The BabaDeda loader family has undergone significant advancements in its capabilities, particularly in stealth, evasion, and payload flexibility. Discovered during April 2026, this evolved framework continues to conceal malicious payloads within seemingly legitimate installer packages while expanding its functionality. The attack methodology begins with a social engineering exploit known as ClickFix, which encourages users to execute commands via trusted operating system utilities. This initial step transitions into a sophisticated multi-stage loader that employs several tactics, including hidden PowerShell commands, in-memory shellcode, DLL sideloading, and external payload storage.
Kaspersky uncovered a sophisticated phishing campaign by the APT group Armored Likho, deploying a previously undocumented Python-based infostealer dubbed BusySnake Stealer. The campaign targets government agencies and electric power sectors across Russia, Brazil, and Kazakhstan through spear-phishing emails containing malicious EXE or LNK attachments. BusySnake Stealer features advanced obfuscation using PyArmor Pro, extracts credentials from browsers using DPAPI and NSS libraries, captures screenshots, logs keystrokes, scrapes cryptocurrency wallets and 2FA tokens, and establishes reverse SSH tunnels for remote access. The threat actor leverages AI-generated code for first-stage payloads and distributes components via GitHub repositories. The stealer maintains persistence through scheduled tasks and communicates with C2 infrastructure to receive commands dynamically, representing a significant evolution in the group's technical capabilities.
Children are targeted by a sprawling ecosystem of websites exploiting their interest in Roblox and Minecraft through offerwall reward schemes and phishing campaigns. These sites promise free in-game currency in exchange for completing tasks, collecting personal data, enrolling minors in paid subscriptions, and violating platform terms of service that can result in account bans. The infrastructure relies on cheap, disposable hosting with aggressive domain rotation. Using Internet-wide scan data from Censys, this analysis characterizes two categories: offerwall get-paid-to reward sites and credential harvesting generators. The exposed infrastructure handles children's data with minimal security, monetizing their attention at scale through affiliate commissions while presenting significant privacy and security risks.
PamStealer is a two-stage macOS infostealer distributed as a compiled AppleScript impersonating Maccy, a legitimate clipboard manager, hosted on a fake domain. The first stage uses JavaScript for Automation with Objective-C APIs to download payloads while avoiding shell commands. The second stage is a Rust-based Mach-O binary that validates stolen credentials through PAM before harvesting, reads browser databases directly using bundled SQLite, captures clipboard contents repeatedly via pbpaste, and exfiltrates encrypted data using ChaCha20-Poly1305. It establishes persistence through both modern and legacy login item APIs, masquerades as Finder or System Settings, and tricks victims into granting Full Disk Access through counterfeit alerts. The stealer contacts Ethereum RPC endpoints and employs region-based exclusions targeting Apple silicon systems while avoiding Commonwealth of Independent States countries.
A widespread phishing campaign distributing AsyncRAT and Remcos RATs has been observed targeting organizations across manufacturing, media, professional services, agriculture, and chemical industries globally. The attack leverages malicious Excel spreadsheets sent via emails impersonating business communications like purchase orders and payment advice. When macros are enabled, VBA code retrieves HTA payloads through URL shorteners and Cloudflare Workers infrastructure. The multi-stage infection chain employs heavy obfuscation including Base64 encoding, steganography in PNG files, and character substitution. The campaign intensified during June 2026, affecting organizations across Europe, Asia-Pacific, and the Americas. Infrastructure includes distinctive HTA naming conventions using concatenated positive English words. The operation likely uses automation for payload generation and may leverage LLMs for development efficiency.
Scam advertising campaigns have been identified that impersonate trusted brands to redirect consumers to unrelated online gambling sites. These operations utilize paid social media advertisements on platforms like Facebook, Instagram, and TikTok, combined with fake app store pages and Progressive Web Apps. The campaigns target UK consumers primarily, with variants observed in German and Spanish. Scammers impersonate major brands including financial institutions like Monzo, Revolut, and Barclays, as well as household names such as Tesco, Amazon, Netflix, and Facebook. The scheme involves three stages: paid ads claiming brands have launched official casino products, fake landing pages mimicking app stores, and PWAs that redirect to gambling sites through affiliate tracking links. Typical affiliate payouts range from $50 to $350 per depositing player, indicating significant financial motivation behind these operations.
A sophisticated multi-stage phishing campaign delivers a previously undocumented framework called Avalon through spoofed legal documents hosted on Proton Drive. The intrusion begins with password-protected archives containing ISO images that execute malicious MSBuild projects, loading payloads entirely in memory without conventional executable attachments. Avalon consolidates credential theft, lateral movement, recovery disruption, and ransomware capabilities within a single framework, with its encryption component branded as CrownX. The framework demonstrates hallmarks of AI-assisted development, rapidly combining multiple post-exploitation capabilities that previously required sustained development effort. Avalon targets browsers, cryptocurrency wallets, messaging platforms, VPN configurations, and infrastructure systems while implementing extensive defense evasion techniques against major security products. The framework disrupts recovery by eliminating Volume Shadow Copies, Windows Recovery Environment...
AI agents are increasingly vulnerable to indirect prompt injection (IPI) attacks, where malicious instructions are embedded in web content to manipulate AI-driven workflows. Two campaigns were identified that combine SEO poisoning with CSS/HTML abuse to influence AI decision-making. The first campaign uses fake API documentation to trick AI agents into making fraudulent payments for a fake Python library, incorporating hidden instructions in JSON-LD and CSS-concealed content directing payment of $3.00 via Stripe or approximately 0.0012 ETH to attacker wallets. The second campaign employs typosquatting to impersonate DeBank, a cryptocurrency portfolio tracker, embedding hidden prompts to make the fraudulent site appear as an authoritative source. Testing across 26 LLMs revealed 4 models were vulnerable to the payment scam and 2 models misclassified the typosquatting site, demonstrating measurable real-world impact.
An investigation beginning with a single RedLine Stealer C2 server from VMRay UniqueSignal evolved into uncovering a targeted Business Email Compromise campaign against South Korean maritime infrastructure. The analysis started with IP 194.156.79.122 on port 55615, leveraging fingerprinting techniques through FOFA and VirusTotal to identify additional C2 infrastructure. Pivoting through communicating files revealed spear-phishing emails targeting Kangrim Heavy Industries, a major South Korean marine boiler manufacturer. The campaign delivered Formbook malware through impersonated maritime supply chain companies. Further infrastructure analysis identified seven fraudulent domains hosted on TheHost LLC infrastructure, utilizing similar naming patterns and TLS certificates. The attack demonstrates sophisticated BEC tactics combining malware delivery with social engineering, mimicking legitimate business correspondence within the maritime shipping sector.
Blacksite is a newly identified adversary-in-the-middle phishing-as-a-service offering sold alongside Cloaked.gg, a cloaking platform that conceals phishing infrastructure from automated security analysis. The kit operates as a reverse-proxy that intercepts authentication tokens, session cookies, and 2FA codes in real time, enabling full account takeover even against MFA-protected accounts. Cloaked.gg blocks traffic from AWS, Google Cloud, and Azure networks while serving AI-generated decoy pages to suspected scanners, making malicious URLs appear benign during automated analysis. Priced between $600-$1,000 monthly, the service commercializes sophisticated AiTM techniques, lowering technical barriers for attackers. The pairing of credential theft capabilities with anti-detection infrastructure creates a split-view environment where security tools see harmless content while intended victims are routed to live phishing pages targeting consumer, financial, and enterprise identity systems.
An active SMS phishing campaign targets Serbian road users by impersonating Putevi Srbije, Serbia's state road authority. Victims receive text messages claiming they have unpaid traffic fines with urgent payment demands. The fraudulent links lead to cloned government websites designed to steal payment card details. The infrastructure employs JavaScript-based obfuscation techniques to evade automated security scanners and uses disposable domains with uncommon TLDs. Technical analysis reveals connections to both Darcula and Phoenix Phishing-as-a-Service platforms, indicating fraudsters are combining tools from multiple PhaaS vendors. The operation demonstrates coordinated roles including infrastructure setup, SMS distribution, and data harvesting. Similar campaigns have targeted victims globally across government bodies, postal services, and financial institutions.
The ToddyCat APT group developed a sophisticated tool called Umbrij to compromise Gmail corporate accounts through OAuth token theft. The malware exploits Chromium-based browsers by launching them in headless mode with remote debugging enabled, utilizing the Shadow Token via Remote Debug (STRD) technique. Umbrij automates the entire attack chain: it copies user profiles, launches browsers with debugging ports, connects via Puppeteer Sharp library, and manipulates OAuth flows by impersonating legitimate Google Workspace migration tools. The tool specifically targets client IDs for Google Workspace Migration for Microsoft Outlook and Google Workspace Sync applications, requesting extensive permissions for email, calendar, drive, and contacts. ToddyCat deploys Umbrij through DLL sideloading techniques using signed files from Bitdefender, Visual Studio, and Google Desktop Search. This automated approach enables scalable compromise of organizational email communications while evading traditional security monito...
Mustang Panda orchestrated two concurrent espionage campaigns targeting Indian government entities and hydropower infrastructure between May and June 2026. The campaigns leveraged DLL sideloading via legitimate executables to deploy newly identified malware including SHARDLOADER, MINIRECON, and ZOHOMURK. MINIRECON represents an evolution of Toneshell with WebSocket-based command-and-control capabilities, while ZOHOMURK abuses Zoho WorkDrive cloud services for C2 communications and data exfiltration. Distribution occurred through spear-phishing with lures themed around India-Taiwan cooperation agreements and hydropower projects. The activity demonstrates code overlaps with previous tooling, infrastructure proximity to known operations, and targeting patterns aligned with Chinese strategic intelligence collection priorities. Multiple compromised government systems were identified, with coordination conducted through CERT-In for victim notification and remediation.
In July 2025, threat actors compromised organizations through SEO poisoning campaigns targeting users searching for legitimate IT management tools. Users downloading trojanized installers for ManageEngine OpManager received Bumblebee malware, granting initial access. The attackers exploited the fact that users executing these IT tools were privileged administrators, enabling rapid lateral movement to domain controllers. They dumped credentials using wbadmin, created backdoor accounts with enterprise admin privileges, and installed RustDesk for persistent access. AdaptixC2 beacons were deployed for command and control. The threat actors conducted extensive reconnaissance, dumped LSASS memory across multiple systems, attempted Veeam credential theft, and exfiltrated data via SFTP using FileZilla. The intrusion culminated in Akira ransomware deployment across both root and child domains within 44 hours, with subsequent re-encryption two days later affecting the child domain.
The Gentlemen ransomware-as-a-service group emerged as a top-10 threat actor in the first half of 2026. The group exploits vulnerabilities in internet-facing devices like VPNs and firewalls, potentially collaborating with initial access brokers. They employ comprehensive reconnaissance using tools like SharpADWS, NetScan, and Advanced IP Scanner, capturing network traffic with netsh. The attackers disable security products through BYOVD techniques using vulnerable drivers, and deploy custom Go-based backdoors and ransomware variants. They spread laterally via GPO deployment and PsExec, encrypt files using Curve25519 and XChaCha20, and recently developed a C-based ransomware variant using AES256-GCM and RSA. The group targets multiple industries worldwide, particularly in Brazil, China, Indonesia, Taiwan, and Thailand, with attacks focusing on manufacturing, IT services, healthcare, and financial sectors.
A fresh wave of the Miasma Mini Shai-Hulud supply chain campaign compromised legitimate npm packages under the @immobiliarelabs scope on June 26, 2026. The attack targeted Backstage plugins used for GitLab integration and LDAP authentication, affecting 22 package versions across multiple releases. The malware employs sophisticated techniques including hidden payloads that bypass standard package reviews, steals developer credentials and CI/CD secrets, and exploits GitHub Actions workflows for propagation. The campaign appears linked to a compromised upstream GitHub Action (codfish/semantic-release-action) and leverages deployment-triggered workflows for execution. Stolen credentials include npm tokens, GitHub tokens, cloud credentials, SSH keys, and various authentication secrets, which are exfiltrated to attacker-controlled repositories for further propagation across the ecosystem.
A sophisticated China-aligned cyber espionage campaign targeting India's tax infrastructure was identified between May and June 2026. The operation impersonates the Income Tax Department, Ministry of Finance, exploiting the AY2026-27 ITR filing season to target corporate entities, tax professionals, chartered accountants, and taxpayers. The attack employs spear-phishing emails with malicious attachments mimicking legitimate government utilities. The multi-stage infection chain deploys DcRAT through steganographic payload concealment, fileless .NET execution, AMSI bypass, and Windows service persistence. The threat actor demonstrates operational maturity through active payload rotation achieving 0/66 detection rates, encrypted TLS-based C2 communications, and infrastructure hosted across multiple ASNs linked to China. The campaign shows overlaps with the China-nexus threat actor Silver Fox, featuring screen capture capabilities, data exfiltration, and systematic intelligence collection from high-value India...
A sophisticated supply chain attack campaign linked to Mini Shai-Hulud, Miasma, and Hades malware has compromised LeoPlatform npm packages, GitHub Actions workflows, and the Verana Blockchain Go module. The attack employs binding.gyp install-time execution, Bun-staged JavaScript malware, and encrypted credential exfiltration targeting developer and CI/CD environments. Malicious packages were published through the czirker and llxlr npm accounts in a coordinated burst on June 24, 2026. The campaign steals credentials including npm tokens, GitHub tokens, cloud provider credentials, SSH keys, and AI coding assistant configurations. Attackers use GitHub as dead-drop infrastructure and inject persistence hooks into repositories through orphan branches and fake dependency-update workflows. The RevokeAndItGoesKaboom marker connects this wave to the codfish/semantic-release-action compromise, indicating shared operational tooling.
Since April 2026, a sophisticated multi-stage intrusion campaign has targeted hospitality and hotel organizations across Europe and Asia. The operation uses photo-themed ZIP archives containing malicious shortcut files disguised as images. When executed, these shortcuts initiate an attack chain involving obfuscated PowerShell, Node.js-based implants, and dual registry persistence mechanisms. The threat actor exploits legitimate services like Calendly and Google redirects for phishing delivery, employing authentication laundering to bypass email security controls. The campaign evolved through two waves, introducing .NET DLL compilation, Cloudflare-fronted infrastructure, and refined obfuscation techniques. Post-compromise activities include command-and-control beaconing over non-standard ports, forced shutdowns, and portable executable compilation, suggesting preparation for additional malicious operations.
Throughout 2025, Chinese-speaking threat actors tracked as CL-STA-1062 conducted extensive operations against government entities and critical infrastructure in Southeast Asia, specifically targeting state-owned enterprises in energy and government sectors. Active since March 2022, this cluster was previously identified as UAT-7237 in campaigns against Taiwan's web hosting infrastructure. The attackers employ a hybrid toolkit combining open-source tools like SoftEther VPN, Mimikatz, and VNT with a newly discovered custom backdoor called TinyRCT. This .NET-based backdoor provides capabilities including arbitrary command execution, file enumeration and exfiltration, screen capture, and self-destruct mechanisms. The infection chain typically begins with web application exploitation deploying ASPX web shells, followed by credential dumping, lateral movement, and data exfiltration. Between October and December 2025, at least ten organizations across Southeast Asia were compromised, demonstrating sustained regio...
