Malware’s new masquerade: fake CAPTCHAs, AI c2, and the rise of ClickFix-style attacks

Malware delivery tactics have undergone a marked transformation, moving from the era of basic social engineering—think fake browser updates—to far more cunning technical obfuscation strategies. Campaigns like Smart Ape SG and Evolution illustrate this shift: attackers now deploy deceptive fake CAPTCHAs, such as the ClickFix exploit, to lure users into executing malicious PowerShell commands. These scripts not only install additional malware like Amatera Stealer and NetSupport RAT, but also represent a new breed of multi-stage attacks that blend psychological manipulation with technical sleight of hand.

As attackers have grown more sophisticated, so too have their evasion techniques. Rather than relying on obvious indicators like the sleep function, malware such as Formbook now mimics delays through subtle scripting loops and layers PowerShell obfuscation to slip past signature-based defenses. This technical evolution is paralleled by the deployment of selective phishing kits, which continue to harvest sensitive credentials while leveraging these advanced methods to maximize stealth and impact.

By early 2026, the landscape had shifted decisively toward multi-stage, technically obfuscated web attacks. Modern campaigns, as documented in the SmartApe case study, now employ intricate JavaScript injections, stable redirect chains, and traffic-distribution chokepoints to selectively deliver payloads and evade detection. While these tactics raise the bar for defenders, they also create new opportunities: as researchers note, monitoring these chokepoints at scale can turn the attackers’ own infrastructure into high-confidence detection and tracking assets.

Attackers have increasingly turned to abusing legitimate digital infrastructure, such as ad networks and compromised WordPress sites, to deliver malware at scale while evading detection and complicating attribution. Notably, campaigns like GootLoader have leveraged SEO poisoning and malvertising to funnel unsuspecting users searching for legal templates to booby-trapped WordPress sites, where each malicious ZIP archive is uniquely crafted using 'hashbusting'—randomizing archive fields and concatenating hundreds of files—rendering traditional hash-based detection futile. This evolution in delivery tactics demonstrates how attackers exploit the very systems meant to drive web traffic, transforming everyday digital infrastructure into a stealthy malware distribution network.

The exploitation of typosquatting and parked domains has matured from a nuisance into a sophisticated ecosystem for malware delivery and fraud, with attackers registering lookalike domains of trusted brands—such as NordVPN and ExpressVPN—where 14% of over 980 identified fake VPN domains were found to be malicious. By 2026, parked domains, once considered digital dead ends, had become vectors for scams, malware, and business email compromise, often hidden behind layers of traffic distribution, device fingerprinting, and advanced cloaking techniques that serve decoy content to security scanners while targeting real victims with malicious payloads. This shift is further complicated by the resale of traffic through affiliate advertising networks, where robust KYC checks at large parking platforms are undermined by downstream anonymity, allowing malicious actors to thrive in the shadows.

Malvertising has evolved into a highly professionalized arm of cybercrime, with platforms like 1Campaign enabling attackers to run malicious Google Ads while evading detection through sophisticated cloaking, analytics, and visitor profiling. These platforms, active since at least 2023, empower cybercriminals to extend the lifespan of fraudulent ad campaigns by showing phishing or scam content only to targeted victims, while security reviewers and automated scanners see innocuous blank pages. The integration of direct file distribution, fraud scoring, and large-scale brand spoofing into these campaigns underscores how the abuse of ad tech infrastructure has become both scalable and resilient, blurring the lines between legitimate digital marketing and organized cybercrime.

The meteoric rise of ClickFix-style attacks has fundamentally reshaped the malware landscape, with campaigns like Smart Ape SG and Evolution shifting tactics to exploit users through fake captchas and deceptive prompts. By manipulating victims into executing seemingly benign PowerShell commands—often under the guise of routine security checks or software installations—these campaigns have made social engineering the linchpin of modern malware delivery. As Huntress Labs reported, over half of all malware loader activity in 2025 stemmed from ClickFix attacks, underscoring their dominance and the alarming ease with which attackers turn unsuspecting users into unwitting accomplices.

What sets ClickFix attacks apart is their reliance on increasingly complex, multi-stage infection chains that leverage legitimate tools and shared infrastructure to evade detection. Attackers routinely compromise websites to inject malicious JavaScript, redirecting users into a labyrinth of payloads delivered via living-off-the-land binaries like MSHTA and PowerShell, and culminating in advanced obfuscation techniques such as steganography—embedding shellcode within innocuous PNG images. These multi-layered chains, as seen in campaigns deploying Amatera Stealer, NetSupport RAT, and MIMICRAT, not only target credentials and crypto wallets across both Windows and macOS, but also frustrate analysts and EDR systems with in-memory execution, modular loaders, and traffic patterns that mimic legitimate web analytics.

The adaptability and global reach of ClickFix-style campaigns have become increasingly evident, with attackers exploiting current trends—such as fake AI browser installers mimicking ChatGPT Atlas—and localizing lures in over 17 languages to ensnare victims worldwide. Microsoft and Elastic Security Labs have both warned of infostealer malware rapidly expanding beyond traditional Windows targets, now leveraging compromised legitimate sites, malicious ads, and shared infrastructure like Steam Workshop and Discord to propagate across platforms. This relentless innovation persists even during periods of expected downtime, such as the Thanksgiving holidays, highlighting both the sophistication and persistence of modern threat actors.

Despite the overwhelming prevalence and sophistication of ClickFix and multi-stage malware chains, a critical gap remains in organizational defenses: security awareness. Many employees, even those with security training, report never having been warned about ClickFix tactics, leaving them vulnerable to attacks masquerading as routine tasks like solving captchas or installing software. This lag between the rapid evolution of attacker methods and the slow pace of security training updates continues to fuel the success of ClickFix campaigns, as attackers exploit both technological and human blind spots.

By early 2026, threat actors had begun weaponizing both blockchain and generative AI assistants to create a new breed of evasive malware delivery channels. Campaigns such as Ether hiding leverage blockchain smart contracts—immutable and transparent by design—to host payloads and redirect chains, making takedown efforts nearly futile compared to traditional web infrastructure. At the same time, security researchers at Check Point revealed that generative AI assistants like Microsoft Copilot and xAI Grok are being subverted as stealthy command-and-control proxies, allowing malware to blend its traffic with legitimate queries and sidestep conventional detection mechanisms—no API keys or accounts required.

The democratization of AI tools such as ChatGPT and Claude has dramatically lowered the technical barrier for launching sophisticated malware attacks, fueling a surge in copycat threat actors who can automate and iterate on attack techniques at unprecedented speed. This proliferation is evident in recent campaigns where AI-generated exploit frameworks are paired with advanced evasion tactics, like the time-based logic bombs seen in the XMRig miner uncovered by Trellix. By integrating modular, multi-stage infections and self-destruct mechanisms, these campaigns maximize operational stealth and persistence, underscoring how AI is not just a tool for defenders but a potent force multiplier for attackers.

Next-generation evasion and delivery methods are also exemplified by the resurgence of worm-like propagation, as seen in the XMRig miner campaign that exploited the WinRing0x64.sys driver (CVE-2020-14979) for privilege escalation. By employing techniques like BYOVD (Bring Your Own Vulnerable Driver) and spreading via external storage devices, malware can now achieve lateral movement and resilience even in air-gapped environments. This evolution signals a shift from purely network-based attacks to hybrid approaches that combine software vulnerabilities, physical vectors, and AI-driven automation for maximum reach and persistence.

As web-based malware campaigns have grown more sophisticated, defenders now contend with multi-stage JavaScript injections, deceptive redirects, and fake CAPTCHAs that selectively deliver payloads while evading traditional detection. However, these complex attack chains are not without their Achilles' heel: they rely on stable redirect and traffic-distribution chokepoints, as illustrated by the SmartApe campaign. By monitoring these chokepoints at scale, defenders can transform attackers’ evasion tactics into high-confidence detection and tracking opportunities, turning the tables in the ongoing cat-and-mouse game.

The rapid pivot from traditional malware delivery to web-based attacks—such as the notorious ClickFix exploit, which drove 53% of malware loader activity in 2025—has exposed critical gaps in both technical defenses and user education. Attackers now exploit familiar security interactions, like fake CAPTCHAs and update prompts, to condition users into executing malicious commands, often bypassing even well-trained employees. This evolving threat landscape demands adaptive strategies: defenders must establish behavioral baselines, implement network policies (such as forcing JavaScript files to open in benign editors), and urgently update security training to address emerging social engineering techniques before they become widespread.

Infrastructure tracking has emerged as a crucial defensive tactic, especially as attackers increasingly leverage parked domains and traffic distribution systems to cloak their operations. Security teams are advised to block parked domains—long considered benign but now recognized as high-risk—and to investigate intermittent, unreproducible user warnings as potential signs of cloaking or traffic distribution. This has even sparked discussion of a market opportunity for tools that disguise an organization’s internet presence as that of a security researcher, thereby deterring malware delivery attempts.