👉 Takeaway

Prioritize HTTP.sys, BitLocker, and Remote Desktop patches this cycle. If your environment runs Defender on Windows endpoints, assume RoguePlanet exploitability until Microsoft patches it.

 

Network security

CVE-2026-20245 gives attackers with netadmin-level access the ability to execute arbitrary commands as root in Cisco Catalyst SD-WAN Manager, and Cisco has no patch ready. Mandiant reported the flaw; Cisco has confirmed “limited cases” of exploitation, including configuration changes being pushed to edge devices. The attack requires prior credential access or chaining with two related Cisco CVEs (CVE-2026-20182 and CVE-2026-20127), making this more of a lateral movement accelerant than an initial entry point. If you have SD-WAN Manager internet-exposed without strict access controls, that assessment deserves immediate review.

→ Collect diagnostic files now, and contact Cisco support if you see unexpected configuration pushes to edge devices.

Supply chain

The ongoing Shai-Hulud campaign compromised 37 malicious releases across 19 bioinformatics and scientific Python packages, including Dynamo, Spateo, and Napari-UFISH. The malware executes the moment Python launches by abusing a .pth startup file, then exfiltrates GitHub tokens, AWS credentials, SSH keys, Docker credentials, and Claude/MCP configuration files through auto-created GitHub Actions repositories. Security firm Socket identified the campaign, which now tracks 453 malicious items attributed to this threat group. Bioinformatics teams and data science environments are the primary targets, but any developer who ran pip install in the wrong environment needs to investigate.

→ If you installed any affected packages, rotate all secrets immediately and restore environments from clean backups.

Browser security

CVE-2026-11645 is an out-of-bounds read and write flaw in V8, Chrome's JavaScript engine, that allows a remote attacker to execute arbitrary code inside Chrome's sandbox via a crafted HTML page. Google fixed 74 vulnerabilities total in this release and is withholding technical details until most users have updated. An anonymous researcher reported the flaw in April and earned a $55,000 bounty; exploitation in the wild was confirmed before the patch shipped. This is the fifth Chrome zero-day Google has patched in 2026.

→ Update Chrome now. Version 149.0.7827.102 or later is the safe target across Windows, macOS, and Linux.

 

Social engineering

Microsoft Teams allows cross-tenant collaboration by default, which means attackers can message employees from external accounts while impersonating IT staff. The new development: AI-generated voice clones of known colleagues are following up initial text contact, making the social engineering significantly harder to dismiss than a chat message alone. The attack chain escalates from a Teams message to malware deployment to lateral movement and exfiltration using standard Windows tooling, with the 2024 Arup incident (a $25 million deepfake transfer) as the established playbook. The gap between “this sounds suspicious” and “this sounds exactly like Dave from IT” is closing fast.

→ Audit cross-tenant Teams access, restrict Quick Assist permissions, and establish out-of-band verification before anyone gets remote access to any system.

AI defense

Google is rolling out “fake call detection” to Android 12 and later devices globally this month, using RCS encryption to verify whether an incoming call's number matches the actual originating device. If the verification signal is absent or mismatched, the recipient sees a warning before answering. The feature works when both parties use Phone by Google with RCS enabled, is on by default for Pixel phones, and is rolling out broadly over the coming weeks. Impersonation fraud losses reached $2.95 billion in the United States in 2024 alone.

→ Confirm RCS is enabled on Android devices in your organization and verify that the Phone by Google app is installed for the feature to work.

 

Surveillance

WIRED found that Meta had quietly embedded facial recognition technology called “NameTag” into the Meta AI companion app installed on more than 50 million phones, where it had been sitting dormant since at least January 2026. The code was designed to identify strangers captured by the smart glasses camera and surface personal information in real time. Meta removed it within 48 hours of the report publishing after more than 70 advocacy groups, including the ACLU, demanded it be scrapped. The EFF's Threat Lab independently verified the findings through static analysis. Internal documents from 2025 show Meta had timed the rollout to coincide with a “dynamic political environment” to reduce criticism.

→ Meta removing the code is not the same as Meta abandoning the capability. The technology works. Expect it to return in a different packaging.

Privacy law

Oregon Senate Bill 1587, effective June 5, prohibits state and local government bodies from sharing personally identifiable information with data brokers unless those brokers first provide written attestation that the data will not be used for federal immigration enforcement. The law covers names, addresses, dates of birth, Social Security numbers, and biometric data. Public bodies can reject attestations they believe contain misrepresentations. This is a direct legislative counter to the documented practice of ICE and CBP purchasing location data from the commercial advertising ecosystem to track and arrest individuals.

→ This is a state-level counter-move to federal surveillance infrastructure, and it will face legal challenges. Seven other states are watching closely.

 

Practical play

🔒 Patch or Block: The Three-Day Window on a Critical VPN Zero-Day

CISA gave federal civilian agencies until June 11 to patch CVE-2026-50751, an authentication bypass in Check Point Remote Access VPN being actively exploited by Qilin ransomware affiliates. The vulnerability allows an unauthenticated attacker to establish a VPN session on affected Mobile Access and SSL VPN deployments. Exploitation started May 7 and surged over the weekend, with at least one confirmed Qilin post-compromise case. The key condition is running the deprecated IKEv1 key exchange without mandatory machine certificate authentication.

→ If you are not a federal agency, the June 11 deadline does not apply legally, but attackers are not checking your compliance calendar. Patch now, or verify you are not running IKEv1 without machine certs, or both. Full details: BleepingComputer

 

👉 Takeaway

A live video call is no longer reliable identity verification. Any process that depends on visual or audio confirmation alone needs a secondary out-of-band verification step added immediately.