Synthetic Identity and Deepfake Defence

Detection drills and onboarding hardening

This is the first of the platform's hands-on modules. It explains how synthetic identities, forged documents and deepfake audio and video are produced, and then works through the detection of them: presentation and liveness attacks, the artefacts left in synthetic media, and the hardening of a know-your-customer and onboarding workflow against both attack classes. By the end you will have a repeatable detection routine that can be applied to any verification decision.

Create an account Back to curriculum

By the end of this module, you can

Explain how GANs, diffusion models and voice cloning produce synthetic identities and media.
Apply a structured routine to flag likely AI-synthesised identity documents and deepfake media in an onboarding or verification context.
Specify controls that harden a KYC and onboarding process against presentation and injection attacks, within South African identity-verification realities.

Skills taught

Synthetic-media & document-forgery detection
Presentation-attack awareness
Onboarding-control design
Judging when to escalate rather than trust a biometric result

Lessons in this module

How synthetic media is made: GANs, diffusion and voice cloning

Explain, at a working conceptual level, the three generation mechanisms behind synthetic faces, media and voices, and work out from each mechanism what a detector can and cannot rely on.

25 min

How synthetic identities are assembled, and where they stay weak

Describe the assembly of a synthetic identity and its supporting document set, and identify the structural weak points a verification process can target.

20 min

Detecting synthetic media: artefact classes and their limits

Recognise the principal artefact classes in synthetic images, video, audio and generated documents, and place artefact-spotting correctly as one probabilistic layer in a multi-layer routine.

25 min

Presentation attacks, injection attacks and what liveness establishes

Distinguish presentation from injection attacks on biometric verification, explain what liveness detection does and does not establish, and specify the capture-integrity controls that address the injection gap.

20 min

The detection routine: five steps, every time

Apply Perspica's five-step verification routine (Context, Channel, Artefact, Corroboration, Escalation) to any onboarding or verification decision, and produce the calibrated record it requires.

20 min

Hardening the onboarding workflow: controls, evidence, and the regulator's question

Specify a layered set of onboarding controls against presentation and injection attacks, governed override paths included, and express the design as RMCP-ready, regulator-facing documentation.

20 min

Case 03.2

Case file: The applicant who cleared every check

An applied fact pattern worked against a model resolution, followed by the timed assessment (35 minutes, pass mark 70 percent).

25 min

How it lands across the four desks

Fraud & AML

This module gives your desk a repeatable routine for flagging AI-synthesised identity documents and deepfake media at onboarding, and for deciding when to escalate rather than rely on a biometric check on its own.

Compliance & Risk

You learn what a defensible verification control looks like now, and how to evidence it to a regulator. This is the control-design and documentation side of the module.

Wealth & Advisory

Onboarding-facing advisers apply the same detection routine at client take-on, protecting both the client relationship and the firm from a synthetic-identity account opening.

Credit & Underwriting

Underwriters meet the same synthetic-document risk at application stage. The routine in this module is designed to stop a fabricated identity before it reaches a credit decision.

Key literature · 8 sources

Every module rests on a verified scholarly and institutional evidence base. The full core and further reading lists open with the module.

Tolosana, R. et al. (2020) 'DeepFakes and Beyond: A Survey of Face Manipulation and Fake Detection.' Information Fusion 64: the field survey of face-manipulation families and detection.
Verdoliva, L. (2020) 'Media Forensics and DeepFakes: An Overview.' IEEE Journal of Selected Topics in Signal Processing 14(5): the forensic-detection frontier and its limits.
Mirsky, Y. & Lee, W. (2021) 'The Creation and Detection of Deepfakes: A Survey.' ACM Computing Surveys 54(1): creation-and-detection taxonomy underpinning Lessons 1 and 3.
Ramachandra, R. & Busch, C. (2017) 'Presentation Attack Detection Methods for Face Recognition Systems.' ACM Computing Surveys 50(1): the PAD literature behind Lesson 4.
Goodfellow, I. J. et al. (2014) 'Generative Adversarial Nets.' NIPS 2014: the adversarial training mechanism and its strategic consequence for detection.
U.S. Federal Reserve (2019) 'Synthetic Identity Fraud' payments-fraud insights: the definitional treatment carried from Module 2 into Lesson 2's assembly analysis.

Create an account