Security Evangelist at Gen

Senior Principal Research Engineer

Junior Data Scientist

AI Researcher

Research Manager

Our team invests countless hours in the exploration, discovery and investigation of emerging trends such as AI and futuristic technology. Stay informed about the evolving landscape of cyberthreats through our diligent efforts and insights.

Threat Researcher at Gen

Research

<h2>Introduction&nbsp;</h2>Infostealers are constantly evolving, and so are the techniques they use to bypass <a href="https://security.googleblog.com/2024/07/improving-security-of-chrome-cookies-on.html" target="_blank" rel="noreferrer noopener">Application-Bound Encryption (ABE)</a>. In recent weeks, Vidar has been among the most actively developed stealers and, apart from <a href="https://x.com/GenThreatLabs/status/2064836539952361976" target="_blank" rel="noreferrer noopener">multiple updates to its string obfuscation</a> and <a href="https://x.com/GenThreatLabs/status/2061462524604649840" target="_blank" rel="noreferrer noopener">a reworked approach to protecting its configuration</a>, it has also introduced a novel technique for bypassing ABE. And while there have been many other changes in Vidar lately, with new versions dropping every week, in this blog post we focus solely on the ABE bypass and its technical aspects.&nbsp; &nbsp;<h2>The ABE Bypass&nbsp;</h2>From a high-level perspective, Vidar’s approach to bypassing ABE is somewhat similar to that of Remus/Lumma, in that both try to extract the <code>v20_master_key</code> directly from the browser’s memory. What sets them apart, however, is how they achieve it. Since we’ve already covered the inner workings of ABE in earlier posts (<a href="https://www.gendigital.com/blog/insights/research/remus-64bit-variant-of-lumma-stealer" target="_blank" rel="noreferrer noopener">Remus</a>, <a href="https://www.gendigital.com/blog/insights/research/voidstealer-abe-bypass" target="_blank" rel="noreferrer noopener">VoidStealer</a>), we don’t revisit them here and focus instead on the bypass itself. For readers interested in the broader background, we encourage checking those earlier posts.&nbsp;&nbsp;In short, the end goal of infostealers is to obtain the <code>v20_master_key</code>, as it alone is sufficient for decrypting any ABE-protected data tied to a specific application. This key can be obtained in one of two ways: either from disk, where it is protected by two layers of <code>CryptProtectData</code> (one applied in the context of the logged-in user and one as <code>NT AUTHORITY\SYSTEM</code>), or from the browser’s memory – the direction Vidar took.&nbsp;In memory, however, the key is kept protected with <code>CryptProtectMemory</code> using the <code>CRYPTPROTECTMEMORY_SAME_PROCESS</code> flag. This ensures that even if an attacker manages to read the encrypted form of the key from memory, they cannot simply decrypt it (they would need to invoke the complementary <code>CryptUnprotectMemory</code> function directly from within the browser process). That leaves attackers with two problems to solve: how to find the encrypted key in memory, and how to get it decrypted.&nbsp;<h2>Locating the v20_master_key in Browser’s Memory&nbsp;</h2>Vidar’s approach to locating the encrypted <code>v20_master_key</code> in memory involves scanning the entire memory of a running browser for a specific pattern. But that first requires a browser process to scan.&nbsp;When the target browser is already running, Vidar reuses the existing process. However, it doesn’t read its memory directly. Instead, it creates a fork of it using <code>NtCreateProcessEx</code> with <code>SectionHandle</code> set to <code>NULL</code>, instructing the kernel to clone the address space of the process specified by the <code>ParentProcess</code> as a copy-on-write mapping. The resulting process has no threads and is never resumed – it exists purely as a static memory snapshot.&nbsp;&nbsp;To perform the fork, Vidar first opens a handle to the target browser via <code>OpenProcess</code> with <code>PROCESS_QUERY_INFORMATION | PROCESS_CREATE_PROCESS | PROCESS_VM_READ</code>, falling back to <code>PROCESS_CREATE_PROCESS</code> if the initial call fails. It then passes this handle as the <code>ParentProcess</code> argument to <code>NtCreateProcessEx</code>, requesting <code>PROCESS_ALL_ACCESS</code> on the resulting child. If the fork fails, Vidar falls back to scanning the live browser process directly.&nbsp;

If no target browser process is found running, Vidar creates a new isolated desktop via <code>CreateDesktopA</code>, assigns it a name of the form <code>v20_%d</code> (where <code>%d</code> is replaced with the browser’s index in Vidar’s configuration), and launches the target browser in there. In the case of Vidar version 2.1, the browser is launched with the command line <code>--no-first-run --disable-gpu about:blank</code>, although the exact command line varies across versions. Since modern browsers use a multi-process architecture, there are typically multiple processes matching a given browser. Vidar enumerates all of them (up to 64) and applies the fork-and-scan procedure to each one independently.&nbsp;

With a handle to the forked process in hand, Vidar proceeds to enumerate its virtual memory regions using <code>NtQueryVirtualMemory</code>. It iterates through the entire browser’s address space, collecting regions that match a specific profile: committed (<code>MEM_COMMIT</code>), private (<code>MEM_PRIVATE</code>), and either readable (<code>PAGE_READONLY</code>) or read-write (<code>PAGE_READWRITE</code>). Vidar records up to 4096 memory regions matching these requirements, along with their base addresses and sizes. The actual pattern scanning is then performed in parallel with up to 64 worker threads, depending on the system.&nbsp;&nbsp;

Each thread scans its assigned regions using <code>NtReadVirtualMemory</code>, searching for a 32-byte pattern. When a match is found, the thread reads 8 bytes at offset +32 from the match, interprets them as a pointer, and validates that it points to committed, private, and readable memory (the expected memory state for the encrypted <code>v20_master_key</code>).&nbsp;&nbsp;

This naturally raises the question of what pattern Vidar searches for. The answer is a 32-byte signature: <code>76 32 30 00 ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? 03 00 00 00 00 01 ?? ?? ??</code>, where <code>??</code> denotes a wildcard (Vidar uses the byte <code>AA</code> as the wildcard marker in its internal representation). Notably, the first four bytes correspond to the ASCII <code>v20\x00</code>, which is a common prefix used for ABE-related data.&nbsp;&nbsp;

We assess with high confidence that this pattern is intended to match a node structure within Chromium’s <code>Encryptor::KeyRing</code>, the structure that holds the keys used by the <code>Encryptor</code> class. Now, this part is admittedly a bit tricky as it relies on several assumptions we made about the internal memory layout, as well as on implementation-specific representations of various C++ standard library types <code>(std::map</code>, <code>std::string</code>, <code>std::vector</code>, <code>std::optional</code>), which is why we frame this as a high-confidence rather than a certainty.&nbsp;&nbsp;Our reasoning is as follows: <code>KeyRing</code> is <a href="https://github.com/chromium/chromium/blob/cf5d502bcaedb22927e3829a18be1ff6d8b6f598/components/os_crypt/async/common/encryptor.h#L132" target="_blank" rel="noreferrer noopener">defined</a> as <code>std::map&lt;/*tag=*/std::string, std::optional&lt;Key&gt;&gt;</code>, and <code>std::map</code> is typically implemented as a red-black tree, where each tree node stores three pointers (left child, parent, and right child) followed by the key-value pair. In this case, the pair is in the form of <code>std::pair&lt;/*tag=*/std::string, std::optional&lt;Key&gt;&gt;</code>, meaning that the <code>std::string</code> and the <code>Key</code> object might be laid out consecutively in memory within each node. For a <code>v20_master_key</code> entry, the tag corresponds to the string <code>v20</code>, which at 4 bytes including the null terminator fits within the small string optimization (SSO) buffer and is therefore stored inline rather than heap-allocated. The <code>Key</code> class that immediately follows <a href="https://github.com/chromium/chromium/blob/cf5d502bcaedb22927e3829a18be1ff6d8b6f598/components/os_crypt/async/common/encryptor.h#L93-L98" target="_blank" rel="noreferrer noopener">contains three member variables</a>: <code>std::optional&lt;mojom::Algorithm&gt; algorithm_</code>, <code>std::vector&lt;uint8_t&gt; key_</code>, and <code>bool encrypted_ = false</code>.&nbsp;Putting it all together and applying it to a concrete example from a browser memory dump, we get the following layout:&nbsp;

Please note that while the exact layout may contain minor inaccuracies on our part, the overall structure should be roughly correct. For easier orientation, the following diagram illustrates how the individual structures are nested within each other:&nbsp;

At this point, it should be clear how the scan pattern maps onto the searched structure. The 32-byte match covers the <code>std::string</code> tag <code>v20\x00</code> and the beginning of the <code>Key</code> object up to and including <code>algorithm_</code>. The pointer at offset +32 from the start of the match then lands precisely on <code>key_.begin</code>, the first member of <code>std::vector&lt;uint8_t&gt;</code> that holds the encrypted <code>v20_master_key</code>.&nbsp;Matches from all threads are collected into a shared buffer and subjected to a majority-voting scheme: candidates consisting mostly of zeroes are discarded, the rest are pairwise-compared, and each is scored by the number of identical copies found. This voting is performed independently for each browser process, producing one best candidate per process.&nbsp;<h2>Decrypting the v20_master_key via APC Injections&nbsp;</h2>For each browser process that produced a match, Vidar now holds a candidate address pointing to a potential encrypted <code>v20_master_key</code>. However, as already noted, since the key is protected with <code>CryptProtectMemory</code> using <code>CRYPTPROTECTMEMORY_SAME_PROCESS</code>, it can only be decrypted by invoking <code>CryptUnprotectMemory</code> from within the browser process itself. Vidar solves this by injecting an Asynchronous Procedure Call (APC) into the live browser process.&nbsp;Vidar supports two APC injection methods, selected based on whether ESET or Bitdefender is detected on the system. If either of these products is present, Vidar uses the first method, which it refers to as “classic” (based on decrypted strings) – it creates a suspended thread in the target browser via <code>CreateRemoteThread</code> (with <code>NtTestAlert</code> as the thread start address and the <code>CREATE_SUSPENDED</code> flag), queues an APC using <code>NtQueueApcThread</code>, resumes the thread, and then waits for it to complete via <code>WaitForSingleObject</code> with a randomized timeout. When the thread resumes, it calls <code>NtTestAlert</code>, which flushes the APC queue and executes the injected routine.&nbsp;

If neither product is detected, it uses the “special” method – it finds an existing thread in the target browser using a combination of <code>CreateToolhelp32Snapshot</code> and <code>Thread32First/Thread32Next</code>, opens it via <code>NtOpenThread</code>, and calls <code>NtQueueApcThreadEx2</code> with the <code>QUEUE_USER_APC_FLAGS_SPECIAL_USER_APC</code> flag. Unlike regular APCs, special user APCs execute immediately. Therefore, the target thread does not need to be in an alertable wait state.&nbsp;

In both cases, the APC routine to be executed is <code>CryptUnprotectMemory</code>, called with the following three arguments: the address of the encrypted <code>v20_master_key</code> candidate, its size (32 bytes), and <code>CRYPTPROTECTMEMORY_SAME_PROCESS</code> for the <code>flags</code> parameter.&nbsp;&nbsp;Since <code>CryptUnprotectMemory</code> decrypts the data in place, once the APC executes, the key is decrypted directly in the live browser's memory. To read it, Vidar creates a second fork of the browser process and reads the key from the same address using <code>NtReadVirtualMemory</code>. It then compares the result against a snapshot of the encrypted key taken from the first fork before the APC call – if the data is unchanged, decryption failed, and Vidar moves on to the next browser process’s candidate.&nbsp;If the data differs, Vidar uses the same second fork to perform an additional verification: it scans the forked process’s memory for the byte sequence <code>v20</code> and, treating each match as a potentially ABE-encrypted entry (cookie, password), attempts AES-256-GCM decryption with the key candidate using <code>BCryptDecrypt</code> (successful GCM tag verification confirms that the key is correct).&nbsp;Whenever the data has changed (indicating that decryption of the key took place), Vidar re-encrypts the key in the browser’s memory by injecting another APC with <code>CryptProtectMemory</code> as the routine, preserving the browser’s state. If no candidate produced a successful decryption (i.e., the data remained unchanged after every APC call), Vidar takes a more aggressive approach: it terminates all browser instances, restarts the browser, and repeats the entire scan-decrypt-verify cycle.&nbsp;By abusing APC injections instead of the more traditional approach of injecting code via <code>NtWriteVirtualMemory</code> that performs the decryption, Vidar trades one suspicious behavior for another. APC injections are arguably less common and may thus slip past some detection systems, though they are still well-known techniques monitored by AV and EDR solutions. But that is the nature of ABE bypasses – each has its own advantages and disadvantages, and that’s exactly why infostealers keep inventing new ones. &nbsp;<h2>Indicator of Compromise (IoC)&nbsp;</h2>459daa809751e73f60fbbe4384a7d1653c36bb06945e4eb3635270924241100a (Vidar 2.1)&nbsp;

A Technical Walkthrough of How Vidar Defeats Application-Bound Encryption

Inside Vidar’s ABE Bypass: From Memory Scanning to APC Injections

You have a summer flight booked. You have seen the headlines about jet fuel shortages, cancelled routes and higher travel costs. Then a text message arrives: your flight has been affected, and you need to click a link to confirm your refund or rebook before the offer expires.&nbsp;That message feels plausible because the situation around it is real. In April, <a href="https://www.iata.org/en/pressroom/2026-releases/2026-04-17-01/" target="_blank" rel="noreferrer noopener">IATA warned</a> that Europe could start seeing flight cancellations by the end of May if jet fuel shortages materialized. Airlines and governments have since been trying to reassure travelers, while also publishing guidance on passenger rights, refunds and re-routing.&nbsp;&nbsp;That is exactly the kind of confusion scammers like. They do not need to cause the disruption, they only need to arrive at the right moment, with the right brand logo and the right sense of urgency.&nbsp;<h4>Why travel disruption works so well as a scam hook&nbsp;</h4>Most scam messages fail because they arrive out of context. A random refund email from an airline you never use is easy to ignore. A refund message from the airline you are flying with next week, during a period of public travel disruption, is different.&nbsp;The scammer’s job is to make you think: “This could be real.”&nbsp;That can happen in several ways. You might receive a fake cancellation email that links to a refund form. You might search for an airline’s customer service number and land on a fake support page. You might post publicly on social media asking an airline for help, then receive a reply from an account that looks official. Or you might get a WhatsApp message that knows your hotel name, travel dates or booking reference.&nbsp;Each version uses the same pressure point: you are worried about losing money, losing your flight or being stranded.&nbsp;Criminals are already using disruption this way. In March, <a href="https://www.theguardian.com/money/2026/mar/15/travel-scam-airline-accounts-fake-refunds-iran-war-flight-disruption" target="_blank" rel="noreferrer noopener">The Guardian reported</a> that fake airline support accounts were targeting travelers affected by Middle East flight disruption. Some victims were looking for help or refunds, then received replies from accounts pretending to represent airlines. The scam moved into direct messages, where victims were asked for details and sent links that led to money being taken instead of refunded.&nbsp;&nbsp;The jet fuel story gives scammers another pretext. A message can claim that your flight was cancelled, your fare changed, your fuel surcharge failed, or your refund must be claimed within a few hours. None of those claims needs to be true to be effective.&nbsp;<h4>The scam may start before the message arrives&nbsp;</h4>A fake travel message becomes much harder to spot when it includes real information.&nbsp;This is the idea behind <a href="https://www.gendigital.com/blog/insights/research/reservation-hijack-scam" target="_blank" rel="noreferrer noopener">Reservation Hijack scams</a>, a form of targeted phishing where cybercriminals use real booking details to make a message look legitimate. In some cases, attackers may use information exposed through breaches or compromised travel systems. In other cases, they collect details from emails, social media posts, previous leaks or fake customer service interactions. Gen has previously described Reservation Hijack scams as scams that use real hotel booking details to make fraudulent payment or verification requests feel authentic.&nbsp;&nbsp;That distinction matters. A scam message does not need your credit card number to fool you. Knowing your name, your destination, your travel dates and your booking platform may be enough.&nbsp;<h4>What these scams can look like&nbsp;</h4>A travel disruption scam usually does not begin with a strange offer. It begins with something boring and administrative. That is part of the trick.&nbsp;You may see messages like:&nbsp;“Your flight has been cancelled due to fuel supply issues. Claim your refund here.”&nbsp;“Your rebooking is pending. Confirm passenger details within two hours.”&nbsp;“Additional airport tax or fuel surcharge is required to keep your booking.”&nbsp;“Your hotel reservation will be cancelled unless payment is verified.”&nbsp;“Due to high call volume, please continue this request on WhatsApp.”&nbsp;The exact wording changes, but the structure is familiar. There is a problem, a deadline and a link. The message may also move you away from the official channel into SMS, WhatsApp, Telegram, direct messages or a phone call. Once that happens, the scammer controls the conversation.&nbsp;There is another detail worth watching: some messages do not ask for money immediately. They start by asking for your booking reference, last name, phone number or one-time password. That can still be dangerous. Those details can help scammers access a booking, impersonate you to customer support, reset an account or make the next message more convincing.&nbsp;<h4>Your real passenger rights can be used against you&nbsp;</h4>If an airline cancels your flight, you may have rights to a refund, re-routing or assistance, depending on where you are travelling and the rules that apply. The European Commission said passengers affected by cancellations continue to have rights to reimbursement, re-routing or return, airport assistance and compensation for last-minute cancellations in relevant circumstances. It also said airlines cannot retroactively add fuel surcharges to already published final ticket prices.&nbsp;&nbsp;That information is useful for travelers, but it is also useful for scammers.&nbsp;A fake refund message becomes more believable when people already know refunds are possible. A fake rebooking link becomes more believable when airlines are genuinely changing schedules. A fake “fuel surcharge” demand becomes more dangerous when fuel prices are in the news, even though legitimate retroactive charges may not be allowed in many cases.&nbsp;This is why the safest route is not to judge the message only by whether the story sounds plausible. During travel disruption, many scam stories will sound plausible. Judge the channel instead.&nbsp;<h4>How to check before you click&nbsp;</h4>Start from the airline or booking platform’s official app or website. Do not use the link in the message. Open the app yourself, type the website address yourself, or use the contact details listed on your original booking confirmation.&nbsp;Check whether the same cancellation, refund or payment request appears inside your account. If the message is real, the airline or hotel should be able to confirm it through an official channel.&nbsp;Be careful with social media support. If an account replies to your public complaint, check whether it is linked from the airline’s official website. Look at the handle carefully. Fake support accounts often copy logos and names, but have small differences in spelling, low follower counts or recently created profiles. The Guardian reported that fraudsters often scan social media for passengers trying to contact legitimate airline accounts and reply to them first, which makes the scam harder to spot.&nbsp;&nbsp;Do not share one-time passwords, payment card details or login credentials with anyone who contacts you unexpectedly. A real airline or hotel does not need your banking OTP to send you a refund.&nbsp;Be especially suspicious of pressure. “You have 30 minutes to claim your refund” is not customer service. It is a tactic.&nbsp;<h4>What to do if you already clicked&nbsp;</h4>If you entered payment details, contact your bank or card provider immediately. Ask whether the transaction can be blocked or reversed, and request a new card if needed.&nbsp;If you shared login details for an airline, hotel or booking platform, change the password from the official site or app. Use a unique password, not one you use elsewhere. Turn on two-factor authentication where it is available.&nbsp;If you shared a one-time password or authentication code, assume the account may have been accessed. Check your booking details, payment methods and recent account activity.&nbsp;Report the scam to the airline, hotel or booking platform being impersonated. If the scam came through social media, report the account as impersonation. If money was taken, report it to the relevant fraud authority in your country.&nbsp;<h4>The real lesson for summer travel&nbsp;</h4>Travel disruption changes how people behave. They search for support numbers in a hurry. They post publicly because customer service queues are long. They accept strange instructions because they are worried about losing a flight or refund.&nbsp;Scammers build around that behavior.&nbsp;The safest assumption is simple: if a message about your flight, hotel or refund creates urgency and gives you a link, pause. Go back to the official source. A real cancellation will still be there when you check through the airline or booking platform. A scammer needs you to act before you do.&nbsp;

Travel disruption gives scammers the one thing they need most: a believable reason to rush you

Your flight was cancelled. Is the refund message real? 

A fake job interview page is easier to trust than a fake ticket shop.In samples reviewed by Gen threat researchers, scammers used branded hiring pages that looked like ordinary recruitment flows: a company logo, a recruiter profile, a 30-minute meeting, and a button to continue with Google or Facebook. There was no attachment to open and no software to install. The credential theft attempt sat inside a step many job applicants already expect: signing in to schedule a call.FIFA is an obvious lure ahead of the 2026 World Cup. On May 22, ESET reported fake FIFA-themed websites that copied ticket and merchandise flows to steal money and personal data. Our researchers later found another FIFA-themed domain, hxxps://fifa-mg[.]shop, which followed a naming pattern not mentioned in that report. Domains like these can be short-lived, often appearing for hours or days before they are blocked, abandoned, or taken down.Our researchers also found a different kind of FIFA lure: hxxps://fifahiring[.]com, a fake hiring page that encouraged users to schedule a call with an alleged FIFA recruiter. When a user tried to register with a personal Gmail address, the page asked for a business email instead. The page was steering victims toward credentials with more value: accounts that could provide access to a company environment.

<h4>The FIFA hiring page did not point to FIFA</h4>The fake page looked believable at first glance.

It used FIFA branding, presented a recruiter identity, and copied parts of the experience users see on legitimate recruiting and scheduling platforms. The recruiter’s name and photo appeared to match a real LinkedIn profile, which made the page feel more credible. For publication, we should keep that person’s identity redacted.The domain did not match FIFA’s real hiring infrastructure. FIFA’s legitimate careers pages are jobs.fifa.com and fifa.pinpointhq.com. The fake page used neither. It also loaded the favicon from the real FIFA recruitment site hosted at pinpointhq.com, giving the page a small visual cue borrowed from the legitimate service.Other signals pointed in the same direction. The domain and certificate were newly created, from May 23. The site used commodity hosting infrastructure, including Vercel and Render, rather than a corporate environment. Those signs are not enough on their own, but they fit the rest of the evidence.The login flow was the clearest part. Credentials were posted to:fifeq2026eqbackeq[.]onrender[.]com/api/loginThe Google sign-in prompt was not a real Google authentication window. It was rendered inside the page itself. The visible controls and links did nothing, and the page did not send the user into a genuine Google sign-in flow. According to the analysis, the only Google asset loaded was the favicon, apparently to make the fake window look familiar.A page can look like it opened a Google login without ever handing control to Google.Researchers also observed hxxps://careers-fifahiring[.]com, which later became inactive, and hxxps://fifajobs[.]com, which showed the same content and was indexed by Google.&nbsp;<h4>The recruitment lure went beyond FIFA</h4>The FIFA page was one version of a wider recruitment phishing pattern. The brand can change. The flow stays familiar: recruiter profile, meeting prompt, business email, fake sign-in.Researchers found similar fake hiring pages abusing other major brands, including Heineken, Hilton, Coca-Cola, Netflix, PepsiCo, Delta,&nbsp;and Spotify. In the Heineken case, the page appeared to use the identity of someone who, according to LinkedIn, really works as a recruiter at the company.

Additional hunting found two other recruitment-themed kits:hxxps://aquent-talent-jobs[.]com hxxps://hays-talent-opportunities[.]comThe Hays-themed version targeted Facebook credentials instead of Google credentials. Researchers assessed that the Aquent and Hays pages shared code, although they differed from the earlier FIFA examples.

This flexibility makes the campaign harder to frame as a single World Cup scam. FIFA provides the timely hook, but the mechanism is recruitment phishing. A scammer can swap the logo, recruiter name, and sign-in provider while keeping the same basic path to the victim’s credentials.<h4>Why fake recruitment pages work</h4>Recruitment gives attackers a clean excuse to ask for things that might otherwise raise suspicion.A candidate expects to click a scheduling link. They expect a recruiter profile. They expect a calendar prompt. In some cases, they may also expect to use a work email, especially for corporate, contractor, or partner-facing roles.The fake login window turns that ordinary flow into credential theft. Users are used to “Continue with Google” and “Continue with Facebook” buttons, but the button itself proves nothing. A genuine authentication flow should take the user to the provider’s domain and controls. A fake page can draw a convincing copy while keeping the victim inside the attacker-controlled site.The business email prompt is also a warning sign. A page that rejects a personal email and asks for a work address may be trying to raise the value of the stolen account.<h4>What job seekers should check before signing in</h4>Start from the company’s official careers page, not from a link in a message, ad, search result, or forwarded post. For FIFA, that means checking jobs.fifa.com or fifa.pinpointhq.com, not trusting a domain because it contains “fifa,” “career,” “hiring,” or “jobs.”Look closely at the domain before entering credentials. Scam pages often add plausible hiring words around a brand name: career, talent, jobs, hiring, portal, or connect. A polished page on a new domain is still a new domain.Treat the sign-in step as the real checkpoint. If a page offers Google or Facebook login, check whether the flow actually opens a real provider-controlled authentication page. A fake window embedded inside the original site, broken controls, inactive links, or a provider logo used only as decoration should stop the process.Use security software that includes web and phishing protection. These tools can block known malicious pages before they load, which is especially useful when scam domains look polished, appear in search results, or are shared through otherwise legitimate channels.<h4>What companies can do</h4>Recruitment pages are public by design, which makes them easy to copy.Security teams should monitor for newly registered domains combining their brand with hiring-related terms. They should also watch for pages that reuse favicons, recruiter photos, job descriptions, and calendar-booking language from legitimate hiring flows.HR and recruiting teams should make the official hiring path easy to verify. Candidates should not have to guess whether a domain is legitimate. A clear careers page, consistent recruiter communication, and warnings about unofficial scheduling links can remove some of the ambiguity attackers rely on.<h4>Conclusion</h4>The fake FIFA hiring page worked because it did not feel unusual. It looked like a normal step in a hiring process.Attackers do not need to invent new behavior when they can copy one people already trust. A branded scheduling page, a recruiter profile, and a familiar sign-in button can be enough to move a victim from interest to credential entry.The decision point is the login. Before entering a work email or password, check the domain, check where the authentication flow really goes, and verify the role through the company’s official careers site. Security software with web and phishing protection adds another layer, especially against short-lived scam pages that may only stay online for hours or days.IOCs:hxxps://fifahiring[.]com hxxps://careers-fifahiring[.]com hxxps://fifajobs[.]com hxxps://fifa-mg[.]shop hxxps://heineken-careers[.]com hxxps://aquent-talent-jobs[.]com hxxps://hays-talent-opportunities[.]com fifeq2026eqbackeq[.]onrender[.]com/api/login

Scammers are copying recruitment and calendar-booking flows to make fake Google and Facebook sign-ins look routine. 

Fake hiring pages abuse FIFA and other major brands to steal work credentials 

Threat Research Team Lead

Since January 2026, we have detected 7.4 million malicious incidents tied to residential proxy traffic, affecting 572,000 users in our telemetry. In this model, the household whose connection is being used is often not the buyer of the traffic. It is the exit point.That is the uncomfortable part of residential proxy networks. A home router, laptop, phone, smart TV, or other always-on device can quietly become part of a commercial network that routes strangers’ internet traffic through ordinary consumer IP addresses. To the websites on the other end, that traffic looks as if it came from a real household, because it did.The technology itself is not inherently malicious. Businesses use residential proxies for market research, ad verification, brand protection, security testing, and services that need to see the web as a real user would. But the same feature that makes these networks useful also makes them attractive for abuse: traffic from a household IP address is hard to block without risking false positives against legitimate users.The problem for consumers is simpler. Many people do not realize their device or connection can be enrolled into this kind of network at all. Some users opt in through pay-for-bandwidth apps. Others are enrolled through “free” VPNs, bundled software development kits inside unrelated apps, or unwanted software that installs proxy components without meaningful consent.Here is how the market works, who pays for it, and what it can cost the household that unknowingly hosts it.<h2>What is a residential proxy</h2>A residential proxy is an internet connection – typically a home laptop, phone, smart TV, or router – that has been enrolled into a third-party network and is being used to route someone else's traffic. To the rest of the internet, that traffic looks like it is coming from an ordinary household IP address assigned by a consumer ISP, not from a datacenter or&nbsp;VPN provider.Residential proxy networks are sold commercially. The model is two-sided: some people pay to tunnel their traffic out through someone else's connection (much like a VPN, except the exit point is a real home), while others get paid to provide those exit endpoints. Some are structured as paid peer-to-peer networks, where any participant's connection can serve as an exit for any other.Buyers pay for access to pools of millions of these "real user" IPs because they are extremely difficult to block, or rather, blocking them carries real consequences: the traffic originates from genuine households, so any block risks shutting out a legitimate user – effectively a false positive.Common buyers include market research firms, price-comparison services, brand-protection vendors, security teams testing their own defenses, and consumer privacy tools. At the abusive end of the market, the same infrastructure used by operators running automated purchasing bots, scrapping campaigns, ad fraud, credential stuffing, phishing and other attacks where hiding the true origin of the traffic matters. Demand is also being shaped by large-scale web scraping, including&nbsp;<a href="https://blog.includesecurity.com/2026/06/the-smart-tv-in-your-livingroom-is-a-node-in-the-aiscraping-economy/">scraping for AI training data</a>. As anti-bot systems become better at identifying traffic from datacenters and known VPN providers, residential IP addresses become more attractive because they look like ordinary users. That does not make every residential proxy user malicious. It does mean consumers may be carrying traffic for businesses and actors they have never heard of.<h2>How consumer devices end up in these networks</h2>There are several ways a home device can become part of a residential proxy network. Some are explicit. Others are easy to miss.The most obvious route is the pay-for-bandwidth model. Apps such as Honeygain, IPRoyal Pawns, and PacketStream pay users to share their internet connection. The user opts in, but they may not understand who will buy access to that connection later, what kind of traffic will pass through it, or what happens if that traffic triggers abuse reports.A second route is the “free” VPN or proxy app. Instead of charging a subscription, the app monetizes the user’s connection by reselling bandwidth. In practice, the user may think they are getting a privacy tool while also becoming infrastructure for other people’s traffic.A third route is through bundled software development kits, or SDKs. A developer adds a third-party component to an otherwise unrelated app, such as a flashlight, wallpaper, game, or utility. The developer gets paid when the app is active. The user sees the host app, not the proxy network behind it.The fourth route is malware or potentially unwanted software. In those cases, the proxy component is installed silently or deceptively, with no meaningful user consent at all.<h2>Why the "consent" is rarely meaningful</h2>Even when consent technically exists, it is often not the kind of consent most people would recognize.The disclosure may be buried in a long terms-of-service document. The wording may be vague, with phrases such as “share your unused bandwidth,” “support the app,” or “join our network.” These phrases do not clearly tell the user that arbitrary third parties may route traffic through the user’s home IP address.There is also the question of downstream use. Users are rarely told who buys access to the network or what those buyers do with it. Their connection could be used for benign market research. It could also be used for scraping, automated account creation, ad fraud, credential stuffing, or phishing infrastructure.The problem gets worse when the proxy component sits inside another app. The host app’s privacy policy may mention a partner SDK by brand name, but that still does not make it clear to a normal user that their device can become a proxy endpoint.Uninstalling the visible app may not always resolve the issue cleanly either, especially when proxy components are installed separately or bundled with other unwanted software.<h2>The risks consumers actually carry</h2>When a household device becomes a residential proxy node, the user takes on risks created by someone else’s traffic.The first risk is attribution. Abusive, fraudulent, or illegal activity routed through the device appears to come from the household’s IP address. ISP abuse notices, platform bans, fraud checks, and law enforcement requests point first to the subscriber, not to the person who bought the proxy traffic.The second risk is performance. Continuous third-party traffic can consume bandwidth, slow the home network, hit data caps, or violate ISP terms of service. The user may only notice that the internet feels slower, a bill is higher, or a service suddenly blocks access.The third risk is security exposure. Proxy clients run persistent background processes and maintain network connections so they can relay traffic. That adds always-on networking software to the device, often with limited visibility for the user.The fourth risk is reputation. A home IP address used for abuse can end up on blocklists used by banks, retailers, streaming services, and online platforms. The result can be account friction, failed logins, extra verification steps, or outright blocking from services the household actually uses.There is also a family angle. A child or another household member may install a free game, VPN, or utility without understanding the consequences. The legal, financial, and reputational risk still lands on the account holder.<h2>What our telemetry shows</h2>Since the beginning of 2026, we have detected 7.4 million malicious traffic incidents associated with residential proxy traffic, affecting 572,000 unique users in our user base.&nbsp; This is not a measurement of the entire residential proxy market. It reflects what we observe in our telemetry: malicious traffic, detections, and related activity affecting protected users. The real footprint is far larger, because residential proxy networks span major consumer platforms and many devices outside our visibility.The activity is global, but it is concentrated in a smaller set of countries. India leads by a wide margin, with roughly 83,000 affected users, followed by Vietnam with 61,000 and Brazil with 39,000. The Philippines (23,000), Poland (18,000), Indonesia (18,000) and Thailand (14,000) form a second tier, with Egypt, Spain, Ukraine, Argentina, Mexico and Italy each accounting for 10,000 to 13,000. The spread across South and Southeast Asia, Latin America, and Eastern and Southern Europe reflects where the host apps that carry this software are most widely installed.That geographic spread likely reflects a combination of app distribution, local install patterns, user appetite for free or pay-for-bandwidth apps, and the economics of bandwidth resale. It should not be read as a simple ranking of where residential proxy abuse originates.The chart below breaks down the threats we observed in this proxy-related traffic. Two categories dominate: phishing and malicious advertising. Behind them sits a long tail of consumer-facing fraud, including e-shop scams, generic scams, financial scams, and fake-tutorial pages, alongside outright malware such as trojans and other malicious traffic.&nbsp;

In the malicious proxy traffic we observed, scams and malware dominate. The important point for consumers is not only that their connection may be used by someone else. It is that, when abuse happens, their household IP address can be the return address.<h2>What consumers can do</h2>Most people will not identify a residential proxy component by name. But there are practical warning signs.Review “free” VPNs, proxy apps, bandwidth-sharing apps, and utilities from unknown publishers. Check whether they mention bandwidth sharing, peer-to-peer networking, traffic routing, or network participation in their terms. If an app offers payment for leaving your device online, assume your connection may be resold unless the provider explains exactly how traffic is vetted and who can buy access.On phones and computers, remove apps you do not recognize or no longer use. On home routers, check the list of connected devices and look for anything unfamiliar. If your ISP sends abuse notices, websites start blocking you, or your home IP address suddenly fails fraud checks, a proxy component is one possible explanation.Security software can also help by detecting unwanted proxy clients, bundled components, and related malicious activity. But this should not be a guessing game for consumers. Software that turns a household connection into a commercial exit node should be named clearly and made easy to remove.<h2>The takeaway&nbsp;</h2>With residential proxiesthe user’s connection becomes an infrastructure for someone else’s business.The issue is not that residential proxy technology exists. It has legitimate uses. The issue is whether people know when their devices are part of it, understand what kind of traffic may pass through their connection, and have a real choice to opt out.The fix is not to pretend consumers will read every EULA. This software should be named clearly, shown plainly, and made easy to remove before a household connection becomes someone else’s infrastructure.

Your home connection could be routing traffic for strangers. Here's how residential proxy networks work, how devices get enrolled and what our telemetry reveals about the risks for consumers.

Who's Really Using Your Home Internet Connection?

<h4>Key Points&nbsp;</h4><ul><li data-list-item-id="e43a143fd054561c50dba51071686adff">Gen Threat Labs has been tracking GoFlateLoader, a widespread Golang loader used to deliver multiple infostealers, including Amatera, Remus, Lumma, Vidar and StealC.&nbsp;</li><li data-list-item-id="e3d3b5d0814ea5b26aa75a8d3d6013a39">GoFlateLoader appears both in x86 (32-bit) and x86-64 (64-bit) variants, matching the bitness of the payload it is supposed to execute.&nbsp;</li><li data-list-item-id="e51ccafcfb8c418d894aeb685004a0ba5">The loader is designed for in-memory payload execution and is deliberately inflated with a massive PE overlay to hinder detection</li><li data-list-item-id="ec4bafef8a1113d24bc2b4b3a943228d1">The two main delivery paths are via supposedly cracked software and a <a href="https://research.checkpoint.com/2026/impersonation-click-hijacking-and-tds-inside-a-malware-distribution-ecosystem/" target="_blank" rel="noreferrer noopener">malicious TDS recently documented by Check Point Research</a>, which may redirect users to dedicated landing pages serving password-protected archives containing GoFlateLoader.&nbsp;</li></ul><h4>Introduction&nbsp;</h4>Not every threat that matters is technically sophisticated, and that is also the case with GoFlateLoader, which is a rather simple loader written in Go, whose sole purpose is to decode and execute the payload in memory. What stands out the most is not what the loader does but rather what it does not do – it comes without anti-debugging, anti-VM, or sandbox-evasion checks, and also lacks API hashing or CFG obfuscation, the kind of tricks that loaders almost always come with. Instead, GoFlateLoader relies on one of the simplest yet still effective tricks to stay under the radar – it appends a massive PE overlay at the end of the file, deliberately inflating the binary's size (hence the name GoFlateLoader).&nbsp;Nevertheless, it is important to understand that malware does not necessarily need to be elegant or stealthy. In the end, the only thing that matters to attackers is whether it can successfully deliver the final payload. And if that is the only metric, GoFlateLoader appears to be working fairly well for its operators, as we continue to see it widely distributed across our user base. Since the beginning of April, we have protected more than 33,000 unique users from it, with the most affected countries including Brazil, India, Argentina, Mexico, Turkey and Spain.&nbsp;

<h4>GoFlateLoader’s Execution Flow&nbsp;</h4>If we strip away the obfuscation noise that the Go compiler and the author’s decoy code add, the loader’s execution flow is short and linear:&nbsp;<ul type="disc"><li data-list-item-id="efa51af3b2859fd9c684efca0c82d3b66">Copy the encoded payload blob from the <code>.rdata</code> section onto the stack.&nbsp;</li><li data-list-item-id="e879bb970c6d1a81b8a42512222aceaa4">Decode the payload using a small, multi-stage, custom byte-level transformation into a valid PE.&nbsp;</li><li data-list-item-id="ebff4f340af15b75cbea0d682b26b7645">Parse the PE headers of the decoded payload (image base, image size, sections, data directories, etc.).&nbsp;</li><li data-list-item-id="e37fb26cad3d18a4a0c8bf9b2595d661f">Allocate an RWX memory region using <code>VirtualAlloc</code> with <code>MEM_COMMIT | MEM_RESERVE</code> and <code>PAGE_EXECUTE_READWRITE</code> flags, trying the payload’s preferred <code>ImageBase</code> first and falling back to an OS-assigned address if that fails.&nbsp;</li><li data-list-item-id="ebbde6d77cdc942c0a6bac79f4f8a6c28">Map the decoded payload into the allocated memory region by copying the PE headers first, then walking the section table and writing each section to its target virtual address.&nbsp;</li><li data-list-item-id="ed16738229e0f5c70a981c0c5f9b715f9">Apply relocations (only if preferred base mapping fails).&nbsp;</li><li data-list-item-id="e76cde4bae1be6c21b897c350c6fd15db">Resolve imports by walking the standard <code>IMAGE_IMPORT_DESCRIPTOR</code> table and rebuild the IAT in place via <code>LoadLibrary</code> and<code> GetProcAddress</code>.&nbsp;</li><li data-list-item-id="e786887a3d34fc7ce344ee8375944799e">Transfer execution to the final payload using Go’s <code>syscall.Syscall</code>.&nbsp;</li></ul>&nbsp;This is essentially a textbook example of manual PE loading with the payload reconstructed and executed entirely in memory so that it never touches the disk.&nbsp;&nbsp;

Notably, GoFlateLoader transfers execution to the final payload via <code>syscall.Syscall</code>. In Go, <code>syscall.Syscall</code> has the signature <code>Syscall(trap, nargs, a1, a2, a3)</code>, where <code>trap</code> is the function pointer to invoke, <code>nargs </code>is the argument count, and <code>a1</code>, <code>a2</code>, and <code>a3</code> are the first three arguments passed to the target function. Here, trap is set to point to the final payload’s entry point, <code>nargs</code> is set to <code>1</code>, and since the execution is transferred to that entry point, the remaining arguments serve only as filler values and are never used by the invoked payload. In other words, GoFlateLoader is not issuing a <code>syscall</code> at all, but abusing <code>syscall.Syscall</code> as a generic call gate to hand control to the payload.&nbsp;&nbsp;It is worth emphasizing that GoFlateLoader consistently sets these arguments to the hardcoded values <code>1</code>, <code>2</code>, <code>3</code>, and <code>4</code>, which is unusual enough to stand out as a potential detection opportunity (combined with other patterns).&nbsp;

The loader also carries substantial junk/decoy code, which varies across builds and serves no purpose other than padding the control flow and complicating static analysis.&nbsp;&nbsp;&nbsp;<h4>Inflated PE Overlay&nbsp;</h4>GoFlateLoader’s defining characteristic is its massive size (typically 700-950 MB), caused by an artificially inflated PE overlay appended to the original loader.&nbsp;

This is a deliberate design choice intended to bypass detection, as excessively large executable files may pose challenges for AVs and EDRs, primarily due to performance and resource constraints. As a result, to preserve performance and avoid noticeable impact to end users, these solutions often enforce practical file-size limits for deep scanning and/or emulation.&nbsp;The same logic extends to automated analysis pipelines – bloated files frequently cause timeouts during heuristic analysis and effectively prevent the sample from being uploaded to cloud-based sandboxes or threat intelligence platforms, which enforce their own strict upload limits to manage bandwidth and storage costs. A prime example is VirusTotal, which enforces a strict 650 MB upload limit. The fact that GoFlateLoader consistently sits just above this threshold strongly suggests that VirusTotal is one of the platforms it specifically aims to evade.&nbsp;&nbsp;In most cases, GoFlateLoader’s PE overlays are filled with null bytes, though we have also observed builds with random padding.&nbsp;&nbsp;

Crucially, adding such a PE overlay comes at virtually no distribution cost for attackers, as the inflated data compresses to a fraction of its original size (assuming constant padding).&nbsp;&nbsp;<h4>Delivered Final Payloads&nbsp;</h4>The final payloads distributed through GoFlateLoader are predominantly information stealers, with the most prevalent currently being Amatera, <a href="https://www.gendigital.com/blog/insights/research/remus-64bit-variant-of-lumma-stealer" target="_blank" rel="noreferrer noopener">Remus</a>, and Lumma. Beyond these, we have also observed it distributing Vidar, StealC, and SvitStealer.&nbsp;GoFlateLoader comes in both x86 and x86-64 variants, with each variant matched to the architecture of the payload it is intended to execute.&nbsp;&nbsp;As for how GoFlateLoader binaries are distributed, we have observed at least two notable delivery paths. The first is through supposedly cracked software. The second involves the payload being served via <a href="https://research.checkpoint.com/2026/impersonation-click-hijacking-and-tds-inside-a-malware-distribution-ecosystem/" target="_blank" rel="noreferrer noopener">the malicious TDS recently described by Check Point Research</a>, which redirects users to a landing page that offers a downloadable archive and separately displays the password required to extract it. This approach provides several advantages for the threat actor. First, compression significantly reduces the binary size due to the constant-padded PE overlay. Second, because the archive password is displayed separately on the landing page, it makes it harder for AV and EDR solutions to obtain it and decrypt the archive. As a result, the payload may remain concealed until the victim manually extracts it.&nbsp;&nbsp;&nbsp;<h4>Summary&nbsp;</h4>GoFlateLoader is a reminder that prevalence and sophistication are not the same thing. At its core, it is a rather simple manual in-memory PE loader written in Go, with no novel anti-analysis tricks under the hood. Whatever evasion it achieves comes not from obfuscation or other anti-analysis tricks but from its size – the appended PE overlay is built to slip past size-constrained scanners and automated analysis pipelines. Yet, despite its low level of sophistication, GoFlateLoader remains an active and relevant threat, used to deliver a wide range of prevalent information stealers.&nbsp;&nbsp;Gen products protect our users against both GoFlateLoader and the payloads it delivers.&nbsp;&nbsp;<h4>Indicators of Compromise (IoCs)&nbsp;</h4>Note: The files listed in the IoCs that are not archives all exceed 650 MB in size and are therefore not available on VirusTotal.&nbsp;b88c5744975d2abb447aecc6c090fee9f8580413f4612eecdc6ed1973e8a1739 (password-protected archive containing GoFlateLoader x64 variant loading Remus; pwd: 1234)&nbsp;ed5ae7f36453c5a23e9868a5729d67e0549a11f6dea54f5f52d654a8f51d4902 (archive containing GoFlateLoader x64 variant loading Remus)&nbsp;841c9297cb8a2e0ff89433d13c05bfc760eb2e98e251cb8fa785d2ad7cbac05f (archive containing GoFlateLoader x86 variant loading Amatera)&nbsp;ece7c48eb411b24f26762ede83badb4a644c41d5777129381ac2541804d64fc2 (archive containing GoFlateLoader x86 variant loading Lumma)&nbsp;421ce2d2f49c23bbe9f60ef3b9cd38d7eb912ce02e56a61837656210069bd9e2 (archive containing GoFlateLoader x64 variant loading Vidar)&nbsp;121c2dc793b3873f75a29ec02241f94136de19c049382a50a50d0d5b99507073 (GoFlateLoader x64 variant loading StealC)&nbsp;2415db5081cec9bfd14ad6da1a66169fd96f13a49010c319a73d1ed6fafd4efa (GoFlateLoader x64 variant loading Vidar)&nbsp;d9917ade3b4c125a95b5d3e6343cde26145dfbf569bd7e2a843fd0c6fc8ddc28 (GoFlateLoader x64 variant loading Remus)&nbsp;4cf6893756f441522b94b36f10e5de0e47aeed4743f95c51650746d1ecf97e3d (GoFlateLoader x64 variant loading SvitStealer)&nbsp;8b89d6c9152d3aab97aadd515ecb69ca72654db2f25425759ba4b646853d737d (GoFlateLoader x86 variant loading Lumma)&nbsp;90ce4ff9da23ac150da0a8e17930cab1e369aa349fdc1b65691b70369145664a (GoFlateLoader x86 variant loading Amatera)&nbsp;

 Intentionally Inflated Go Binaries That Use Massive PE Overlays to Bypass Detection

GoFlateLoader: A Widespread Golang Loader Delivering Multiple Infostealers

Scam advertising does not thrive on social platforms because it is invisible. It thrives because it is designed to look legitimate, just long enough to reach victims before detection catches up.In our first investigation,&nbsp;<a href="https://www.gendigital.com/blog/insights/research/scam-ad-machine-meta">The Scam Ad Machine</a>, we showed the scale of the problem and how widespread scam ads have become on Meta platforms, with roughly 1 in 3 ads approved for publication in the UK and EU being a scam.&nbsp;This second part builds on that foundation, focusing on how these campaigns actually operate in practice.Across our investigation of Meta platforms, we saw the same pattern again and again, scammers combining normal advertising features, deceptive presentation tricks, and short-lived campaign tactics to stay in circulation. This is not a single evasion trick. It is an operating model.Below, we break down the techniques we found at scale, and why many of them are difficult to catch through automated moderation, manual review, or public transparency tools.<h2>How scammers hide in plain sight</h2><h3>Multi-facet ads, hiding one scam among legitimate links</h3>One of the most common techniques we uncovered is what we call multi-facet advertisements.These ads bundle multiple images, products, and links into a single ad. Most of the links point to legitimate marketplaces or well-known ecommerce platforms such as Amazon, eBay, AliExpress, or Etsy. Hidden among them is a single malicious or scam-related link, often leading to fake ecommerce stores, fraudulent service pages, account scams, or impersonation pages tied to public figures.The trick is simple: make the scam look like part of a normal shopping experience.By surrounding one malicious destination with a cluster of legitimate brands and products, the ad looks ordinary at first glance. That lowers suspicion for users and can also make review harder if the ad is assessed superficially. It also creates collateral damage, legitimate brands and products end up visually associated with fraudulent campaigns.From a detection perspective, this should be a solvable problem. A robust review process could inspect the full set of linked domains and compare what is shown with what is actually being promoted. In practice, however, we identified large volumes of these campaigns running simultaneously from different advertisers on Meta platforms, often linked to newly created advertisers and low-history accounts.

<h3>URL masking, when the ad shows one domain and sends users to another</h3>This is one of the most effective deception techniques we observed.Scammers make an ad appear to link to a trusted domain, such as a major marketplace or brand, while the click sends the user somewhere else. The ad displays a familiar name, but the landing page belongs to unrelated infrastructure controlled by the attacker.For users, especially on mobile, that visible domain cue is often enough. Many people do not inspect the final URL before clicking, and once they land on an attractive page with compelling offers or urgency signals, many will continue browsing.In our observations, Meta’s API may expose the domain shown in the ad, not necessarily the final destination the user reaches after clicking. That makes this technique harder to detect programmatically and significantly complicates analysis by transparency researchers, watchdogs, journalists, regulators, and law enforcement.In other words, what the ad appears to be, and where it actually leads, can diverge in ways that are difficult to verify at scale.

The actual link leads to the website selling products that are completely unrelated to the advertised ones.

<h3>Typosquatting, deceptive subdomains, and misleading previews</h3>Why abandon a tactic that still works?A classic technique remains highly effective, registering domains that closely resemble legitimate brands. A small spelling change, an extra character, or a visually similar character can be enough to fool users at a glance. In fast-scrolling environments, these differences are easy to miss.Scammers also use deceptive subdomain structures to look more credible. Instead of using a plainly suspicious domain, they may deploy subdomains such as go.[domain], shop.[domain], or offers.[domain]. These can look more legitimate to users and may also create inconsistencies in detection coverage. In some cases, we observed situations where one domain variant was flagged while a related subdomain remained active.Another layer of deception appears in the creative itself. In ads using domain impersonation or typosquatting, we also observed mismatches between the ad’s visual preview and its actual content. For example, a video thumbnail may show a legitimate product, a game scene, or unrelated imagery, while the underlying content promotes a scam. If neither an automated system nor a human reviewer clicks through or expands the media, the discrepancy may go unnoticed.This combination, deceptive domain naming plus misleading visuals, increases both click-through potential and evasion resilience.An example of an advertisement using the prefix “go.”. The link “GO.DATING4SINGLES.COM” does not lead to the website, but “DATING4SINGLES.COM” does.

<h3>Unicode text obfuscation that breaks detection, not readability</h3>Sometimes the URL is not the main problem, the text is.Many scam ads rely on familiar triggers, high-return investments, miracle products, fake financial services, urgent offers, or celebrity-backed promises. In plain text, these patterns should be relatively easy to flag for additional review.Scammers increasingly avoid this by using obfuscated characters and Unicode tricks. To a human reader, the text may look normal. To an automated system, parts of the same string may be transformed into unusual Unicode sequences, invisible joiners, or look-alike characters that break keyword matching and reduce classifier reliability.The result is elegant from the scammer’s perspective, readable to people, unreliable for machines.We observed this technique at scale in scam ads circulating across Meta platforms, often in combination with other evasive tactics such as domain impersonation and celebrity references.

<h3>Hidden impersonation, names that look real to people but not to machines</h3>Impersonation remains one of the fastest ways to manufacture trust in scam advertising.Scammers know that names associated with wealth, authority, or credibility can dramatically increase engagement, whether it is a celebrity, a business leader, or a title such as “Dr.” or “PhD.” Those same names should also be obvious candidates for automated scrutiny.Scammers work around this by manipulating text so that it looks normal to a user but is encoded differently for machines. One common technique we observed uses special spacing characters or joiners that visually preserve a name while changing its underlying structure. To the naked eye, it may read as a well-known person’s name. To a detection system, it may be processed as a different token entirely.This technique was especially common in scam ads designed to manufacture credibility through famous names, credentials, or authority signals.This advertisement uses the name of dentist Dr. Margaret Hunt to gain users' trust. At the same time, a special character is used in the name to avoid detection.

<h2>Why scam ads keep coming back</h2>The problem is not only that scam ads evade detection. It is that they regenerate faster than enforcement can disrupt the operation behind them.This is the Hydra problem of scam advertising. Remove one ad, and several variants can reappear through new accounts, duplicate creatives, and short-lived campaigns built to expire before meaningful intervention happens.Our investigation did not only examine what scam ads looked like. It also examined what happened to them over time, how they were removed, how quickly similar variants reappeared, and how easily the underlying operations continued.What we observed suggests a recurring pattern, many interventions remove individual ad instances, while the broader scam operation remains fully capable of regeneration.<h3>Takedowns that remove ads, but not the operation</h3>One of the first patterns we noticed was that removals often appeared to happen on a single ad instance basis.We observed cases where the same scam ad, or functionally identical copies of it, continued running even after one instance had been removed. This happened across different advertisers and, in some cases, within the same scam-linked advertiser account. The text, target audience, visuals, and destination could remain effectively the same, while only one instance was taken down.Based on these observations, moderation often appears to act on the ad object that was detected, rather than reliably escalating to advertiser-level or campaign-cluster disruption.For scammers, this is a gift. If enforcement targets copies one by one, duplication becomes a business advantage.<h3>Short-lived campaigns, high volume beats persistence</h3>Scammers do not need every ad to survive. They need enough ads to run long enough.A common pattern we observed was the use of many short-lived advertisements, often with small budgets and short runtimes. Instead of relying on one long campaign, scammers distribute risk across large numbers of disposable ads. If some are removed, others remain active and continue reaching users.This strategy fits the takedown pattern above. If moderation is slower than campaign turnover, short-lived scam ads become highly efficient. The ad dies, the operation continues.We also repeatedly observed advertiser pages with very limited organic activity, minimal followers, random page names, and behavior patterns consistent with throwaway infrastructure, yet still capable of launching scam ads.Permanence is not the business model, throughput is.<h3>User reporting is too slow for disposable scam ads</h3>We also tested the user reporting path directly by reporting scam ads from platform accounts.In multiple cases, we observed the same ads still active after reporting, and in some cases visible again on subsequent checks. This does not necessarily mean reports are ignored, but it does suggest that user reporting alone is too slow or too inconsistent to function as an effective frontline defense against high-volume scam advertising.This becomes even more problematic when campaigns are designed to run briefly and aggressively. By the time an ad is reviewed manually, it may already have completed its useful life, reached a large audience, and been replaced by multiple variants.At this scale, platform moderation faces a difficult asymmetry. Scammers can generate and rotate ads faster than manual review can reasonably keep up.<h3>Advertisers accounts are cheap to replace</h3>Another structural issue is the ease of creating or replacing advertiser accounts.Even when accounts are blocked, scammers can often resume operations quickly by shifting to new accounts, new pages, or repurposed profiles. During our investigation, we observed large numbers of suspicious advertisers with limited page history, low follower counts, recently created pages, and naming patterns consistent with disposable setups.This creates a persistent regeneration advantage. If the cost of losing an advertiser account is low, and the cost of creating a replacement is also low, enforcement becomes a recurring operating expense for the scammer, not a meaningful deterrent.That is the Hydra problem in practical terms, remove one head, and the infrastructure behind it keeps producing more.<h3>Scam ads are engineered for resilience</h3>Scam advertising is not static. It is adaptive, distributed, and engineered for resilience.From URL presentation tricks and Unicode obfuscation to disposable ad waves and rapid advertiser regeneration, scammers exploit structural weaknesses in how social advertising ecosystems are built, moderated, and audited.Reactive takedowns matter, but they are not enough when enforcement is slower than replication.Meaningful progress requires systemic changes, stronger advertiser verification, better destination-level transparency, campaign-level enforcement, and faster cross-platform disruption mechanisms.The Hydra metaphor matters here because it captures the real failure mode. Removing individual ads without disrupting the regeneration model does not solve the problem, it manages the symptoms while the system keeps producing new heads.

How scam ads hide in plain sight, survive takedowns, and keep coming back

The Scam Ad Machine: Part II

Over a 23-day period, Gen Threat Labs analyzed 14.5 million ads running on Meta platforms across the EU and UK, representing more than 10.7 billion impressions. Nearly one in three of those ads (30.99%) pointed to a scam, phishing or malware link. In total, scam ads generated more than 300 million impressions in less than a month. The activity was highly concentrated, with just 10 advertisers responsible for over 56% of all observed scam ads. Repeated campaign clusters were traced to shared payment and infrastructure linked to China and Hong Kong, indicating organized, industrial-scale operations rather than isolated bad actors. Read on to learn about how this ecosystem operates, why existing enforcement fails to contain it and what the data reveals about who is driving scam advertising at scale.<h2>Advertising built to persuade</h2>Today, social advertising is being used to defraud at scale across some of the largest platforms.&nbsp;This is not about isolated policy failures or a handful of malicious advertisers slipping through moderation. It is about an advertising system that is structurally attractive to criminals and consistently delivers results for them.Across the internet, ads have quietly become one of the most efficient delivery mechanisms for scams, phishing and malware. Today, dangerous ads don’t look suspicious; they look professional, familiar and seem to target your exact needs. On social networks, the same optimization engines designed to maximize engagement and conversion are being repurposed to maximize victimization. And that’s not accidental.<h3>When advertising becomes the attack channel</h3>Most people still picture cyberattacks starting with a sketchy download or a suspicious email or text. Increasingly, they start with something far more normal: an ad. Malvertising (malicious advertising) has surged because it gives criminals what every marketer wants: instant reach, precise targeting and scale. Gen telemetry shows that malvertising has become the single largest threat to individuals, accounting for <a href="https://newsroom.gendigital.com/2026-01-20-Gen-Q4-Threat-Report-Shows-Scams-Thriving-Inside-Ads%2C-Feeds%2C-and-Video" target="_blank">41% of all cyberattacks</a>.&nbsp;Malvertising has become the top threat to consumers because they don’t look like they used to. People think they still look like loud pop-ups that scream, “I’m a virus,” or fake subscriptions. The modern model is quieter and more effective. It is a social engineering toolkit that blends into whatever people already trust and are currently paying attention to. We see this clearly in <a href="https://us.norton.com/blog/emerging-threats/threat-report-q3-2024" target="_blank">“scam-yourself” attacks</a> like FakeCaptcha and ClickFix, where victims are nudged into doing the attacker’s job for them, approving a browser prompt, enabling push notifications, copying and pasting commands or “verifying” something that feels routine. Browser push notifications in particular have become a reliable hook, as one click can turn a normal website visit into a persistent stream of scam prompts and redirects.Attackers also borrow directly from legitimate marketing. They ride trends, exploit urgency, and add credibility signals. That includes pairing malicious ads with deepfakes and timely themes, especially in investment and crypto scams. Gen researchers have <a href="https://www.gendigital.com/blog/insights/reports/threat-report-q1-2025" target="_blank">documented this extensively in CryptoCore campaigns</a>, where deepfake videos and hijacked accounts were used to promote fraudulent investment schemes at scale.&nbsp;Another persistent and successful tactic is impersonation. Cybercriminals buy ads on major ad networks (including search ads) to pose as trusted brands. The outcomes range from malicious redirects to phishing sites to drive-by downloads where malware is installed as part of the ad’s click path, often without the victim realizing what happened until later.&nbsp;And malvertising is not confined to shady corners of the web. Even the most trusted websites can unknowingly display malicious ads because ads are delivered through complex, automated supply chains. The net effect is that malicious advertising has become a fast, efficient way to scam and phish people. Gen research on social media platforms shows that malicious advertising comprises roughly <a href="https://us.norton.com/blog/malware/malvertising" target="_blank">30 % of scam incidents observed across social networks</a>, making it one of the most common threats users encounter in feeds and ads.&nbsp;If malvertising is now one of the primary ways criminals reach victims, then measuring what is happening inside major ad ecosystems is not just interesting; it is necessary.<h3>When online fraud metastasizes</h3>What we’ve discovered is not just growth, it’s an uncontrolled spread. What starts as an isolated scam tucked into an obscure corner of the web does not stay there. It spreads, mutates, and embeds itself into channels billions of people use every day. Online advertising, once a tool to connect consumers with products and services, has become part of the attack surface itself.Meta’s family of apps, including Facebook, Instagram, Threads, WhatsApp and Messenger, reaches an astonishing number of people. In Q2 2025, Meta reported that daily active users across its platforms reached about 3.48 billion worldwide. When a system that large becomes a gateway for scams, the impact is no longer isolated.Scam infrastructure does not operate in parallel to these platforms. It operates through them, leveraging their trust signals, engagement mechanics and targeting capabilities to propagate at speed.That is the context for what we measured.<h3>How we measured the problem</h3>Our Gen Threat Labs knew Meta ads were a problem and set out to measure the exact scale. Rather than chasing individual scam examples, we built a large-scale measurement pipeline around Meta’s Ad Transparency API, which provides visibility into ads that are currently active or were recently active on Meta platforms. Each day, we collected ads that were active or had recently been active, contained English ad text, and were delivered to users in regions covered by Meta’s Ad Transparency Library, primarily the EU and UK. While the ads themselves often promoted content intended for a global audience, our analysis focused on ads visible through the transparency framework rather than attempting to infer targeting outside of what the API explicitly exposes. We intentionally avoided narrowing the scope further in order to capture how scam advertising behaves at scale, not just in obvious edge cases.This work exists because regulation forced visibility. In the EU and UK, ad transparency requirements expose advertiser behavior in ways that cannot be replicated elsewhere, enabling independent measurement of scam advertising at an industrial scale. Where such regulation is absent, scam activity does not disappear; it simply becomes harder to measure. Transparency does not create abuse; it reveals it.For this analysis, we focused on consumer-facing ads and did not intentionally collect or analyze political advertising. Political ads are self-declared by advertisers, so while some may still appear in the dataset due to misclassification, they were not the target of this study.Even with these constraints, the dataset quickly became massive.<h3>14.5 million ads, billions of impressions, one pattern</h3>Over a 23-day period, we collected 14.57 million ads, representing 10.76 billion impressions delivered across the EU and UK. Even as a partial view of the ecosystem, the scale was impossible to ignore.Using intelligence and classification technology, we identified ads pointing to infrastructure associated with e-commerce scams, phishing campaigns, malware distribution and other consumer-facing threats. This allowed us to move beyond ad text and examine what users were actually being sent to.The results were not subtle. 4.51 million ads in our dataset were identified as scam-related, meaning nearly one in three ads (30.99%) pointed to scam infrastructure. In total, these scam ads generated 143.8 million impressions in the EU and 304.11 million impressions across the EU and UK in less than a month.This is not an edge case. It is an incredible volume.

<h3>Why takedowns don’t scale</h3>Volume alone does not explain why this problem persists. So, we looked at behavior over time.Across the dataset, we repeatedly observed scammers reusing the same infrastructure, identical domains, and near-identical ad text across many campaigns. In multiple cases, when one ad disappeared from the ad library for policy violations, other ads using the same domain and messaging remained active until their campaigns naturally expired.From the outside, it seems enforcement of scam ads is reactive. Ads appear to be removed one at a time, often following reports or reviews, while related campaigns using the same infrastructure continue running. Even when scam components are reused at scale, generalization appears slow, even if they are delivered by the same advertiser.We cannot see Meta’s internal detection signals, so we are careful not to assign intent. But the observable outcome is clear: known scam building blocks often remain usable long after one instance is taken down.At this scale, that distinction matters.

<h3>A small number of advertisers cause outsized harm</h3>Scale tells us how big the problem is. Enforcement shows us why it persists. But, to understand who is actually driving the harm, we asked a different question: are scam ads driven by countless small actors or by a smaller number of highly active criminal groups?The answer was clear.Scam advertising was also highly concentrated. The top 10 scam advertisers alone accounted for 56.1% of all scam ads, representing 2.53 million unique scam advertisements and 57.92 million impressions. A relatively small number of advertiser entities were responsible for a majority of observed harm.These advertisers are not anonymous hobbyists. They are organized, persistent operators running industrial-scale campaigns.We repeatedly traced clusters of campaigns back to payers and infrastructure linked to China and Hong Kong, operating fleets of short-lived pages created almost exclusively to run ads. Western brand names and English storefronts cycled rapidly, while payers and beneficiaries rotated or disappeared from disclosures altogether. What did not rotate were the mechanics: the same domains, URL patterns and ad behaviors resurfaced across supposedly unrelated advertisers.&nbsp;<h3>What comes next</h3>The numbers alone are unsettling. Millions of scam ads. Hundreds of millions of impressions. A small number of actors are driving disproportionate harm, repeatedly and at a scale.This isn’t about imperfect moderation. It’s about a system that, in practice, allows attackers to move faster than constraints can keep up. In our next blog post, we’ll move this research from measurement to mechanics. We break down the techniques scammers use to hide malicious ads in plain sight, including how they manipulate displayed URLs, chain redirects, blend malicious links with legitimate ones and exploit the limits of transparency tools themselves.&nbsp;Because understanding the size of the problem is only the first step. Understanding how it persists is what allows it to be disrupted.

Efe Karabeyli

Posts by Efe Karabeyli

The Scam Ad Machine: Part II

The Scam Ad Machine