Synthetic Identity Fraud: How It Works, Red Flags, and How to Detect It
Synthetic identity fraud is a crime in which offenders combine real and invented details, often a valid Social Security number with a made-up name and date of birth, to create a person who does not exist, build credit and then draw it all down. A Federal Reserve paper cites an estimate that it cost US lenders $6 billion in 2016.
What is synthetic identity fraud?
Synthetic identity fraud is a crime in which perpetrators combine fictitious and sometimes real information, such as Social Security numbers (SSNs) and names, to create a new identity and use it to defraud lenders, government agencies or individuals. It gets past onboarding because the person on the application doesn't exist, so the checks built to catch a stolen identity have no real victim, no fraud alert and no mismatch to find.
Synthetic identity fraud is a form of identity fraud, the wider term for using someone's identity details, real or invented, to obtain money, credit or services. Classic identity theft takes over a real person's existing account or opens a new one in their name. A synthetic identity is a new person stitched together from parts, often a valid SSN that nobody is actively using paired with a made-up name and date of birth. FinCEN uses the same idea when its November 2024 deepfake alert defines a synthetic identity as a combination of real and fake personal information used to fabricate a person or entity.
The cost lands on lenders. The Federal Reserve's 2019 paper cites an Auriemma Group estimate that US lenders lost $6 billion to synthetic identity fraud in 2016, with an average charge-off of more than $15,000 per case and up to 20 percent of all credit losses that year. The Fed's later synthetic identity fraud toolkit puts US financial institution losses at an estimated $20 billion in 2020 and notes that the losses are often booked as ordinary credit losses. Both figures are dated estimates; nobody publishes a reliable current total, partly because so many cases are never labelled as fraud.
This differs from first-party fraud and third-party fraud in a way that matters operationally. There's no victim to call and no police report, and when the identity stops paying it simply goes silent. That's what makes it so effective against standard fraud controls, and so costly when it appears at scale in a portfolio.
How does synthetic identity fraud get past onboarding?
A synthetic identity gets past onboarding by looking like a genuine thin-file customer and by building a real credit history before it asks for real money. The lifecycle runs in three phases: creation, cultivation and bust-out.
Creation. The fraudster picks an SSN, often one belonging to a child or someone else with little or no credit activity, and pairs it with a fabricated name, date of birth, a mail-drop address and a disposable phone number. Some now add a generated face on a forged ID document, which FinCEN has seen used to open accounts.
Cultivation. The first credit application is often declined, but the inquiry can be enough for a bureau file to start. The identity then holds a small secured or retail card and pays on time. The Fed paper describes piggybacking, where the synthetic is added as an authorized user on an account with good credit, and cites an ID Analytics estimate that nearly 50 percent of synthetic identities use it to build credit.
Bust-out. Once limits are high enough, every line is drawn down at once on high-value, liquid goods and services such as electronic devices, gold or cash advances, and the identity stops paying. The accounts charge off as credit losses.
One of the largest cases ever charged shows the same three steps at scale. In February 2013, the US Attorney for New Jersey charged a ring that allegedly created more than 7,000 false identities, obtained tens of thousands of credit cards and caused more than $200 million in confirmed losses. The complaint described a "make up, pump up, run up" process, more than 1,800 drop addresses and sham companies that reported false credit history to the bureaus.
Onboarding checks pass these identities because each one answers a narrower question than it seems. A bureau check asks whether a credit file exists, and a cultivated synthetic has one. A document check asks whether the ID looks genuine, and a generated image can. An SSN format check asks whether the number is valid, and the SSN often is valid; it just belongs to someone else.
This pattern frequently intersects with bust-out fraud when the same ring operates dozens of synthetic identities in parallel, and the proceeds often move through money mule networks.
What are the types of identity fraud?
The main types of identity fraud are misuse of an existing account, opening a new account with someone's details, and misuse of personal details for other purposes, with synthetic identity fraud as the variant where the identity itself is invented. That three-way split is how the US Bureau of Justice Statistics measures identity theft in its national survey.
- Existing account misuse: unauthorized use of a credit card, debit card, bank, email, social media or other account the victim already has.
- New account fraud: using someone's personal information to open a new account in their name.
- Other misuse: using personal information for other fraudulent purposes, such as filing a tax return, getting medical treatment, applying for a job or claiming government benefits.
- Synthetic identity fraud: a fabricated person built from real and fake data, so there's no single victim whose account or name was taken.
Lists of the "4 types of identity theft" or "5 types of identity theft" usually come from splitting that last category into tax, medical, employment and benefits fraud, or from adding synthetic and child identity theft. There's no official count. For scale, BJS estimated that 23.9 million US residents age 16 or older, about 9 percent, were identity theft victims in the 12 months before its 2021 survey, most often through misuse of an existing credit card or bank account. Fewer than 1 percent reported a new account opened in their name, which is the type that onboarding controls are meant to stop.
What does a deepfake mean for KYC?
A deepfake is AI-generated or manipulated image, audio or video content that resembles real people, objects, places or events and would falsely appear to be authentic. That's the legal definition in Article 3(60) of the EU AI Act. For KYC, a deepfake means the face on the ID document, the selfie or the video call may have been generated rather than captured.
NIST's July 2025 identity proofing guidelines (SP 800-63A-4) describe the attack precisely. Fraudsters pair generative AI tools with injection attacks, inserting forged images or video between the capture device and the system doing the comparison, to defeat automated document checks, biometric matching and visual review by staff. NIST says that live document capture and presentation attack detection make this harder, and that even these mechanisms aren't sufficient for every case. Its guidelines require presentation attack detection whenever biometrics are collected and compared remotely.
Fake KYC is the wider problem: an account opened on documents or data that don't belong to a real, present applicant, whether the documents were edited, borrowed, bought or generated. Deepfake KYC is the newest form of it. The questions about deepfake fraud in banking beyond onboarding, such as voice cloning and payment scams, are covered in that article.
Why do randomized SSNs and credit privacy numbers matter?
The Federal Reserve's 2019 paper Synthetic Identity Fraud in the U.S. Payment System explains why the SSN geography check catches less every year. The SSA began assigning SSNs at random on June 25, 2011, which removed the geographic meaning of the first three digits. Geographic checks no longer work for newly issued numbers. The paper cites an ID Analytics estimate that nearly 40 percent of synthetic identities use a randomized SSN, and another that 85 to 95 percent of applicants identified as synthetic identities were not flagged as high risk by traditional fraud models.
The same paper describes credit privacy numbers (CPNs). These are nine-digit numbers that some non-reputable credit repair agencies sell to people who want to re-establish credit, and fraudsters use them in place of an SSN to apply for credit and set up new credit profiles. In reality a CPN may be a valid SSN that the SSA has not yet issued, and law enforcement officials consider using one to obtain credit illegal, with or without criminal intent. On an application form a CPN looks exactly like an SSN, so a nine-digit number that SSA records do not match to the applicant should go to review rather than be treated as an ordinary thin file.
The direct fix in the US is the SSA's electronic Consent Based SSN Verification service (eCBSV). It lets a permitted entity, which means a financial institution or its service provider, check whether an SSN, name and date of birth combination matches SSA records, and it returns a yes or no answer that names the field that doesn't match and flags a death record if there is one. It needs the number holder's written consent, and it's a paid service. A synthetic identity built on a real SSN with a different name or date of birth should come back as a no match, which is why eCBSV belongs early in the onboarding flow.
How are generated faces and deepfake identity documents used?
In FIN-2024-Alert004, issued on November 13, 2024, FinCEN reported that criminals have used generative AI to alter or create the images on identity documents such as driver's licenses and passports. They combine those images with stolen or entirely fake personal information to create synthetic identities. FinCEN's analysis of BSA data shows accounts opened this way being used to receive and launder the proceeds of other schemes, including check fraud, credit card fraud and authorized push payment fraud, and as funnel accounts.
Institutions often found these documents by re-reviewing a customer's account opening documents. FinCEN's red flags include:
- A photo with visible signs of alteration, or one that does not fit the customer's other details, such as a date of birth that makes them much older or younger than the photo suggests
- Multiple identity documents that are inconsistent with each other
- A third-party webcam plugin during a live verification check, or a request to change communication method after repeated glitches
- A customer who declines to use multifactor authentication
- A reverse image search that matches the identity photo to an online gallery of AI-generated faces
- Geographic or device data that does not match the identity documents
- A new account with rapid transactions, high payment volumes to risky payees such as gambling sites or digital asset exchanges, or many chargebacks or rejected payments
FinCEN notes that no single red flag is necessarily suspicious by itself. When a SAR relates to this activity, it asks filers to include the key term "FIN-2024-DEEPFAKEFRAUD" in SAR field 2 and in the narrative, along with any key terms for the underlying typology.
Which onboarding check catches which synthetic identity attack?
No single onboarding check catches every synthetic identity attack, so the useful question is which check answers which attack and where each one runs out. The map below is FluxForce's own analysis, built from the sources cited on this page; treat it as a starting point for your own control testing, not as a regulatory standard.
| Attack | Check that catches it | Where the check runs out |
|---|---|---|
| Real SSN paired with a different name or date of birth | SSA eCBSV match on SSN, name and date of birth, with consent | Needs written consent and a paid enrollment; US only |
| Credit privacy number sold as a "new SSN" | eCBSV or bureau SSN-to-name history; route any mismatch to review | A CPN that matches nothing looks like a thin file unless someone checks |
| Randomized post-2011 SSN | Name and date of birth match against SSA records, not geography | Geographic issuance checks no longer work for these numbers |
| Cultivated thin file with piggybacked tradelines | Authorized-user tradeline review and cross-lender application velocity | Looks like a genuine first-time borrower on a single bureau pull |
| AI-generated or edited ID document | Document checks plus re-review of opening documents, reverse image search | A clean-looking image passes a visual check |
| Injected selfie or deepfake video | Live capture with presentation attack detection and injection controls | NIST says these mechanisms don't cover every case |
| Ring of identities on shared phones, addresses or devices | Linking applications on shared attributes across the portfolio | Invisible when each application is reviewed alone |
| Bust-out after months of good payments | Monitoring for synchronized drawdowns and limit requests across products | Fires late; the loss may already be booked |
Reading across the rows, identity-data checks stop the cheapest synthetics at the door, media checks stop the generated documents, and only portfolio-level linking catches the patient rings.
What are the red flags of synthetic identity fraud?
The red flags of synthetic identity fraud are mismatches in the identity data, signs of a cultivated rather than lived credit history, links between applications that should be unrelated, and coordinated drawdowns. Treat them as reasons to look closer, never as proof on their own.
At application
- SSN that SSA or a bureau holds under a different name or date of birth
- Nine-digit number that matches no record and is presented by an applicant with no prior history
- Phone number, email address or device shared with other recent applications
- Mail-drop or shared address appearing on many thin-file applications
During cultivation
- Rapid growth in tradelines driven by authorized-user additions rather than the applicant's own accounts
- Perfect minimum payments with no wider sign of a real life behind them, such as utility or rental history
At bust-out
- Several accounts drawn down within days, mostly on cash advances and easily resold goods such as electronics or gold
- Contact details that go dead immediately after the drawdown
What are the biggest identity theft and synthetic identity fraud cases?
One of the biggest synthetic identity fraud cases charged in the US is the 2013 New Jersey ring, which prosecutors called one of the largest credit card fraud schemes the Department of Justice had ever charged. It allegedly used more than 7,000 fake identities and more than 25,000 fraudulent credit cards to cause over $200 million in confirmed losses. Investigators traced $60 million in proceeds through 169 bank accounts, and the ring wired millions of dollars abroad, including to India and the United Arab Emirates.
That case is useful for compliance teams because it shows every element on this page: drop addresses, sham companies acting as data furnishers to the bureaus, bought tradelines, and complicit merchants converting credit into cash. It also shows why synthetic identity fraud belongs in the AML program. The proceeds didn't stay in the credit card system; they moved through bank accounts and overseas wires.
What do regulators expect for synthetic identities?
Regulators expect identity checks that can catch an invented person as well as a stolen one, and suspicious activity reporting when a synthetic identity is found. FATF Recommendation 10 requires financial institutions to identify the customer and verify that identity using reliable, independent source documents, data or information.
In the US, the Customer Identification Program rule at 31 CFR 1020.220 requires banks to verify identity within a reasonable time after account opening, and to use non-documentary methods when, among other cases, the customer opens the account without appearing in person. Banks must file a SAR on a transaction of at least $5,000 that they know or suspect involves funds from illegal activity or is designed to evade BSA requirements, under 31 CFR 1020.320. FinCEN's deepfake alert adds the key term for GenAI-linked identity fraud described above.
Section 314(b) of the USA PATRIOT Act gives financial institutions a safe harbor to share information with each other to identify and report possible money laundering or terrorist activity, which may include fraud. That matters here because a synthetic ring usually hits several lenders at once, and no one lender sees the whole picture. FinCEN published a new 314(b) fact sheet on June 12, 2026, and rescinded the December 2020 version.
How do you explain a synthetic identity decision?
A risk score alone gives the analyst little to work with. To decide on a thin-file applicant, the reviewer needs to see which signals drove the result: an SSN that the bureau holds under a different date of birth, a device shared with other recent applications, a phone number only days old. Those reasons belong in the case file, because the analyst who declines the application or closes the account has to be able to defend that decision later.
US lenders have a legal reason too. Under Regulation B, 12 CFR 1002.9(b)(2), a statement of reasons for adverse action must be specific and indicate the principal reasons, and saying that the applicant failed to reach a qualifying score on the creditor's scoring system is insufficient. A lender that denies credit because a model flagged an applicant as a likely synthetic identity still owes the applicant specific reasons, so those reasons need to be recoverable from the model output and the case record.
How do you measure whether the controls work?
Alert volume says little about whether a synthetic identity program is working. These measures say more, and most come from data a lender already holds:
- Confirmed synthetic identities approved at onboarding, tracked by booking month so each change in controls can be compared with earlier cohorts
- First-payment defaults and charge-offs later confirmed as synthetic, as a share of all credit losses
- Synthetic accounts caught before the bust-out compared with those found only after charge-off
- Analyst time spent per confirmed synthetic case
- Approval rates for genuine thin-file applicants
The last measure keeps the program honest. Refusing every thin file would stop most synthetics, but it would also turn away the real first-time borrowers a lender wants, and that cost has to be weighed against each synthetic identity stopped.
How does synthetic identity fraud relate to other typologies?
Synthetic identity fraud sits next to several typologies that compliance teams see together.
- Bust-out fraud is the end point of most synthetic schemes: a patient credit-building phase, then every line drawn down at once.
- Money mule accounts are sometimes opened under synthetic identities to receive and pass on the proceeds of other fraud, which FinCEN also saw with GenAI-made documents.
- First-party fraud involves a real person misrepresenting their own circumstances; in synthetic identity fraud the applicant doesn't exist, so the SAR narrative and the detection signals differ.
- Identity verification and liveness detection are the main preventive controls at the door.
- Enhanced due diligence can be triggered after onboarding when a profile shows synthetic signals, and it means a full file review that goes beyond a sanctions and PEP check.
The wider onboarding controls are set out in our guides to KYC identity verification and customer due diligence.
Where do AI agents fit in catching synthetic identity fraud?
AI agents fit in the evidence work around each application: running the identity, document and linking checks in parallel and handing a person one file with the gaps marked. FluxForce is the Agentic OS for Regulated Industries. AI agents that investigate AML, sanctions, fraud and KYC alerts and prepare the case. Your analyst makes the call, with evidence an examiner can replay. You decide how much each agent does on its own, and every one has a kill switch.
In KYC / KYB Onboarding, the agents check ID documents for tampering and expiry, run selfie and liveness checks, flag synthetic and borrowed identities and match the applicant's details across sources. They screen every connected person and set a first risk rating from the factors in your own risk policy.
Your team approves every account; missing items are listed rather than hidden, and every check and decision is timestamped. The file stays linked to the customer, so a later review can see exactly what was known on the day the account opened.
Fit and limits, stated plainly: the agents don't replace eCBSV, bureau data or your liveness vendor; they use the data sources you connect. Synthetic identities that pass every source check on day one can still only be caught later, by monitoring for the bust-out pattern. Onboarding time and catch rates depend on your products and market, which is why FluxForce measures them on your own applications in shadow mode before anything changes.
Get the KYC POC test plan to see which of these checks would have flagged the synthetic identities already in your book.
How FluxForce detects synthetic identity fraud
FluxForce AI agents monitor synthetic identity fraud-related patterns in real time, surface red-flag activity for analyst review, and produce evidence-backed decisions with full audit trails.














