Two Netlogon Flaws Hit Domain Controllers as FortiBleed Lands in Finland

A heavy day for offensive practitioners: a no-auth Netlogon RCE PoC and a separate Zerologon-patch bypass both target domain controllers, while NCSC-FI confirms the FortiBleed credential-harvesting campaign is now hitting Finnish organizations. Add new Exchange Online spoofing and Entra Conditional Access bypass research, an actively-exploited Cisco flaw, and a fresh round of AI-security findings.

Offensive & Red Team

• CVE-2026-41089, a Windows Netlogon RCE via a CLDAP stack buffer overflow, now has a public PoC. A single crafted UDP packet to port 389 overflows a 528-byte stack buffer inside LSASS on any unpatched domain controller — no authentication required — currently demonstrated as a reliable DC crash/reboot (~60 seconds) with code-execution potential. Dark Web Informer. Patch DCs and restrict CLDAP exposure.
• Onelogon is a newly disclosed Netlogon authentication bypass that defeats the Zerologon patch, letting an attacker take over computer accounts or compromise an entire AD — though it requires a non-standard configuration to be present. al3x_n3ff / Florian Roth. Worth reviewing your DCs’ secure-channel config.

Cloud & Identity

• Ghost-Sender abuses a widely-present Exchange Online misconfiguration to spoof both internal and external senders, enabling universal phishing as a trusted sender. InfoGuard Labs. Strong purple-team test case for tenant hardening and detection.
• NetSPI detailed an Entra Conditional Access bypass via nested app authentication (NAA/BroCI token exchanges), allowing persistent Microsoft Graph token issuance post-compromise. Microsoft has since fixed it, but the token-exchange technique remains instructive. NetSPI.

Threat Activity

• FortiBleed has reached Finland: NCSC-FI confirms it is mapping affected Finnish organizations and issuing remediation guidance for a campaign it calls one of the year’s most significant. NCSC-FI. Researchers attribute it to a financially-motivated Russian-speaking initial-access broker using a custom Golang sniffer that has captured 110M+ credentials across 430,000+ FortiGate firewalls (and Sophos/MSSQL targets) since February 2026, leveraging a curated password list from prior breaches. The Hacker News, Unit 42.
• Two Scattered Spider members pleaded guilty on day one of their UK trial over the August 2024 Transport for London attack, which cost ~£39M and exposed data on ~10M customers. Thalha Jubair (20) and Owen Flowers (18) face sentencing July 16; Flowers had twice violated release conditions. Krebs on Security, NCA via vx-underground.
• A new macOS ClickFix variant uses fake-CAPTCHA Terminal commands to silently download, mount, and launch infostealers from DMG files in the background, hijacking cryptocurrency wallet data. BleepingComputer, Unit 42.

Vulnerabilities & Exploits

• CVE-2026-20230, a high-severity SSRF in Cisco Unified Communications Manager, is now exploited in the wild; a PoC was already public when Cisco shipped patches in early June. BleepingComputer, SecurityWeek.
• PixelSmash, a critical RCE in FFmpeg’s MagicYUV decoder, lets a crafted media file execute code in any app using libavcodec — RCE on Jellyfin under certain conditions, and DoS in Kodi, Emby, Nextcloud, PhotoPrism, and OBS. Now patched. JFrog, BleepingComputer.
• An eight-year-old use-after-free in Samsung KNOX exposed Galaxy devices from the S9 through S25 to kernel-level attacks. SecurityWeek.
• GIGABYTE confirmed a UEFI admin-password bypass via untrusted BootNext into WinRE, enabling offline attacks given physical access or Windows privileges — and is calling it a design issue. CyberInsider.
• GitHub updated actions/checkout to block pwn-request attacks that abuse pull_request_target to run attacker code with full workflow privileges, effective June 18. The Hacker News.

Supply Chain & Malware

• The Klue OAuth-token supply-chain attack widened: LastPass confirmed customer CRM data was accessed from its Salesforce environment, and the Icarus extortion crew added more victims, positioning itself as a potential CLOP-style supply-chain competitor. BleepingComputer, Dark Reading.
• A malicious npm package, @ts-apis/ts-utils, posing as a TypeScript utility library, drops a char-code-encoded PowerShell credential harvester that steals SSH/AWS/kube/Solana keys plus IMDSv2/ECS/GCP cloud creds, then opens an interactive C2 with a Sliver implant. Nextron’s generic SUSP_DecimalEncoded_PowerShell YARA rule caught it. Nextron Research.
• A set of npm packages masquerading as PostCSS tools deliver a Windows RAT, published over the past month by a single npm user. The Hacker News.
• Cordyceps: malicious pull requests are spreading through CI/CD workflows across major projects including Microsoft’s Azure Sentinel, Google’s AI Agent Development Kit, Apache Doris, Cloudflare’s Workers SDK, and Python’s Black. Dark Reading.
• A worth-watching reverse-engineering thread on a multi-stage Discord-delivered stealer: a fake rivals_toolkit.exe Electron app loads a bundled Discord.exe loader that spawns a JVM to run an obfuscated Java payload across four staged droppers — notable both for cross-language IPC and for leaving the AI agent’s own code-generation notes in the payload. vx-underground.

AI & Model Security

• Anthropic’s Mythos model reportedly found a 29-year-old, Heartbleed-style memory-leak flaw in Squid (“Squidbleed”) that had silently leaked plaintext HTTP requests, credentials, and session tokens for decades; maintainers have patched it. A US official separately said Mythos surfaced vulnerabilities in classified government systems within hours — though finding them is not the same as exploiting them. The Register, SecurityWeek.
• Anthropic’s Fable 5 — the “safe” guardrailed build of Mythos Preview meant to block cyberattack generation — was jailbroken within days of release. Schneier on Security.
• Security firm AIR built a benign fake AI-agent skill, pushed it through a popular skill marketplace and an Instagram ad, and says it reached ~26,000 agents — including some on corporate accounts — while every skill-security scanner tested marked it safe. The Hacker News.
• OpenAI expanded its Daybreak initiative with the full GPT-5.5-Cyber model, an updated Codex Security plugin, and a 25+ partner network, claiming it outperforms Anthropic’s Mythos on a cybersecurity benchmark and shifting focus from finding to auto-patching vulnerabilities. The Decoder.

Data Breaches

• Tata Electronics confirmed a cyberattack weeks after the World Leaks extortion group claimed to have stolen and published confidential client documents — reportedly including Apple and Tesla files. World Leaks also listed Reliance Power, putting two Indian conglomerates on its portal within 24 hours. The Record, BleepingComputer.
• A threat actor is advertising an alleged Notion database of 110M+ user records (emails, password hashes, signup/login IPs, workspace metadata). The claim is unverified and volume alone is not proof of a breach — but Notion users should enforce MFA and watch for suspicious logins. Dark Web Informer.
• Healthtech firm Xsolis disclosed a January phishing-driven breach affecting ~1.4M individuals’ personal and health data, with no reported misuse. BleepingComputer.

Industry & Policy

• President Trump signed an executive order setting hard federal post-quantum cryptography deadlines: key establishment must migrate by Dec 31, 2030 and digital signatures by Dec 31, 2031, with national security systems on a separate track. The Hacker News, The Record.