In today’s cybersecurity news…
Google Chrome installs 4GB AI model on devices
Computer scientist and lawyer Alexander Hanff reports that recent versions of Google Chrome automatically download a roughly 4GB Gemini Nano AI model to user devices without explicit consent when default AI features are enabled. He says the file installs silently, can re-download after deletion, and at Chrome’s scale could generate between 6,000 and 60,000 tonnes of CO2-equivalent emissions, raising privacy and environmental concerns. Hanff argues the behavior may violate privacy laws and calls for an opt-in prompt. (Cybernews)
Daemon Tools disk app backdoored in supply-chain attack
Kaspersky reports that the Daemon Tools disk imaging app was compromised in a monthlong supply-chain attack starting April 8th, with malicious updates signed by the developer infecting Windows users who downloaded versions 12.5.0.2421 through 12.5.0.2434. The malware collects system data and sends it to attacker servers, with thousands of machines across more than 100 countries affected. A subset of organizations received more advanced backdoors capable of executing commands and evading detection. Kaspersky says the attack was highly sophisticated and likely targeted, urging users to scan systems and monitor for suspicious activity. (Ars Technica)
Crypto’s ‘decentralised finance’ sector hit by investor exodus
The Financial Times reports that the DeFi sector is seeing a sharp investor pullback, with nearly $14 billion withdrawn after two major hacks, including a $290 million exploit tied to North Korean actors that destabilized open source protocol Aave and triggered a bailout. The attacks exposed structural risks in interconnected DeFi protocols, leaving Aave with more than $200 million in bad debt and pushing the market down to about $86 billion, near a yearly low. (Financial Times)
Iran cybersnoops still LARPing as ransomware crooks
Rapid7 researchers report that an Iranian state-linked group, likely MuddyWater tied to the Iranian Ministry of Intelligence and Security, masqueraded as the Chaos ransomware gang to conceal an espionage campaign. The attackers used Microsoft Teams phishing, social engineering, and remote tools to steal credentials, deploy Darkcomp backdoors, and move laterally across networks, while staging fake ransomware messages without encrypting files or seeking payment. Rapid7 says the operation was designed to obscure attribution and distract defenders, with stolen data ultimately published, suggesting intelligence-gathering or prepositioning for future attacks rather than financial motives. (The Register)
Huge thanks to our sponsor, Vanta
Sandbox bug lets attackers execute code on hosts
A critical vulnerability in the widely used Node.js sandboxing library vm2 lets attackers escape the sandbox and execute arbitrary code on the host system. The flaw stems from improper handling of exceptions, where WebAssembly features can bypass JavaScript-level protections and expose host objects, enabling access to sensitive Node.js internals. A proof-of-concept exploit is available, and users are urged to upgrade to version 3.10.5 or later. (BleepingComputer)
New offline CISA initiative for cyberattacks
CISA launched a new initiative called CI Fortify to help critical infrastructure operators maintain operations during cyberattacks by preparing to disconnect from internet and telecom dependencies and operate in isolation. The guidance emphasizes network segmentation, rapid recovery, and resilience as officials acknowledge that nation-state actors like China-linked Volt Typhoon may already be embedded in systems and difficult to remove. (The Record)
New Cisco DoS flaw requires manual reboot
Cisco has patched a high-severity denial-of-service flaw affecting its Crosswork Network Controller and Network Services Orchestrator products. The bug lets unauthenticated attackers remotely exhaust connection resources and crash systems, leaving them unresponsive until a manual reboot is performed. Cisco says there is no evidence of active exploitation but urges customers to upgrade to fixed versions, noting similar DoS flaws have been exploited in past attacks. (BleepingComputer)
Arctic Wolf thins out the pack
Arctic Wolf laid off 250 employees, or less than 10% of its workforce, as part of a restructuring to shift more investment toward AI, including its Superintelligence platform and agentic SOC offerings. The cuts impacted roles across sales, product, and marketing as the company looks to operate more efficiently while competing in the crowded MDR and EDR markets. There’s a broader industry trend going on of reallocating resources toward AI-driven security capabilities. (The Register)