Synthetic Identity Fraud Is Becoming an Identity-Bypass Chain

Synthetic identity fraud used to mean a manufactured identity built from a mix of real and false personal data (a real national ID number, a fabricated name) used to open an account that had no history and therefore no red flags. That definition is now too small. In modern attacks the synthetic identity can include a cloned voice, a generated face, a stolen photo, a breached email address, a broker record, a virtual camera feed and a live phishing session. The goal is no longer only to open a fake account. It is to pass whatever identity check sits between the attacker and trust.

This article is the advanced layer. For the narrower question of how leaked data defeats ordinary identity checks, read our guide to how leaked data is used to bypass KYC checks. Here the focus is the wider chain: synthetic identity fraud assembled from breach data, AI voice, virtual-camera injection, real-time session interception, and the exposed identity material that makes the whole sequence believable.

Understanding how each layer is defeated requires understanding where the source data comes from. Every component is assembled from publicly available or previously leaked information. The technique is the application. The data is the fuel. For a wider view of how credential leaks and breach data are weaponised, see our Credential Leaks & Breach Response hub.

Synthetic Identity Fraud Is No Longer Just a Fake Person

The classic definition comes from lending. A fraudster pairs a real, unused identifier with fabricated details, nurtures the profile until it looks creditworthy, then borrows and disappears. The Federal Reserve has estimated this pattern costs US lenders billions and accounts for a meaningful share of unsecured charge-offs. That version of the problem has not gone away — US lenders faced an estimated $3.3 billion in synthetic-identity exposure in the first half of 2025 alone, and the Deloitte Center for Financial Services projects at least $23 billion in losses by 2030.

What has changed is the material. A synthetic identity is now assembled from the same components used to defeat live verification: a breached email address and password, a broker record confirming an address, a photo scraped from LinkedIn, a voice cloned from a podcast, a document number lifted from a telecom breach. The account-opening fraud and the verification-bypass attack draw on the same inventory. That is why it is more useful to treat synthetic identity fraud as an identity-bypass chain than as a single fake persona: each stage is satisfied by a different assembled component, and the components are cheap and already circulating.

The Raw Material: Breach Data, Broker Records and OSINT

Before any synthetic voice is generated, before any face model is built, before any phishing panel is deployed, an attacker sources the components. This sourcing phase is pure OSINT — no system access required, no tools beyond a browser. These are the identity packs assembled from ordinary breach data, enriched with public detail until they resemble a real person.

Voice samples are sourced from LinkedIn video introductions, company YouTube channels, earnings call recordings published on investor relations pages, conference talk recordings, podcast appearances, and media interviews. A 3–5 second clip is sufficient for modern voice synthesis tools to clone with 85% accuracy. Executives, sales leaders, and anyone with a public-facing role have typically published far more than that. A single earnings call contains hours of clean audio.

Facial data is sourced from LinkedIn profile photos, Instagram, Facebook, company team pages, and press photography. At scale, it is sourced from breach databases: the Odido breach released passport and ID-document numbers for millions of Dutch customers, and the IDMerit leak exposed approximately one billion KYC verification records including facial imagery from 26 countries. These are not hypothetical sources — they are actively indexed and traded on dark web forums. iProov identified 31 new crews selling packaged KYC bypass toolkits in 2024 alone, and found a dedicated dark web operation farming identity data specifically for verification spoofing.

Organisational context for vishing pretexts is sourced from LinkedIn (job titles, team structures, who reports to whom), company websites (executive biographies, office locations, technology partnerships), press releases (recent system migrations, new software deployments, organisational changes), and job postings (which reveal the internal technology stack in detail). A vishing caller who opens with a reference to the company's recent Okta migration or mentions a colleague by name is not guessing — they read the job posting and the LinkedIn profile.

Layer 1: The Voice

AI voice synthesis has crossed a threshold that makes voice-based authentication and phone-based identity confirmation unreliable as standalone controls. Modern cloning tools require 3–5 seconds of audio, achieving 85% voice accuracy. The output replicates not just pitch and tone but cadence, hesitation patterns, filler words, and accent — the subtle signals human listeners use to judge authenticity. This is the point where a vishing call collapses the identity layer: the voice itself becomes the credential, and the credential is forgeable.

The documented impact is not theoretical. In early 2024, a UK energy firm lost €220,000 after an employee received instructions from what sounded exactly like the company's CEO. In early 2025, a European energy conglomerate lost $25 million when a cloned CFO voice issued live wire transfer instructions — pauses, tone, and context all matching. In February 2025, the group UNC6040 cloned a CFO voice to infiltrate a Canadian insurance company, resulting in $12 million in unauthorised transfers and exfiltration of sensitive financial data.

At the corporate vishing level used by ShinyHunters and Scattered Spider, voice cloning is combined with caller ID spoofing and a pre-researched pretext. The caller knows the target's name, their manager's name, the name of the IT ticketing system, and which SSO platform is in use. That context is assembled entirely from open sources before any call is placed. Deepfake-enabled vishing surged 1,600% in Q1 2025 compared to the end of 2024.

Layer 2: The Session

Multi-factor authentication was designed under an assumption that credentials and authentication factors would be captured in isolation — at different times, via different channels. Adversary-in-the-middle (AiTM) phishing invalidates that assumption by sitting between the target and the legitimate service during live authentication. Where it succeeds, the next stop is account takeover after identity verification has already passed.

Tycoon 2FA was the most operationally significant example. The mechanism: a phishing link redirects the target to a proxy server that forwards all traffic to the real Microsoft or Google login page. The user authenticates — entering their password and approving their MFA prompt — entirely successfully. The proxy captures the session cookie that the authentication service issues after successful MFA completion. The attacker then uses that cookie to access the account with no further authentication required. MFA was not bypassed — it completed correctly. It was simply irrelevant to the outcome. By mid-2025, Tycoon 2FA had reached over 500,000 organisations and accounted for 60% of all Microsoft-blocked phishing attempts.

On 4 March 2026, a Europol-coordinated operation seized 330 domains behind Tycoon 2FA, a platform tied to more than 64,000 attacks. The takedown removed one platform, not the technique — Tycoon infrastructure began recovering within the same day, and a crowded successor market was already running in parallel: Mamba 2FA, Salty2FA, Sneaky 2FA, Whisper 2FA, the open-source Evilginx, and newer commodity kits such as Starkiller. AiTM is now a subscription product, not a specialist capability, and the infrastructure for running it is rebuilt faster than it can be dismantled.

Layer 3: The Face

KYC liveness detection was designed to distinguish a live person from a static photograph. The requirement that a user blink, turn their head, or track a moving target was intended to make spoofing with printed images impractical. It solved that problem. It did not anticipate the video injection attack, which does not attempt to fool liveness detection with a photograph — it replaces the camera feed entirely before the KYC system receives it.

On Android, the camera API is a software interface. An application requesting camera access receives a video stream — but the operating system does not inherently guarantee that stream originates from the physical camera sensor. A virtual camera application installs itself as a camera driver and intercepts API calls before they reach hardware. The KYC application requests the camera; it receives the virtual camera's feed instead. iProov's Red Team demonstrated this using two free tools: Faceswap, an open-source desktop application that applies generative AI to produce real-time face-swapped video, and Virtual Camera: Live Assist, a free Android app that replaces the device camera feed with an incoming video stream. The face model is generated from photographs sourced during the OSINT phase. The result: the KYC system receives a video of a face blinking, turning, and tracking on command. All passive and active liveness checks pass. The attacker is not in the room.

The volume is climbing sharply. Having recorded a 783% increase in injection attacks in 2024, iProov's 2026 Threat Intelligence Report found injection attacks against iOS rose 741% across 2025, including a 1,151% surge in the second half of the year — the technique spreading from Android to what had been the harder target. The cost of the toolchain is effectively zero: the tools are free, the documentation is public, and the attack requires no programming knowledge. Face swaps are now the deepfake of choice. The defensive countermeasures — where they exist — are covered in our guide to practical deepfake detection and verification protocols.

Layer 4: The KYC Record

The three layers above defeat verification in real time. The fourth removes the need to. A completed KYC record — a verified name, document image, selfie, and approval log — is itself a reusable artifact, and these records now leak and trade at scale. The IDMerit exposure alone put roughly one billion KYC verification records into circulation. Once a verified record exists, an attacker does not have to pass the check; they can present the artifact of a check that already passed, or use its fields to seed a new application elsewhere.

This is where the account-opening definition of synthetic identity fraud meets the verification-bypass definition. A document template bought for under $20, paired with a real breached identity and a face model, produces an application that clears remote onboarding. The resulting account has no fraud history precisely because the identity is new to the institution — and it is then used for money laundering, fraudulent transfers, or simply sold as a pre-verified account. The KYC record is not the end of the verification process. In this market it is a raw material for the next one.

Why This Defeats Fragmented Verification

No single control here was badly designed. Voice authentication, MFA, liveness detection and document checks each stop the crude version of their own attack. The failure is structural: verification is fragmented across channels — a phone line, a login page, a camera, a document scanner — and each channel is validated in isolation. The attacker does not defeat any one control at its strongest. They assemble one satisfying component per channel, sourced from the same leaked inventory, and present each in the place it is expected.

A cloned voice satisfies the phone. A proxied session satisfies the login. An injected feed satisfies the camera. A leaked record satisfies the document. Individually, each is plausible. Together they compose an identity that no single checkpoint was designed to see whole. This is why adding another isolated factor rarely helps: the weakness is not the strength of each check but the absence of anything that evaluates the identity across all of them at once.

What Detection Actually Requires

Each layer has a genuine countermeasure. None of them are defaults.

Against voice cloning: designated call-back protocols to verified numbers (not numbers provided during the call), live code-word systems rotated per session, and policies that prohibit credential changes over inbound phone calls regardless of caller identity. Voice alone cannot be trusted as an authentication factor.

Against AiTM: FIDO2 hardware security keys are the only control that structurally defeats session-cookie theft. Unlike TOTP codes, a FIDO2 key cryptographically binds authentication to the specific domain being authenticated — a proxy relaying traffic from a phishing domain cannot obtain a valid FIDO2 response for the real domain. Google reported zero successful phishing attacks against its 85,000+ employees after FIDO2 deployment. The caveat is the enrolment flow, not the key: in July 2025 the PoisonSeed campaign showed that a vishing pretext can persuade an IT administrator to remove a hardware key and re-enrol a weaker TOTP factor, degrading the account to a level AiTM can defeat. The key is only as strong as the policy governing when it can be removed.

Against KYC injection: hardware attestation (verifying the video stream originates from a real camera sensor), depth-sensor liveness using IR mapping (a flat video feed cannot produce a 3D facial depth map), and emulator detection through sensor fingerprinting. The standards bodies now formalise this: NIST's SP 800-63-4, finalised in 2025, adds explicit controls against injection attacks that deliver deepfakes and presses providers to offer alternatives to face biometrics rather than treat a selfie as proof. These controls exist. They are not universally implemented.

What This Means for Executives and Organisations

The practical consequence is most visible in remote account opening. bunq, operating under a Dutch banking licence, opens accounts in five minutes via its mobile app; Revolut uses the same core model — document scan plus selfie liveness check, fully remote. These processes removed friction by reducing the number of identity signals to one document and one face, which is exactly the surface the injection chain maps onto. Synthetic identity fraud cost US lenders billions and is growing at over 20% annually, and the same techniques now target the enterprise directly rather than only the bank.

The forecast has become the present. A September 2025 Gartner study found that 37% of cybersecurity leaders had encountered deepfake incidents during video calls, and the Ponemon Institute reports that 41% of organisations have experienced deepfake attacks targeting executives. The exposure is not evenly distributed: executives, entrepreneurs, and high-visibility professionals publish more voice, more imagery, and more organisational context, so the identity that can be assembled about them is more complete. That is also the corporate risk that infrastructure-focused tools miss — the human identity layer sits outside the assets they map, as we cover in the identity attack surface infrastructure tools miss.

How Exposure Reduction Breaks the Chain

Every layer of this chain runs on identity data that has already leaked or been published. The voice, the face, the document number, the organisational context — none of it is generated from nothing. That is the one point of control a person or organisation still holds: not the strength of any single verification control, which they do not operate, but how much of their identity material is sitting in circulation waiting to be assembled. The same broker and breach data feeds executive impersonation and CEO fraud, which is why CEO fraud starts with broker and breach data long before the fraudulent instruction is sent.

This is where the response calibrates by reader. An individual or executive starts with a Mirror investigation to map what is findable before it is used; a high-risk individual reduces the exposure behind impersonation with The Lockdown; an organisation reviews the identity exposure around its key people through a Corporate Audit; and where the raw material is broker records specifically, The Eraser removes the listings that support pretexting.

Sources

1. DeepStrike — Vishing Statistics 2025: AI Deepfakes & the Voice Scam Surge
2. American Bar Association — The Rise of the AI-Cloned Voice Scam
3. ScamWatchHQ — The $200 Million Deepfake Disaster (2025)
4. Microsoft Security Blog — Inside Tycoon2FA
5. CrowdStrike — Tycoon2FA Persists Following Takedown
6. Biometric Update — Face-Swapping App Fools Liveness Detection
7. iProov — Threat Intelligence Report 2026 (iOS injection attacks +741%)
8. iProov — Dark Web Identity Farming Operation Discovery
9. NIST — SP 800-63-4 Digital Identity Guidelines (2025)
10. Deloitte Center for Financial Services — Synthetic Identity Fraud ($23B by 2030)
11. Sardine — Neobank Fraud & Synthetic Identity Fraud
12. Zyphe — Beating KYC With $20 Deepfakes
13. The Hacker News — PoisonSeed: Social Engineering Against FIDO Keys
14. Biometric Update — Social Engineering Against FIDO Keys
15. bunq Help Centre — Open an Account in 5 Minutes

Frequently Asked Questions

What is synthetic identity fraud?
Synthetic identity fraud is the use of an identity assembled from a mix of real and fabricated data. Traditionally that meant pairing a real identifier with a fake name to open a creditworthy-looking account. In current attacks the same idea extends to verification bypass: attackers combine breached records, a cloned voice, a generated face and a stolen session to satisfy each identity check between them and access.

How is synthetic identity fraud different from identity theft?
Identity theft impersonates one real person using their real details. A synthetic identity is a composite — real fragments (an address, a document number, a photograph) stitched together with fabricated or generated elements, so it does not map cleanly to a single victim. That is what lets it pass checks designed to catch obvious mismatches, and what makes it harder to detect after the fact.

Can AI voice cloning and deepfakes really pass identity checks?
Yes. Three to five seconds of audio can clone a voice with about 85% accuracy, and free tools can inject a real-time face-swap into a KYC camera feed so that liveness checks pass. Injection attacks are rising sharply year on year. The stronger countermeasures — hardware attestation, depth-sensor liveness, FIDO2 keys, call-back verification — exist but are not universally deployed.

How do you protect against synthetic identity fraud?
Institutions need controls that evaluate an identity across channels rather than one check at a time. For an individual or organisation, the controllable variable is exposure: the voice, imagery, document numbers and organisational detail that make a synthetic identity believable are sourced from leaked and published data. Reducing that circulating material — through a Mirror exposure audit or broker removal — shrinks the raw material the chain depends on.