generated by Imagen and in audio generated by Google's systems. Meta and OpenAI have published research on robust watermarking for AI-generated text and images.

The fundamental limitation of watermarking is that it only identifies AI-generated content as such — it cannot identify manipulated real content, and watermarks can potentially be removed by adversarial processing (though this is non-trivial for well-designed watermarks).

The Arms Race Problem

Every detection method published in a research paper gives deepfake developers a training target. The history of GAN development shows that discriminators are effectively free optimisation signals for generators — publication of a detector is simultaneously the publication of a roadmap for defeating it.

Media Literacy and Institutional Resilience

Researchers increasingly argue that technical solutions cannot solve what is fundamentally a social and epistemic problem. Media literacy — the capacity to critically evaluate media sources, apply verification heuristics, and tolerate uncertainty — is a necessary complement to technical tools.

Studies from the Reuters Institute and MIT Media Lab suggest that source credibility heuristics — asking "who published this, and why?" before asking "is this video real?" — are more reliable guides to accuracy than automated detection tools in most everyday contexts.

Institutional resilience includes: newsrooms with established deepfake verification workflows; platforms with rapid-escalation channels for suspected political synthetic media; election authorities with public communications plans for the scenario in which a deepfake goes viral 48 hours before polls open; and legal frameworks that assign liability clearly enough to create deterrence.

The SIFT method (Stop, Investigate the source, Find better coverage, Trace claims) developed by researcher Mike Caulfield has been adopted in media literacy curricula across the US and Europe as a practical framework applicable before any technical analysis.

Synthesis

The most robust defence against synthetic media is not a single tool but a layered system: cryptographic provenance at capture, platform-level watermark verification, automated detection as a triage signal (not a verdict), trained human reviewers for high-stakes cases, and populations with sufficient media literacy to apply scepticism proportionally — to fakes and to real content being falsely denied.

Lesson 4 Quiz

Detection, Provenance & Resistance — Technical and social defences.
What is the "generalisation problem" in deepfake detection?
Correct. Every new generation architecture can defeat existing detectors until those detectors are retrained — a perpetual asymmetry favouring attackers.
The generalisation problem is that detectors perform well on known fakes but poorly on fakes made with new methods — creating a persistent cat-and-mouse dynamic.
What is a critical weakness of the C2PA content credentials system as currently deployed?
Correct. Content credentials only function end-to-end if every platform in the distribution chain preserves the metadata — currently, most do not.
The key weakness is metadata stripping at upload — major social platforms remove C2PA data, making provenance verification impossible for most shared content.
What does rPPG detection measure in deepfake analysis?
Correct. Remote photoplethysmography detects the tiny colour fluctuations driven by blood flow — biological signals that current synthesis models typically do not replicate.
rPPG (remote photoplethysmography) detects heartbeat-driven colour changes in skin — a biological signal that deepfakes currently fail to reproduce authentically.
What does the SIFT method stand for, as used in media literacy education?
Correct. SIFT was developed by Mike Caulfield and has been adopted in media literacy curricula across the US and Europe as a practical verification framework.
SIFT stands for Stop, Investigate the source, Find better coverage, Trace claims — a practical media literacy framework applicable before any technical analysis.

Lab 4: The Provenance Chain

Detection limits and layered defence — designing a resilient media system

Scenario

You are advising a major news organisation on its synthetic media policy. They want to know: when a viral video arrives at their verification desk, what is the full decision tree? Your AI assistant can walk through detection tools, provenance checks, source heuristics, and the SIFT method. Design a verification workflow that is practical under newsroom time pressure.

Starter prompts: "What should a newsroom check in the first 10 minutes after a suspicious video is flagged?" · "When should automated detection tools override human judgment, and when should they not?" · "What are the limits of watermarking as a defence?"
Newsroom — Verification Policy Advisor
Media Forensics
A newsroom workflow needs to be fast enough to be competitive and rigorous enough to avoid amplifying synthetic content. There's real tension there. Should we start with the first-10-minutes triage, or the broader policy question of when to publish versus wait?

Module 2 Test

Deepfakes & Synthetic Media — 15 questions · Pass at 80%
1. Who coined the term "deepfake" and when?
Correct. The Reddit username "deepfakes" appeared in late 2017 posting celebrity face-swap videos.
The term originated with an anonymous Reddit user in 2017, not from academia or Hollywood.
2. In a GAN, which component learns to distinguish real images from synthetic ones?
Correct. The discriminator's feedback signal drives the generator to produce increasingly convincing outputs.
The discriminator is the "judge" — its inability to detect fakes is the measure of the generator's success.
3. What is the "liar's dividend"?
Correct. The existence of deepfakes gives bad actors a plausible denial mechanism for real footage of their wrongdoing.
The liar's dividend is the epistemic benefit to bad actors of a world where everything could be fake — real evidence can be dismissed.
4. According to Sensity AI, what proportion of deepfake videos online depicted non-consensual intimate imagery?
Correct. Image-based abuse dominates the deepfake ecosystem — political deepfakes receive outsized media coverage relative to their actual volume.
Over 95% — the deepfake crisis is primarily a gender-based harm crisis, not a political one, despite media framing.
5. In the Hong Kong deepfake fraud (2024), what was the approximate amount transferred?
Correct. HK$200 million ≈ US$25 million — the largest known deepfake financial fraud at the time.
The transfer was HK$200 million, approximately US$25 million.
6. What is "temporal consistency" in deepfake synthesis?
Correct. Frame-to-frame stability is expensive to achieve and is the primary quality bottleneck in real-time deepfake systems.
Temporal consistency is about frame stability — preventing the synthesised face from flickering, which is the most visible artefact in lower-quality deepfakes.
7. What did the Biden robocall incident in New Hampshire (January 2024) prompt the FCC to rule?
Correct. The FCC issued a targeted ruling requiring consent and disclosure — one of the first US federal regulatory responses to AI voice cloning in political communications.
The FCC ruled that AI-generated voices in robocalls require explicit consent and disclosure.
8. The fake Zelensky surrender video (March 2022) was distributed via which channel to increase credibility?
Correct. The trusted channel provided credibility the synthetic content itself lacked — illustrating that deepfakes operate within broader information ecosystems.
Hacked Ukrainian news sites — the delivery channel's trust transferred to the fake content, amplifying its initial impact.
9. What is the "generalisation problem" in automated deepfake detection?
Correct. Every new synthesis method can defeat existing detectors — creating a perpetual asymmetry that favours those creating fakes.
The generalisation problem: detectors that excel on known deepfake architectures often perform near-chance on novel synthesis methods.
10. What is rPPG in the context of deepfake detection?
Correct. rPPG detects biological signals — subtle colour fluctuations from blood flow — that current deepfake synthesis does not authentically replicate.
rPPG is remote photoplethysmography — it measures biological heartbeat signals in facial video that synthesis models typically fail to reproduce.
11. What critical weakness does C2PA content credentials have in current deployment?
Correct. Without end-to-end metadata preservation across every platform in the distribution chain, C2PA provenance is broken at the point of social media upload.
Most social platforms strip metadata on upload — making C2PA provenance verification impossible for the vast majority of shared content.
12. What is Google's SynthID designed to do?
Correct. SynthID embeds invisible watermarks that survive moderate compression and editing, enabling later identification of AI-generated content.
SynthID is Google's imperceptible watermarking system for AI-generated images and audio — embedded at creation to enable later identification.
13. In the Slovakia election deepfake case (2023), what platform carried the audio clip to tens of thousands before fact-checking?
Correct. The audio spread on Facebook, and Meta's invocation of election-silence rules — rather than removing the content — became a case study in platform policy failure.
The audio spread on Facebook. Meta's policy decisions in the 48 hours before the election became a significant case study in platform accountability.
14. What does the SIFT media literacy method recommend as its first step?
Correct. SIFT begins with Stop — a metacognitive pause that interrupts the automatic sharing impulse before any verification step is taken.
The S in SIFT is Stop — interrupting the impulse to immediately share or react, creating space for the subsequent verification steps.
15. Which of the following best describes a "layered defence" against synthetic media, as recommended by researchers?
Correct. No single tool is sufficient — resilience requires combining technical, institutional, and educational layers, each covering the others' blind spots.
A layered approach combines multiple systems: provenance standards, watermarking, AI detection, human reviewers, and population media literacy — no single layer is sufficient.