Hello there 👋
Welcome back to the Kraven Security weekly newsletter, triaging the week. We round up the week’s top news stories, highlight our featured article, give you some learning resources, and finish with a few personal notes about what’s happening at the company. Enjoy!
Top News Stories
Malicious CPU-Z and HWMonitor Installers Distributed via Official CPUID Site
CPUID has fallen victim to a supply-chain attack, resulting in the distribution of malware-laden versions of its flagship tools, CPU-Z and HWMonitor. Threat actors hijacked the official website to swap legitimate installers for trojanized versions, deploying RedLine Stealer.
Key takeaways:
🎯 Target: Hardware enthusiasts, gamers, and system administrators who downloaded or updated CPU-Z or HWMonitor directly from the official CPUID website during the breach window.
💡 Insight: This incident underscores the “trust paradox” of modern software distribution; even when users follow best practices by avoiding third-party mirrors and going directly to the source, a compromised backend can turn a reputable vendor into an unwitting malware distributor.
☑️ Recommendation 1: Users who downloaded software from CPUID recently should immediately perform a full system scan with updated EDR or antivirus tools and change all stored browser credentials, as RedLine Stealer is highly effective at harvesting sensitive data.
☑️ Recommendation 2: Before executing administrative-level diagnostic tools, manually verify the file’s digital signature; legitimate CPUID installers are typically signed by “CPUID,” and any “Unknown Publisher” warning or mismatched certificate should be treated as a critical red flag.
Nearly 4,000 US Industrial Devices Exposed to Iranian APT Targeting
Censys reports that nearly 4,000 Rockwell Automation PLCs are exposed to the internet in the U.S. These industrial controllers are critical to the water and energy sectors, making them prime targets for Iranian-affiliated threat actors seeking to cause physical disruption.
Key takeaways:
🎯 Target: Critical infrastructure operators in the United States utilizing Rockwell Automation and Allen-Bradley PLCs that are directly accessible via the public internet.
💡 Insight: The scale of this exposure highlights a persistent “security through obscurity” fallacy; threat actors are increasingly using simple scanning tools to identify unshielded industrial assets, bypassing the need for complex exploits to gain control over physical systems.
☑️ Recommendation 1: (Short-term) Immediately disconnect all PLCs and industrial control systems from the public-facing internet and move them behind a secure firewall or a Virtual Private Network (VPN) protected by Multi-Factor Authentication (MFA).
☑️ Recommendation 2: (Short-term) Change all default administrative credentials and disable unused communication ports on all Allen-Bradley devices to prevent “low-effort” hijacking by opportunistic APT groups.
☑️ Recommendation 3: (Long-term) Implement a robust OT network segmentation strategy that enforces a strict “air-gap” or DMZ between IT and OT environments, ensuring that no industrial device is reachable without passing through multiple layers of authenticated security.
Adobe Fixes Zero-Day Under Active Attack
Adobe released an urgent update for a critical zero-day vulnerability (CVE-2023-26369) in Acrobat and Reader. The flaw allows arbitrary code execution via malicious PDFs and is currently being exploited.
Key takeaways
🎯 Target: Enterprise users and individuals on both Windows and macOS systems who utilize Adobe Acrobat or Reader for document management.
💡 Insight: What makes this dangerous is its “zero-day” status. Attackers were already exploiting this flaw to bypass security controls before a patch was available, highlighting the ongoing risk posed by file-based delivery methods in targeted campaigns.
☑️ Recommendation 1: (Immediate Action) Manually trigger an update in Adobe Acrobat or Reader by navigating to Help > Check for Updates to ensure you are running the latest version (23.006.20322 or higher).
☑️ Recommendation 2: (Short-term) Enforce “Protected View” via Group Policy or local settings to ensure PDFs from untrusted locations open in a restricted sandbox.
☑️ Recommendation 3: (Long-term) Evaluate your organization’s reliance on desktop-based PDF readers; where possible, transition to browser-based viewers or “Secure-by-Design” applications that offer superior isolation against memory-corruption exploits.
Rockstar Games Analytics Data Leaked Following Third-Party Breach
Rockstar Games has confirmed a security incident involving the theft of millions of business records, which the ShinyHunters extortion gang has now begun leaking online. The breach originated from a compromise of the third-party analytics platform Anodot, which provided the attackers a pathway into Rockstar’s Snowflake data environment.
Key takeaways
🎯 Target: Rockstar Games’ corporate analytics and business intelligence data, specifically involving Grand Theft Auto Online revenue metrics, player behavior tracking, and internal game economy data.
💡 Insight: This incident demonstrates a “double-hop” supply chain attack: by compromising a niche AI-powered analytics tool (Anodot), attackers bypassed direct defenses and exfiltrated massive datasets from a primary cloud data warehouse (Snowflake).
☑️ Recommendation 1: (Immediate Action) Conduct an audit of all third-party SaaS integrations and service accounts connected to your cloud data warehouses to ensure no unauthorized “shadow” integrations exist.
☑️ Recommendation 2: (Short-term) Enforce the Principle of Least Privilege for API keys and service tokens used by third-party tools, ensuring they have only “read-only” access to the specific datasets required for their functions.
☑️ Recommendation 3: (Long-term) Mature your Third-Party Risk Management (TPRM) program to include continuous technical monitoring of vendor security posture rather than relying solely on annual compliance questionnaires.
Cluster of 108 Malicious Chrome Extensions Exposed Stealing Google and Telegram Data
A huge campaign involving 108 malicious Chrome extensions has been harvesting sensitive user identities and hijacking sessions of over 20,000 users. These extensions masquerade as legitimate tools/games while exfiltrating credentials to a single shared backend infrastructure.
Key takeaways
🎯 Target: Individual users of Google and Telegram, particularly those using “gaming” utilities, YouTube/TikTok enhancers, and translation tools.
💡 Insight: The campaign utilizes a sophisticated “multi-publisher” strategy, operating under five distinct identities to bypass store reputation checks, and leverages Chrome’s declarativeNetRequest API to strip security headers from platforms like YouTube and TikTok, enabling ad injection and session theft every 15 seconds.
☑️ Recommendation 1: (Short-term) Audit your browser extensions immediately; remove any published by Yana Project, GameGen, SideGames, Rodeo Games, or InterAlt, and terminate all active Telegram Web sessions from your mobile app to invalidate stolen tokens.
☑️ Recommendation 2: (Long-term) Transition from a “How to” to a “How I” security mindset by documenting and auditing your digital footprint regularly, and enforcing the use of hardware-based 2FA (like YubiKeys), which are resistant to session-theft and OAuth2 abuse seen in this campaign.
AI-Driven “Pushpaganda” Scam Hijacks Google Discover to Spread Scareware
Researchers unmasked “Pushpaganda,” a massive ad fraud campaign weaponizing AI-generated content and SEO poisoning to infiltrate Google Discover. This operation coerces users into enabling browser notifications that deliver scareware and fake legal threats, generating millions in illicit revenue.
Key takeaways
🎯 Target: Android and Chrome users globally who rely on the Google Discover feed for news and personalized content.
💡 Insight: This campaign represents a shift toward “AI-velocity” fraud, in which threat actors use generative AI to mass-produce deceptive news stories that bypass traditional content filters and hijack trusted, high-traffic discovery platforms.
☑️ Recommendation 1: (Short-term) Audit your browser and mobile device settings to immediately disable notifications from unknown or suspicious websites, especially those that use alarming language or “legal notices” to grab attention.
☑️ Recommendation 2: (Long-term) Organizations should pivot their awareness training toward “algorithmic literacy,” teaching users that content appearing in trusted feeds like Google Discover or social media trends can still be weaponized via AI-driven misinformation and SEO manipulation.
Fake Ledger Live App Slips into Apple’s App Store, Draining $9.5M
A fraudulent Ledger Live app bypassed Apple’s App Store, stealing $9.5 million from over 50 users in one week. Victims lost significant savings by entering recovery phrases into the malicious app, which remained active for seven days.
Key takeaways
🎯 Target: High-net-worth cryptocurrency investors and retail users who rely on the Apple App Store’s reputation for security to manage their hardware wallets.
💡 Insight: This incident highlights a significant lapse in Apple’s “walled garden” security; the attackers utilized a complex laundering network involving over 150 KuCoin deposit addresses to rapidly move stolen funds across Bitcoin, Ethereum, and Solana blockchains.
☑️ Recommendation 1: (Short-term) If you installed “Ledger Live” from the App Store between April 7 and April 13, 2026, assume your seed phrase is compromised; immediately move all remaining funds to a new wallet generated with a completely fresh recovery phrase.
☑️ Recommendation 2: (Long-term) Never type your 24-word recovery phrase into any digital device or app; enter it directly on your physical hardware device. Always download wallet management software exclusively from the manufacturer’s official website rather than searching third-party app stores.
Microsoft Hardens Windows Against Malicious RDP Phishing Attacks
Microsoft’s April 2026 updates combat phishing campaigns weaponizing Remote Desktop (.rdp) files by introducing educational warnings and disabling risky resource sharing by default.
Key takeaways
🎯 Target: Enterprise employees and remote workers who rely on Remote Desktop Protocol (RDP) for daily operations, a vector famously exploited by state-sponsored groups like APT29.
💡 Insight: The update moves Windows toward a “Zero Trust” posture for RDP files by introducing a mandatory one-time educational prompt and a security dialog that, by default, disables all local resource redirections (such as drives and the clipboard) even for signed files.
☑️ Recommendation 1: (Short-term) Deploy the April 2026 cumulative updates (KB5082200 for Windows 10; KB5083769 or KB5082052 for Windows 11) across all workstations immediately to activate these built-in RDP safeguards.
☑️ Recommendation 2: (Long-term) Update your organization’s security awareness training to emphasize that legitimate RDP files should always be digitally signed; instruct users to treat any “Unknown Publisher” warning as a critical phishing indicator and report it to the SOC immediately.
Critical MCP Vulnerability in Nginx UI Under Active Exploitation
The “MCPwn” vulnerability (CVE-2026-33032) in nginx-ui allows unauthenticated attackers to control Nginx servers via a single HTTP request. This actively exploited flaw bypasses authentication due to a 27-character omission in the Model Context Protocol (MCP) integration.
Key takeaways
🎯 Target: Organizations and developers utilizing nginx-ui (v2.3.3 and earlier) to manage Nginx configurations, particularly those with web interfaces exposed to the internet or untrusted networks.
💡 Insight: The vulnerability stems from a “fail-open” authentication gap where the /mcp_message endpoint, responsible for destructive actions like rewriting configurations and reloading the server, was left unprotected by mandatory authentication middleware.
☑️ Recommendation 1: (Immediate Patching) Administrators must upgrade to nginx-ui v2.3.4 or later immediately to apply the missing security middleware and secure the MCP transport.
☑️ Recommendation 2: (Emergency Mitigation) If patching is not immediately possible, disable the MCP feature entirely or restrict the IP whitelist to trusted management hosts, as the default “allow-all” configuration leaves servers vulnerable to any host on the network.
☑️ Recommendation 3: (Post-Incident Audit) Conduct a thorough review of Nginx access logs and configuration directories (e.g., conf.d/) for unauthorized server blocks or malicious access_log directives that could be used for credential harvesting.
N8n AI Workflow Automation Weaponized in Phishing Campaigns
Cisco Talos reports a 686% surge in phishing exploiting AI platforms like n8n. Attackers use trusted cloud infrastructure to deliver RATs and conduct device fingerprinting via malicious AI workflows.
Key takeaways
🎯 Target: Organizations and individual users who interact with “trusted” cloud subdomains (e.g., *.app.n8n.cloud), particularly those targeted via spoofed Microsoft OneDrive notifications.
💡 Insight: Attackers use n8n webhooks to serve dynamic, malicious HTML payloads; because the browser sees the data coming from a trusted n8n domain, it often bypasses reputation-based security blocks that would normally flag external malware, hosts.
☑️ Recommendation 1: (Immediate Visibility Audit) Configure security alerts for any internal traffic directed toward n8n.cloud or similar low-code subdomains that are not part of your organization’s authorized tech stack.
☑️ Recommendation 2: (Behavioral Detection) Shift defense strategies from simple domain blocking to behavioral analysis that triggers alerts when high volumes of data or unauthorized file types (like .exe or .msi) are pulled from automation platforms.
☑️ Recommendation 3: (Secure Email Defense) Deploy AI-driven email security solutions that utilize Natural Language Processing (NLP) to detect the subtle linguistic patterns of “agentic” phishing rather than relying solely on static blacklists.
Signed Adware Weaponized into a Global AV Killer
A campaign involving 25,000+ endpoints has transformed a harmless PUP into a security threat. Signed by Dragon Boss Solutions LLC, this software uses elevated SYSTEM privileges to silently uninstall antivirus products and modify system files to prevent their reinstallation.
Key takeaways
🎯 Target: Global organizations and individual users with systems infected by “monetization” browsers or utilities (e.g., Chromstera, Artificius) that carry the Dragon Boss signed update mechanism.
💡 Insight: The vulnerability wasn’t just in the code, but in the infrastructure; a primary update domain was left unregistered, allowing researchers to hijack the entire supply chain and push any payload directly to thousands of hosts.
☑️ Recommendation 1: (Immediate Forensic Hunt) Audit your environment for WMI event subscriptions named “MbRemoval” or “MbSetup” and scheduled tasks referencing “ClockRemoval.ps1” or “WMILoad” directories.
☑️ Recommendation 2: (Enforce Signing Blocks) Proactively block or alert on any executables digitally signed by Dragon Boss Solutions LLC and review Windows hosts files for unauthorized entries that block security vendor update domains.
☑️ Recommendation 3: (Re-evaluate PUP Risk) Elevate the priority of “Potentially Unwanted Programs” (PUPs) in your security policy. As this case proves, the line between “annoying adware” and “supply chain weapon” can vanish with a single domain registration.
The AI-Assisted Trojan Hiding in Your Obsidian Plugins
Attackers are weaponizing Obsidian plugins to distribute PHANTOMPULSE, an AI-generated Trojan bypassing security perimeters. By exploiting “Shell Commands” and social engineering, they gain remote control over Windows and macOS systems.
Key takeaways
🎯 Target: High-value professionals in the financial and cryptocurrency sectors, typically reached via targeted social engineering on platforms like LinkedIn and Telegram.
💡 Insight: The campaign poses a double threat: it uses AI to generate evasive code and abuses legitimate “living-off-the-app” techniques by hiding malicious logic within trusted Obsidian configuration files (data.json).
☑️ Recommendation 1: (Immediate) Disable “Community Plugins” and avoid opening or syncing vaults from untrusted sources, as malicious settings can execute automatically upon opening a vault.
☑️ Recommendation 2: (Strategic) Security teams should implement EDR telemetry to monitor for anomalous child processes (e.g., PowerShell or AppleScript) originating from Electron-based productivity applications like Obsidian.
The New Face of Vishing: AI-Powered TOAD Attacks
Cybercriminals are using Generative AI to create convincing voice phishing (TOAD) campaigns. These professional, accent-less voices trick employees into disclosing credentials or installing malware during high-pressure support calls.
Key takeaways
🎯 Target: Enterprise employees across all sectors, particularly those in Finance and HR who handle sensitive data or wire transfers.
💡 Insight: This “Hybrid” attack starts with a benign-looking email to bypass filters, but the real threat is the AI-generated voice on the other end of the phone, which eliminates traditional red flags like poor grammar or suspicious accents.
☑️ Recommendation 1: (Short-term) Alert your workforce to the “TOAD” tactic: never call a phone number provided in an unsolicited email to verify an account or security status; always use official, pre-verified internal directories.
☑️ Recommendation 2: (Long-term) Implement a “Zero Trust” communication policy that requires a secondary, out-of-band authentication (like a pre-shared code or internal chat confirmation) before performing any sensitive action requested over the phone.
Google vs. AI Scammers: The Malvertising Arms Race
Google is using Gemini AI to counter a surge in AI-generated malicious ads, blocking 8.3 billion harmful impressions in 2025. By shifting from keyword filtering to intent and behavioral pattern analysis, Gemini aims to intercept “living-off-the-platform” attacks at submission.
Key takeaways
🎯 Target: General internet users and IT professionals searching for popular utilities (e.g., Google Authenticator, Homebrew) or cryptocurrency platforms, who are lured by high-ranking “Sponsored” results.
💡 Insight: Scammers are using GenAI to rapidly produce deceptive ad variants at scale, requiring Google to use Gemini to analyze billions of signals, such as account history and campaign intent, in real time.
☑️ Recommendation 1: (Immediate) Practice “search hygiene” – never click on sponsored results for software downloads or financial logins; instead, navigate directly to the official domain or use bookmarks.
☑️ Recommendation 2: (Strategic) Organizations should deploy advanced DNS filtering and EDR solutions that can flag and block connections to recently registered or high-risk domains typical of malvertising redirect chains.
Marimo RCE Exploited for Blockchain Backdoors
Attackers are exploiting a critical pre-authentication RCE in the Marimo notebook platform (CVE-2026-39987) to deploy backdoors via HuggingFace. This campaign shows threat actors building working exploits from advisories in under 10 hours to hijack AI/ML infrastructure.
Key takeaways
🎯 Target: Data scientists, AI/ML researchers, and developers using the Marimo Python notebook platform, specifically those with internet-facing instances.
💡 Insight: Threat actors are leveraging “trusted” AI platforms like HuggingFace to host malware (via typosquats such as vsccode-modetx) and NKN blockchain protocols for Command and Control (C2), successfully bypassing legacy domain-reputation filters.
☑️ Recommendation 1: (Immediate) Update Marimo to version 0.23.0 or later and audit your systems for the kagent process or ~/.kagent/ directory, which indicates a successful compromise.
☑️ Recommendation 2: (Short-term) Block the malicious HuggingFace space vsccode-modetx.hf.space at the DNS level and rotate any credentials (AWS keys, DATABASE_URL, API tokens) stored in environment variables on your notebook servers.
☑️ Recommendation 3: (Strategic) Restrict notebook access to private networks or VPNs; never expose interactive developer tools like Marimo or Langflow to the public internet without an authenticated proxy.
Feature Video
How many times did you open the MITRE ATT&CK website yesterday?
If you’re like most CTI analysts, that number is somewhere between 30 and 60. And every single lookup is silently killing your productivity.
Here’s what I learned about reclaiming your focus:
🧠 Research shows it takes 23 minutes to return to peak focus after an interruption—even a brief one. Those “quick” ATT&CK lookups aren’t quick at all.
⏱️ 40 lookups/day × 45 seconds = 30 minutes of daily overhead. That’s 130+ hours per year lost to managing browser tabs.
🔧 The ATT&CK Powered Suit is a free browser extension that embeds the entire framework into your workflow. Right-click any text, get instant results. No tab switching required.
🏢 500+ security teams already use it—including JPMorgan Chase, Microsoft, and Verizon.
Feature Course
Learning Resources
Cyber Training
- Zero-Point Security: Advanced training in red team operations, adversary simulation, and offensive development. They equip you with the latest tactics and techniques to succeed in security and defense strategies.
- TCM Academy: A comprehensive suite of courses, including everything from penetration testing to malware analysis. Their hands-on, practical approach to training equips students with the real-world skills needed to succeed in cyber.
- Blue Cape Security: A specialist in Digital Forensics and Incident Response (DFIR) training, offering courses to take you from complete beginner to expert. Learn to defend like a pro.
Tools
- Octoparse: A no-code solution that will save you time, energy, and money. Let me show you how to use it to build your own custom cyber threat intelligence web-scraping tool!
