AI/ML, AI benefits/risks, Identity, IAM Technologies

 INTERPOL launches deepfake scam awareness campaign as AI impersonation risks grow 

Warning concept of deepfake technology and facial recognition misuse, showing cybersecurity threats, digital identity fraud, and the risks of AI-generated face manipulation in online systems

COMMENTARY: As AI makes faces, voices and messages easier to reproduce, CISOs need to rethink how their organizations prove identity before people and systems are allowed to act. 

INTERPOL has launched a public awareness campaign highlighting the growing risk of scams that use AI-generated or AI-altered media. The campaign includes three short videos, themselves created with AI, showing how synthetic media, commonly known as deepfakes, can be used to deceive people. The warning reflects how quickly the barrier to convincing impersonation is falling. Generative AI tools can now create hyper-realistic avatars and reproduce the face or voice of a real person using only a small amount of material found online, without requiring significant technical skill or investment. 

[SC Media Perspectives columns are written by a trusted community of SC Media cybersecurity subject matter experts. Read more Perspectives here.]

What is changing is not only the quality of the fake. It is our ability to trust the identity presented through it. 

For most of human history, identity was closely tied to physical presence. You saw a person, heard their voice and understood the context around the interaction. The telephone removed the face, so we placed more trust in the voice. Email removed both face and voice, shifting trust toward addresses, language and established relationships. Video brought some of those physical signals back and, with them, a greater sense that the person on the other side could be believed. 


Related reading:


Generative AI now interferes with that progression because it can reproduce the same signals we have gradually learned to use as evidence of identity. This is where a deepfake becomes more than manipulated media and becomes an identity problem. 

INTERPOL describes part of that change directly when it warns that “AI is erasing the tell-tale signs of a scam.” The spelling mistakes, poor formatting and awkward language that once gave fraudulent communications away can increasingly be corrected by AI. Synthetic media adds another layer because the scam can now arrive carrying a face or voice the victim already associates with trust. 

For CISOs, the response cannot therefore depend on asking people to become better judges of whether a face “looks fake.” When the visible identity can be manufactured, security teams need another source of evidence. 

One way to approach that problem is to look beyond the identity being presented and examine the identity of the media itself. 

A genuine image or video carries its own history. Its facial structure, pixels, textures, frequency patterns, compression and editing history contain information about how that media came into existence. Synthetic media also leaves traces. The question for security teams is whether those different traces tell the same story. 

That creates a more useful model for deepfake verification: do not trust a single clue. Test the media across seven independent characteristics before accepting the identity it presents. 

Seven tests for the identity behind the media

1. Check the biometric structure. 

Start with facial geometry and three-dimensional consistency. A deepfake may look convincing at normal viewing distance while still carrying distortions in shape, alignment or facial structure. The goal is not to ask whether the person looks familiar, but whether the face behaves like a real biometric identity. 

2. Look for an AI signature. 

GANs, diffusion models and other generative systems can leave statistical fingerprints in the media they create. These traces may not be visible to a person, but they can help security teams determine whether the apparent identity was captured from reality or generated by a model. 

3. Read the file’s digital history. 

Saving, editing, exporting and compressing media leave traces. Inconsistent compression layers or unusual editing patterns can reveal that the file has been manipulated. Security teams should examine whether the media’s history supports the identity and origin it claims to represent. 

4. Inspect the frequency domain. 

AI-generated media can contain patterns that sit beyond normal visual inspection. Frequency analysis gives security teams another view of the file and can reveal inconsistencies that a convincing face or voice may hide. 

5. Test texture and surface detail. 

Skin, hair and other surfaces can reveal unnatural smoothing, repetition or inconsistency. These small details matter because the visual identity may appear correct while the texture underneath shows signs of synthetic construction. 

6. Test the media after degradation. 

Real-world content is compressed, re-uploaded, resized and screenshotted. Deepfake detection should remain useful after those changes. A control that only works on pristine media may fail when the same content reaches an organization through email, messaging or social platforms. 

7. Inspect pixel-level boundaries. 

Fine pixel relationships can expose blending seams where a generated or swapped face meets the surrounding frame. These boundaries can show where one identity has effectively been placed over another. 

Deepfakes will keep improving, so no detection tool will be the final word. The real question for security leaders is whether their organization can prove who it is dealing with before acting on a request. Start building that proof, one verified identity at a time, before the next wave arrives. 

An In-Depth Guide to AI

Get essential knowledge and practical strategies to use AI to better your security program.
Frayyam Asif

Frayyam Asif is the Chief Technology Officer at Shufti, where he leads the engineering and IT teams building the company’s AI-powered identity verification and compliance platform. Before his CTO role, Frayyam held positions at Shufti in earlier capacities and previously served as Chief Development Officer at Programmers Force. He also worked at EMC Solutions and The Punjab Information Technology Board.

Get daily email updates

SC Media's daily must-read of the most current and pressing daily news

By clicking the Subscribe button below, you agree to SC Media Terms of Use and Privacy Policy.

You can skip this ad in 5 seconds