Video: Threat Briefing | Detecting and Preventing Synthetic Identity Fraud | Duration: 3612s | Summary: Threat Briefing | Detecting and Preventing Synthetic Identity Fraud | Chapters: Introduction and Overview (30.03s), Synthetic Identity Overview (165.40999s), Synthetic Identity Fraud (314.51s), Synthetic Identity Threats (452.43002s), North Korean IT Campaign (636.24s), North Korean IT Fraud (658.165s), Organizational Implications and Mitigation (1378.01s), Addressing Purple Delta Risks (2050.015s)
Transcript for "Threat Briefing | Detecting and Preventing Synthetic Identity Fraud":
Everyone and thank you for joining us today. I'm Rosie, a senior threat intelligence analyst at Recorded Future. And I'm joined by my colleague Mitch, a principal threat intelligence analyst specializing in North Korean threat activity. Today, we're going to be discussing the growing threat of synthetic identities and how organizations can identify and defend against these risks using Recorded Future. Before we dive in, we'll begin with a quick introduction about us, who we are. So as I said, I'm Rosie. I sit on the risk insights team within Insit Group, which is Recorded Future's research and analyst function. Our mission is to help business leaders manage, measure and communicate cyber risk by translating the more complicated threat intelligence into clear risk insights and actionable recommendations. Mitch, I'll pass over to you just for a quick intro. Sure, thanks Rosie. I'm Mitch Hazard. I'm a Principal Threat Intelligence Analyst on the Strategic and Persistent Threats team here at INSIC group, where we focus on researching state sponsored cyber activity, and my area of focus is North Korea. Great, thanks Mitch. So we'll get started. So today's presentation is going to be structured into three sections. We'll begin with a threat overview, so defining synthetic identities, what they are, how they're being used for things like fraud, sanctions evasion, and potentially intellectual property theft, and we'll also explore why many current detection mechanisms are falling short. Next, I'll pass over to Mitch, who's going to walk us through a case study of Purple Delta, which is institute research detailing how North Korea has used synthetic identities to embed malicious operatives within organisations and generate revenue for the regime. And then finally, Section three will focus on the threat outlook and implications for organisations. We're translating these findings into those practical implications with an assessment on how the threat is likely to evolve over the next few years, the key risks to monitor, and also some actionable mitigations. And we'll end with a Q and A. Diving straight into our threat overview, I'm going to start by baselining our understanding, so what is a synthetic identity? Well, at its core, a synthetic identity is a fabricated digital persona built from a blend of real, stolen and invented information. It looks legitimate on paper but doesn't correspond to an actual human being. Now within this concept, there are two main types of synthetic identities. Firstly, there are manipulated identities, so these are a genuine identity that's been altered, for example, by adjusting a date of birth, an address, or elements on a credit file while the underlying person remains the same. So an example would be an individual with a poor credit score that modifies part of their financial history in order to appear more credit worthy to a financial organization. There are also manufactured identities, which will be our focus today, and these are fully invented personas built by paying real data points, things like valid national IDs or social security numbers, with fabricated details like fake names, addresses or date of birth. So an example would be a threat actor creating a synthetic profile to open accounts that appear legitimate, but they're tied to no real individual. And as I said, these manufactured identities are going to be our focus for today because they're the most relevant in the threat landscape that we're currently seeing. Now, it's their hybrid nature of these manufactured identities which make them so dangerous. They're real enough to pass automated checks but they're not tied to a natural person. Traditional identity systems rest on three key assumptions: firstly, that a victim will report misuse secondly, that manipulated records will show some form of inconsistency and thirdly, that data models are built around real individuals. Manufactured identities break all three. There's no victim. Their date data is internally consistent because it's invented, and since the identity never belonged to an actual person, it sits outside of traditional forward mechanisms that would track and flag those anomalies. So the result is a quiet, unmonitored gap in identity and access management systems, one that adversaries are now exploiting at scale. So how are they doing practice synthetic identities are being weaponised in two key ways. Firstly, synthetic identity fraud. This is the fastest growing category of financial crime worldwide. Criminal groups don't just create these synthetic personas for a single transaction they cultivate identities over months, sometimes even years, opening bank accounts, establishing credit histories, and building normal looking digital behaviors until that profile is what they call fully seasoned. And once that comes, they can use their identity to secure large loans, establish lines of credit, and then simply vanish, leaving financial institutions to absorb those losses. Now, this trend is extending beyond the banking sector. Banks have been the traditional victims for this, but now we're seeing a move towards e commerce and digital platforms because they're designed for rapid onboarding. They rely on pretty lightweight verification checks that synthetic identities can easily bypass. Fraudsters can then use them for high value purchases and then trigger chargebacks, shifting the cost onto merchants. But the financial impact is only one piece of a much larger threat. What we're seeing now is synthetic identities increasingly being used to gain employment, especially for remote or contractor roles, providing access to corporate networks, credentials and intellectual property. The North Korean IT worker scheme is the most visible example of this today, and Mitch is going to cover that in a little bit more detail later in the presentation. Now, what is overtly clear is that synthetic identity attacks are happening at scale, and this is what the data is showing. The latest reporting by iProve, which is a software company specializing in facial biometric verification technology, they identified the following. So synthetic identity document fraud has increased 300% in 2025 compared to the 2024. At the same time, we're seeing deep fake fraud increasing, so up to 1100% year over year, and over the past year we've also seen a rise in deepfake injection attacks, which rose by around seven eighty three percent. Now this surge is fueled by many things, but most importantly is the availability of AI tools that can produce convincing passports, national IDs, and even synthetic biometric data, so things like faces, fingerprints, and iris patterns. All of this is freely available and on the open source web and often for quite low costs as well. The rise in injection attacks is particularly alarming. So unlike presentation attacks, which involve holding a fake video or image up to a webcam, injection attacks bypass the camera entirely, and they feed synthetic media directly into an identity verification system's input stream. So in practice, this means there's no physical camera involved, there's no actual applicant present, and no visible manipulation for the system to detect. So what this means is adversaries can now animate synthetic faces in real time during onboarding, fraud checks, remote hiring, interviews, all of these things while appearing completely illegitimate to automated verification systems. Alongside better technology on the attacker side, there's also some structural weaknesses on the defender's side. When we look at how enterprises are responding to this threat, there are two systemic gaps that continue to create opportunities for Synthetic Identity attacks. The first is inflated vendor claims. So what we're seeing is a lot of identity verification providers are now advertising deepfake proof solutions, yet many are still failing to detect injection attacks, which are the very techniques that adversaries are adopting the most rapidly. And this gap between marketing and actual capability creates a false sense of security for organizations, leaving them vulnerable while believing they are actually protected. There's also quite a significant knowledge deficit at play. So the latest RSA report in 2005 detailed that nearly half of respondents, including many identity and access management specialists, failed basic questions about identity security, with embarrassing cybersecurity professionals scoring the lowest overall. So the result is the enterprises are purchasing tools they don't know how to evaluate and defending against threats they don't always fully understand. Humans are also unable to bridge that gap. We drastically overestimate our ability to spot synthetic media. In fact, eye proof reporting also shows that only 0.1% can identify synthetic identities, only nine percent can recognize deepfake videos. Over a third of adults 55 have never actually heard of a deepfake and around twenty nine percent would not report suspicious content even if they identified it. So this creates a behavioural gap. We've got awareness with our action, and in high pressure environments where many analysts already face alert fatigue, for example, relying on human intuition is not viable. Now, one of the clearest demonstrations of how effective synthetic identities be is, as I said, the North Korean IT worker campaign internally tracked as Purple Delta. And with that, I'm going to hand over to Mitch, who's going to walk us through this, how this campaign operated and tell us what it means about the future of this threat. So over to you, Mitch. Thanks, Rosie. North Korean IT workers are North Korean nationals with real technical backgrounds who seek remote IT and software developer roles using false identities. Estimates put the scale at up to 10,000 workers worldwide. And it's important to understand that they're not operating as independent fraudsters. They're dispatched, organized, and controlled by North Korean government linked entities, which is why this activity is sustained and systematic. Economically, the incentive is huge. Individual workers can earn up to $300,000 per year and a substantial portion of their earnings is used to fund the North Korean regime. Operationally, they may be physically based in places like North Korea, China, Russia, parts of Southeast Asia, but they can appear local by working virtually through laptop farms, often located in The US, as well as using VPNs, proxies, and remote desktop software, giving them global reach. To get hired and stay hired, they build synthetic identities, including CVs, GitHub portfolios, and social media. And they're often supported by a wider ecosystem that includes front companies, facilitators, and collaborators, who help with logistics, accounts, the financing, and sometimes even interviewing. The takeaway is that this is a state sponsored scalable operation that targets the remote hiring pipeline. And we're going to look at ways you can identify these individuals and yours. So, one practical starting point for triage is email patterns. I want to highlight that none of these are definitive alone. Plenty of legitimate people have similar username styles. But in combination with other signals we will talk about, they can help you spot risk earlier. So common patterns that we see, include a name plus two to four digits, words like dev or code, paired with competence words like good, great, creative, or themed words like star, sun and moon, or dragon, night, god. Two things to keep an eye out for are misspellings in the names and surnames and cases where the email doesn't match the claimed identity on the CV or profile. And one caution here, while these examples are Gmail, North Korean IT workers also use other email providers. So treat this as a pattern basically. So GitHub is often a credibility anchor for technical candidates. So it's a great place to look for inconsistencies and the two perfect candidate patterns. Here's an example profile, and what we're watching for are clusters of signals, just not single tell. So, first the handle format. They are often one to three words, plus two to four numbers, and frequently include terms like dev or code, similar to their emails. Second, the profile optics. Bios that lean heavily on templates and emojis, plus inflated seniority claims. So they'll often have titles such as senior or principal with many years of experience. And that text doesn't always match the actual project history or their profile age. Third, over claiming. So as you'll see in this profile here, many of their profiles have vast or broad tech stacks and skill grids that look more like marketing rather than a realistic specialization. And finally, location. So words such as remote, home, self employed, these are common and are not inherently suspicious in 2026. But when these words show up alongside these other signals we're talking about, they become more meaningful. So the main takeaway is that GitHub can quickly surface whether a candidate's story is coherent or whether it looks manufactured. So, one of the fastest ways to validate a persona is to pivot on the email address. In this example, the GitHub user blue0316 is tied to the email mikecreative0908gmail dot com. When you search for that email though, you'll find that it's associated with a different identity. This identity with the name of Mike Limonce. And it is also connected to a completely different GitHub persona, which is called Fantasy Night. So if a candidate CV claims one name, but their email footprint ties to multiple unrelated identities, that's a major signal. And even if you assume a benign explanation for this, it's still a strong indicator that you should stop and validate before proceeding. So the takeaway here is simple. Email reuse across personas is a high signal lead, and it helps connect what look like separate people into a single operator. So after identity pivots, the next step is network pivots. So if you look at followers and following lists, starred repositories, things like that on a suspicious GitHub profile. You'll often find that clusters of accounts that look remarkably similar. So what you're seeing on this slide is that repeated profile traits, right? The same kinds of bios, the same skill grid aesthetic, similar naming conventions, non authentic profile images create a recognizable pattern. And most importantly, these accounts often follow and star each other, which gives us a relationship signal just beyond vibes. So instead of judging one profile in isolation, we can ask, does this account sit inside a cluster of similar personas? And that's often where confidence increases. Resumes are another place where patterns show up quickly, especially template reuse. So when you compare multiple resumes side by side, you'll often see repeated layouts, repeated phrasing, a gamification of skills with these completion bars or charts, and the same kind of aspirational language that doesn't really map to any sort of concrete outcome. So make sure to cross reference resumes with the candidate's other application materials, such as GitHub profiles or portfolio websites. And these individuals are constantly adapting their tactics, techniques, procedures. So what have we seen recently? Looking forward, there are three key developments to watch out for regarding North Korean IT workers. So first is operational scaling. North Korean IT workers are increasingly what we'll call industrializing the process of renting and buying Upwork, email, and LinkedIn accounts, and even employing interviewees for hire. So the person on the calls, on the interview or on the team meetings may not be the person that's actually doing the work. So this creates a risk of layered deception that's harder to spot unless you're validating identities and continuity. Second is, it's just more than IT. While technical roles remain the highest risk, we're seeing activity in areas like animation and graphic design, civil engineering, and project management. So the implication is that cyber defenses can't just focus on remote software developers. It has to be a broader organizational initiative. And third, APT activity. So we're already seeing signs of this. For example, in some of our previous reporting, we've observed overlap between North Korean IT workers and other threat groups like Purple Bravo. And the concern here is that there is a access handover model where workers get hired into a target and then pass it over to cyber operators. So we should treat this as a plausible escalation path and monitor it accordingly. Now, do we do about it? So the main idea is that you can detect risk at multiple stages, and each stage has both signals to watch for and actions you can take. So at the application stage, look for repeated patterns across candidates, such as similar CVs and cover letter language, GitHub patterns, naming and IOC matches, and do lightweight validation like checking emails and the candidate's public presence. During the interview, the goal is continuity and authenticity, requiring video, avoiding virtual backgrounds when feasible, simple live validation steps, and asking experience questions that are hard to fake with generic answers. For background checks, you should enforce consistency across documents and look for altered ID indicators. And when it's practical, use in person verification for higher risk roles. And then after hiring, assume that some risk remains. There is a potential that we could still accidentally hire an North Korean IT worker. So you should implement strong two factor authentication, monitor for anomalies, and alert on VPN and remote desktop connections, or impossible travel indicators. So this isn't about a one gotcha check, right? That there's no one flag that we should look out for. It's a layered verification across the entire hiring pipeline, proportional all to the risk of the role. And with that, I will hand it back to my colleague, Rosie, who will take it. Thanks Mitch. Yep, so moving on to organisational implications. Obviously Mitch has just walked us through an example of how the threat is manifesting today and touched on how that may unfold in future. I just want to dive in a little bit more to what happens next in the threat outlook over the next twelve to twenty four months. Based on current trends and adversary behaviour, we're assessing four key developments that are most likely to shape the next two years. The first one sounds pretty obvious, but synthetic identities will almost certainly become harder to detect. We're seeing rapid advances in generative AI, identity deception tools and synthetic biometrics, which will likely continue to outpace many existing verification systems. Adversaries, including state aligned actors like North Korea, will increasingly create these identities to appear legitimate across background checks, CVs and their digital fingerprints as well, and we assess that traditional identity proofing will struggle to detect things like deep fake IDs, video spoofing and multi personal fraud without a fundamental redesign of verification processes. At the same time, we assess that state sponsored infiltration via remote hiring will expand. As Mitch mentioned, we've seen North Korean IT workers being deployed abroad to secure remote contracts. Other hostile states may also try and replicate and adapt to this model using remote and contractor roles as a long duration, low risk access vector for intelligence collection, IP theft and covert funding. As Mitch touched on as well, we're seeing a lot in the space of software development, IT support and other technical roles. Over the next two years, given nation states' focus on critical infrastructure, we may well see a shift towards defence, supply chain, fintech and other critical infrastructure sectors as well. Even partial access can enable industrial espionage, long term exploitation, so that's something we would be watching out for. In response to escalating threat, we assess that governments and regulators are going to tighten identity and hiring compliance, so we're expecting more advisories and regulatory frameworks. Some of these may evolve into mandatory compliance regimes for identity verification in hiring and contractor management. We're not seeing the mandatory compliance regimes yet, but, you know, if this continues, we may well do. Certainly, organisations are going to start facing increased pressure to verify not only who the user is at onboarding but whether the identity remains trustworthy over time. And then finally, perhaps a longer term outlook, but identity verification shifting to a continuous zero trust model. So as we've mentioned before, identity assurance will no longer be that one time pre employment event, as we're seeing adversaries exploiting remote work organisations will adopt have to adopt continuous verification models, reasserting trust at every session, high risk action and privileged access requests. So this is likely to include things like enhanced background checks, live biometric validation, ongoing risk scoring and potentially session level authentication tied to that zero trust architecture. Taken together, these trends point to a clear conclusion that synthetic identities are no longer just a fraud problem. They represent a multi dimensional risk that directly impacts financial stability, operational resilience, regulatory requirements and compliance and long term competitiveness as well. So within the Risk Insight team, it's our role to help leaders visualise how emerging threats translate into business risk, and this is done across four core risk categories, the first being financial loss, so this is a reference to direct monetary and recovery costs brand impairment referencing loss of trust and reputation operational disruption so the interruption of critical business operations legal and compliance failure encompassing regulatory violations and legal exposure and then finally, competitive disadvantage, which refers to a loss of strategic market advantage. Now from a synthetic identity perspective, the most immediate and viable impact is financial loss. As we saw in the North Korean IT workers scheme, some of these operatives were demanding salaries of up to $300,000 a year. And identity related crime in general has caused an estimated 8,800,000,000.0 in losses in The US in 2022, with projections for that figure to rise to up to £58,300,000,000 by 2030. But for many organisations, this represents just the first layer of impact, masking longer term costs tied to investigations, control redesign, workforce disruption and regulatory scrutiny as well. Because brand and trust consequences are often more damaging and longer lasting. Confirmed synthetic identity infiltration undermines confidence amongst customers, partners, regulators, investors and employees, particularly once activity is publicly disclosed or linked to any form of enforcement action. When the US Department of Justice, FBI and partner governments publicly disclosed investigations and indictments, companies that had unknowingly hired these operatives faced media scrutiny, questioning of their hiring and access controls, and even where firms were treated as victims, coverage frequently focused on how the infiltration occurred, which affected regulatory and partner trusts, raised questions about sanctions compliance internally, contractor vetting, hiring and risk management practices, particularly for those operating internationally or in sensitive sectors as well. Much of that discussion focused on whether organisations had failed to take reasonable steps to prevent sanctioned individuals from being hired. This is shifting a narrative from a victim of fraud to a failure of due diligence, which directly impacts trust in the long term. Failure to detect synthetic identities within one's environment can also trigger sanctions violations, leading to multi million dollar fines, potential criminal liability, including executive prosecutions, mandatory regulatory investigations, both domestically and internationally, and eventually potential loss of government contracts, clearances, licenses and third party contracts as well. There's also operational disruption, so that occurs when adversaries gain access to internal systems, interfere with workflows or degrade any system reliability, and so the result could be things like service interruptions, delayed operations and increased recovery costs as well. Competitors' disadvantage compounds all of these things. Once inside, adversaries could potentially exfiltrate sensitive data, proprietary processes, strategic plans, for example, and combined with reputational damage, this accelerates competitors and undermines an organization's long term market positioning. So as Mitch has already touched on, defending against synthetic identity infiltration requires a shift in mindset, so identity must be treated as a continuous security function, not just a one time verification step. So how do we do this? Well, we recommend things like securing identity with AI, so leveraging AI and machine learning, advanced analytics to detect synthetic identity sorry, synthetic activity that humans and traditional wall based systems might miss, so things like deep fake injections, synthetic biometrics and anomalous behavior patterns, which move too quickly for manual review. Secondly, validating dynamic digital identity signals. So as we mentioned, static identifiers, things like IDs, resumes, platform profiles aren't sufficient by themselves. It's worth noting that legitimate workers tend to exhibit pretty consistent digital signals across location, their device characteristics, network behavior and working patterns, whereas fortunate workers, as Mitch touched on, they often rely on things like VPNs, proxy service, third party intermediaries, and so that can create some detectable instability over time. So, as we said, organisations should be aiming to continuously validate things like IPs, geolocations, device footprints and behavioural norms. Third, we can't emphasise this enough, is strengthening remote hiring and third party onboarding. Mitigation should include enhanced identity verification for remote freelancers, contractor roles, particularly those with access to sensitive systems, source code or data, and then organisations should be verifying the true location of their workers, scrutinising the use of any intermediaries or account managers to confirm devices before granting access, and you know managers requiring live camera on videos where that's possible as well. And then fourth is sort of aligning with regulations and that transition towards zero trust models as well. So as we said, identity assurance shouldn't be ending at onboarding. Zero trust is not always feasible, but it's a gold standard. So, yeah, it's important to remain identities remain trustworthy over time with periodic rechecks of identity signals, traceability across employees, contractors and vendors as well. So just a quick summary of what we've covered today. As you know, we're seeing an increase in synthetic identities driving financial crime, sanctions evasion and IP theft. Advances in deepfake and generative AI are creating personas that evade KYC and hiring checks. We're seeing an increase in overhyped tools and low awareness as well, and continuous AI led verification, secure remote access, that revalidation are essential to combat this threat. So I'll move on now to any questions that might be submitted. Okay, we've had one question come through here, which is: If we were worried that our current controls couldn't catch something like the Purple Delta case, where should practitioners focus first? I don't know, Mitch, if you wanted to jump in quickly on that one. Sure, I can take that. So I would say that there are a few things that you should focus on. I think the first is, as we've talked about throughout this presentation is, you should continuously require identity checks. The person that interviews for the role, must be the person that shows up and onboards to do the work. I would say in the case of purple Delta, schemes often rely on remote access or like laptop farms. So, you as the employer can kind of control the remote work reality. So you should look for things like, are there any strange behaviors on your work device? What is the location that this worker claims that they are working at? And then things like VPN and proxy usage or remote desktop tooling that would may indicate some sort of suspicious activity. Along those lines, I would say, you should always assume that a North Korean IT worker might get through the checks you have in place in your hiring pipeline. So you should just assume least privilege by default. And then monitor for high signal behaviors. So I would say things like unusual repository cloning, unusual requests for secret access, large data movements, things like that. And I guess I would sum it up by finding that a lot of orgs struggle with it because these checks are scattered across different departments. So HR might have checks, IT might have checks and the SOC might have checks, but they're not communicating with one another. So I think it would behoove organizations to operationalize this as a single high risk, remote hire playbook. I don't know, Rosie, if had anything to add there. Yeah, that's kind of the direction that I was thinking as well. It's correlating those signals across multiple departments and across HR identity and access management, any threat intel teams, but also just making sure, in general, that there is awareness and education within organizations as well. Through training, there are plenty. You know, we sit within the cybersecurity space where this threat is one that's discussed quite frequently. That's not the case across a lot of organisations, so just making sure that awareness is there and that all employees are looking out for this type of behaviour can be useful as well, and because that adds an additional layer of protection and just making sure that there are clear channels where that any suspicious behaviour can be reported as well. Okay, I think that was it for the questions, so I just wanted to thank everybody for joining the webinar today. Thank you to Mitch as well for giving his take on the situation. There is some reporting if you are interested. The Risk Insights team has published a report on synthetic identities, which covers a lot of the points that were mentioned today in this webinar. And then Mitch has also authored some research. I don't know if you wanted to quickly plug that research, Mitch, well. Yes, I don't have it on the slide, but if you are a portal user and you just go to platform and search for purple delta on that threat actor identity card, there should be a whole bunch of recent in secret reporting on their activity. Great, thank you so much. And thanks everybody for joining today.