Fibergrid has operated as a bulletproof hosting provider for nearly a decade, currently hosting 16,700 active fraudulent e-commerce sites. The network exploits stolen African IPv4 address space worth $20-25 million, originally acquired through improper AFRINIC registrations. Despite claiming Seychelles-based operations, multilateration analysis reveals infrastructure concentrated in the United States, United Kingdom, Netherlands, Canada, and other Western countries, primarily within Equinix data centers. Fibergrid operates through a complex web of UK and Estonian shell companies using multiple autonomous systems to evade detection and enforcement. Fake shops constitute 70% of malicious activity on this infrastructure, targeting consumers through search engines and social media with counterfeit goods and payment fraud schemes. Disruption opportunities exist through upstream provider intervention, regional internet registry action, domain-level takedowns, and indicator sharing with security providers.
Since February 2026, multiple incidents involving data theft and extortion have been attributed to activity cluster CL-CRI-1116, also known as BlackFile, UNC6671, and Cordial Spider. These financially-motivated attackers, likely associated with "The Com" collective, employ voice-based phishing combined with credential harvesting through fraudulent login pages. They impersonate IT support staff to steal credentials and bypass multi-factor authentication. The attackers focus on Living Off the Land techniques, abusing legitimate APIs like Microsoft Graph to access SharePoint sites and Salesforce data. They search for confidential information and employee data within SaaS environments, then exfiltrate it through browser downloads or API exports. To pressure victims into paying seven-figure ransoms, attackers send demands via Gmail and compromised email accounts, sometimes employing SWATting tactics against executives.
Xinference, an open-source distributed AI model inference framework, suffered a supply chain attack when attackers compromised PyPI release credentials of maintainers and published three malicious versions (2.6.0, 2.6.1, 2.6.2) on April 22, 2026. The malicious code, encoded in Base64 layers within __init__.py, executes automatically upon library installation or import, collecting cloud credentials, SSH keys, API tokens, database passwords, cryptocurrency wallets, and environment variables. The payload specifically targets AWS environments through metadata service exploitation and uploads stolen data to attacker-controlled infrastructure. The attack affects users who downloaded these versions from PyPI, which has over 680,000 total downloads. Attribution remains unclear as TeamPCP's name appears in the code but the group denies involvement, suggesting third-party impersonation.
A malicious Chrome extension impersonating Google's Authenticator application has been identified as part of an ongoing campaign active since early 2026. The extension requests excessive permissions and contains dormant infrastructure suggesting a staged deployment model where malicious updates can be delivered without requiring further user approval. This extension is linked to at least six others through a shared developer front, with two already carrying fully operational malicious payloads. These extensions utilize hidden iframes to inject attacker-controlled content, deploy fraudulent paywalls for free services, and maintain bidirectional communication with command and control servers. The infrastructure maps directly to the AiFrame campaign, which has reportedly compromised over 260,000 users from 2025 to present, marking a continued evolution of this threat.
Threat actors are leveraging fake CAPTCHA pages to trick victims into sending premium SMS messages as part of an international revenue share fraud (IRSF) scheme. Operating since at least June 2020, this campaign uses traffic distribution systems and social engineering to direct users through multi-stage fake verifications requiring SMS messages to international phone numbers across 17 countries with high termination fees. Each CAPTCHA step triggers messages to over a dozen destinations, generating over 60 SMS messages per victim costing approximately $30. The operation employs back button hijacking, sophisticated tracking cookies, and affiliate advertising networks to maximize reach while obscuring the fraud from detection. Both individual victims and telecommunication carriers suffer financial losses through this deceptive scheme.
X-Labs researchers discovered 10 verified Indirect Prompt Injection (IPI) payloads deployed across live web infrastructure. Unlike direct prompt injection where users send malicious input to AI models, IPI hides adversarial instructions inside ordinary web content. When AI agents crawl or summarize poisoned pages, they ingest and execute these instructions as legitimate commands. The discovered payloads span financial fraud, data destruction, API key exfiltration, and denial-of-service attacks. Attackers employ techniques including CSS invisibility, HTML comments, accessibility attribute abuse, meta namespace spoofing, and system prompt tag impersonation. The shared injection templates across multiple domains suggest organized tooling rather than isolated experimentation. Observed attack intents include unauthorized financial transactions, terminal command execution, content suppression, traffic hijacking, and sensitive information leakage, targeting AI systems that browse web pages, index content for RAG ...
Investigation of DPRK-linked fake IT worker infrastructure began after cryptocurrency researcher ZachXBT identified domain luckyguys[.]site connected to illicit payments. Analysis of 30 days of network activity associated with IP 163.245.219[.]19 revealed concentrated VPN usage patterns, with Astrill VPN (37.5%), Mullvad (32.25%), and Proton VPN (6.25%) being prominent. American and Latvian residential IPs communicated with the infrastructure, showing frequent Astrill VPN usage and connectivity to Gmail, ChatGPT, and Workana freelance platform. A second IP, 216.158.225[.]144, was discovered through X509 certificate analysis. Traffic dropped sharply following public exposure, consistent with adversary behavior of abandoning attributed infrastructure. The activity suggests a distributed network of remote IT workers participating in sanctions evasion workflows, leveraging AI tools and freelance platforms to obtain employment under false identities.
This analysis documents a newly observed ClickFix variant that abuses native Windows utilities, specifically cmdkey and regsvr32, for payload delivery. Victims are socially engineered through fake CAPTCHA pages to execute a malicious command via the Windows Run dialog. The single command chains multiple actions: staging credentials using cmdkey, retrieving a remote DLL via regsvr32 from a UNC path, and executing it silently. The 64-bit DLL establishes persistence through a scheduled task pulled from a remote XML file hosted on attacker infrastructure. This approach avoids traditional malware drops and leverages exclusively trusted Windows components for high stealth. The variant demonstrates continued evolution of ClickFix techniques, moving beyond PowerShell to use command chaining with legitimate system tools.
StepDrainer is a Malware-as-a-Service (MaaS) platform engineered to steal digital assets from cryptocurrency wallets, including fungible tokens and high-value NFT collections. The malware supports more than 20 blockchain networks and incorporates multiple draining techniques, particularly abusing ERC-20 token permissions and NFT approval mechanisms.
The platform includes automated asset transfer capabilities, compatibility with widely used mobile wallets, and encrypted logging via Telegram channels for attacker monitoring. StepDrainer is commercially distributed within cybercriminal ecosystems, with pricing models ranging from approximately $750 for full source code access to $150 for a shared version that imposes a 20% commission on successful thefts.
A sophisticated ClickFix campaign targets both Windows and macOS users through fake CAPTCHA pages that trick victims into executing malicious commands. The macOS variant deploys an AppleScript-based infostealer that harvests sensitive data including keychain databases, credentials, and session cookies from 12 browsers, over 200 browser extensions, and 16 cryptocurrency wallets. The malware employs a persistent, non-closable dialog box mimicking legitimate system prompts to force victims into providing their system password. Stolen session cookies enable attackers to bypass multi-factor authentication by hijacking active sessions. The campaign uses client-side JavaScript to filter victims by user-agent, directing desktop users to OS-specific payloads while ignoring mobile devices. Latest macOS updates include native terminal security warnings designed to alert users against pasting potentially malicious commands.
In early 2026, phishing attacks remain a top threat vector in security operations. This analysis covers a novel attack method exploiting Microsoft's OAuth 2.0 Device Authorization Grant (Device Code Flow) to compromise user accounts. Attackers use phishing emails containing Mailchimp's Mandrill service links to bypass security controls, leading victims to fake Adobe-themed websites. The sites abuse legitimate Microsoft authentication mechanisms to obtain access and refresh tokens, granting persistent delegated access to critical resources like Graph API, Teams, Outlook, and SharePoint. The technique leverages shared client IDs across tenants and family of client IDs (FOCI) for lateral movement. Two variants exist: one using external phishing infrastructure with dynamic code generation, and another relying solely on fake meeting invitations containing pre-generated device codes. The attack is particularly effective as it uses legitimate Microsoft services, making detection challenging.
IOCs related to FlowerStorm phishing‑kit–driven campaign that delivers fake Microsoft authentication pages via compromised domains fronted by Cloudflare. The activity abuses legitimate cloud and CDN services for delivery while credential harvesting occurs on attacker‑controlled infrastructure, with incidental contact to Microsoft services during normal browser behavior.that uses its own web servers to target victims' login credentials and access to their personal details and login details on its servers.
A highly sophisticated multi-stage post-exploitation framework targeting organizations in the Middle East and EMEA financial sectors exploits legitimate digitally signed Intel utilities through .NET AppDomainManager mechanism abuse. The attack leverages trusted binary proxy execution, bypassing EDR and antivirus solutions through JIT-based memory execution and sandbox evasion using computational delays and cryptographic key derivation loops. Initial access occurs via spear-phishing with Arabic-language decoys impersonating Saudi government documents. Once executed, the framework establishes command-and-control communication through Amazon CloudFront CDN domain fronting, employing reflective DLL loading, direct syscall usage, and anti-forensic memory cleanup techniques. The modular plugin-based architecture demonstrates capabilities consistent with advanced persistent threat actors, featuring sophisticated evasion mechanisms including PEB-based API resolution, custom PE export walking, and heap-walking cont...
A compromised Joomla website displayed suspicious product links unrelated to the business. Investigation revealed heavily obfuscated PHP code injected at the top of index.php that contacted external command-and-control servers to receive instructions and manipulate content. The malware acts as a remote loader, assembling strings from two-character chunks to evade signature-based detection. It contacts primary C2 cdn.erpsaz.com and fallback cdn.saholerp.com, sending server fingerprint data and receiving dynamic instructions. Based on responses, it redirects visitors, injects spam content, or serves fake SEO pages to search engines. This approach allows attackers to control compromised sites remotely without modifying local files again, enabling dynamic spam injection, visitor redirection, and search engine manipulation while remaining undetected for extended periods.
A cargo theft threat actor maintained access to a decoy environment for over a month, providing extensive visibility into post-compromise operations. The attacker established redundant persistence using multiple remote access tools, including four ScreenConnect instances, Pulseway RMM, and SimpleHelp RMM. A previously unknown signing-as-a-service capability was employed to evade detection by re-signing ScreenConnect installers with fraudulent code-signing certificates. Extensive reconnaissance targeted financial platforms, payment systems, cryptocurrency wallets, and transportation-specific services including fuel card providers, fleet payment platforms, and load board operators. The activity strongly aligns with financially motivated crimes against the transportation industry, including freight diversion and cargo theft operations.
An active campaign targets Trust Wallet users through malicious QR codes distributed via Telegram, exploiting deep link mechanisms to redirect victims to Netlify-hosted phishing domains. The attack masquerades as a legitimate USDT transfer interface but covertly triggers an ERC-20 approve() transaction, granting unlimited token allowance to an attacker-controlled contract on BNB Smart Chain. This enables persistent fund drainage without further victim interaction. The modular drainer architecture uses config.js for control parameters and main.js for execution logic, with integrated Telegram bot infrastructure providing real-time transaction monitoring. Analysis confirms 52 transaction notifications indicating active exploitation. The campaign employs social engineering through a deceptive dollar-one illusion where victims believe they are initiating small transactions while actually granting unlimited wallet access. Multiple cloned phishing domains demonstrate scalable deployment within a Drainer-as-a-Servic
A sophisticated phishing campaign is targeting YouTube creators using convincing fake copyright strike notifications. The attack dynamically pulls real channel data including profile pictures, subscriber counts, and recent videos to create personalized scare pages. Victims are funneled through a Browser-in-the-Browser attack displaying a fake Google sign-in that captures credentials. The operation functions as phishing-as-a-service, with multiple attackers sharing infrastructure and rotating domains to evade detection. Successful attacks result in complete Google account takeover, allowing hijackers to rebrand channels and livestream cryptocurrency scams to existing audiences. The kit automatically exempts channels with over three million subscribers to avoid detection by security teams.
In spring 2022, a highly sophisticated backdoor named AngrySpark was discovered on a single machine in the United Kingdom. The malware employed a three-stage architecture: a DLL masquerading as a Windows Task Scheduler component, a custom virtual machine interpreter running bytecode instructions, and a beacon that profiles systems while disguising C2 communications as PNG image requests. The malware featured VM-based obfuscation, dual encrypted C2 channels using RSA-4096 and XXTEA encryption, direct syscalls bypassing usermode hooks, hypervisor detection, and CET-aware anti-analysis capabilities. It operated for approximately one year with active maintenance visible through syscall table updates and configuration changes between May 2022 and January 2023. The infrastructure expired in mid-2023 and the operation ceased, with no additional samples or victims identified despite the significant engineering effort invested in its development.
A coordinated campaign of 108 malicious Chrome extensions operated through shared command-and-control infrastructure at cloudapi[.]stream has been identified, collectively accounting for approximately 20,000 installations. The campaign spans multiple threat categories: 54 extensions steal Google account identities via OAuth2, one extension actively exfiltrates Telegram Web sessions every 15 seconds, and 45 extensions contain a universal backdoor enabling arbitrary URL execution on browser startup. Published under five distinct publisher identities (Yana Project, GameGen, SideGames, Rodeo Games, and InterAlt), these extensions masquerade as legitimate tools including Telegram sidebar clients, slot games, YouTube and TikTok enhancers, and translation utilities. All extensions route stolen credentials, user identities, and browsing data to servers controlled by the same operator, with infrastructure confirming a Malware-as-a-Service business model.
On April 7, 2026, a security researcher described an Adobe Reader zero-day vulnerability that has been exploited since at least December 2025. The vulnerability allows threat actors to execute privileged Acrobat APIs via specially crafted malicious PDF files that execute obfuscated JavaScript when opened. Exploitation allows attackers to steal sensitive user and system data and to potentially launch additional attacks and remotely execute code. Recommendations: Reduce the risk by automatically scanning PDF email attachments, blocking suspicious files, training users to be wary of unsolicited attachments, and advising users to temporarily avoid using Adobe Reader to open PDFs. Reference: https://www.sophos.com/en-us/blog/adobe-reader-zero-day-vulnerability-in-active-exploitation
