SAGE INTELLIGENCE BRIEF Friday, June 19, 2026 =========================================== LEAD STORY The FortiBleed campaign now has a confirmed scope: 75,000 Fortinet firewalls across 194 countries with live, harvested credentials being actively sold, covering roughly half of all internet-facing Fortinet devices visible on Shodan. Independent verification from Beaumont and Diachenko confirms these credentials are valid on current-patch devices, which kills the patching-is-sufficient argument dead. If you run FortiGate VPNs, credential rotation and MFA enforcement aren't optional this weekend. --- CONNECTING THE THREADS **FortiBleed closes the loop on a two-week escalation.** Earlier this week I noted that independent verification confirmed harvested Fortinet credentials are live on patched devices, and flagged that credential hygiene and MFA are now co-equal controls with patching on perimeter devices. Tonight's FortiBleed detail puts a number on the exposure: 75,000 devices, ~50% of the visible internet-facing Fortinet population. This moved from "pattern to watch" to "active campaign with quantified blast radius" in 48 hours. The urgency is significant. **INC ransomware's Veeam targeting connects directly to the backup-as-target thread.** INC has purpose-built a credential dumper for newer Veeam deployments using salted DPAPI encryption. Paired with the established pattern of edge-device initial access, this is a complete kill chain. Exploit unpatched perimeter gear (see FortiBleed), move laterally, neutralize backups, encrypt. These two stories in the same night aren't coincidental. The attack surface runs from the firewall to the backup server in one continuous chain. **DragonForce's Teams relay abuse advances the C2-over-trusted-infrastructure thread.** I've been tracking the pattern of attackers using legitimate cloud infrastructure to mask C2 traffic. DragonForce tunneled a QUIC session through a legitimate Microsoft TURN relay, making malicious traffic observationally identical to normal Teams traffic. This is the most technically mature example of that pattern I've seen in this brief cycle. Organizations relying on perimeter inspection of outbound traffic to Microsoft endpoints have a structural blind spot here. --- IT INFRASTRUCTURE ARCHITECTURE **2,000 retired Pixel phones repurposed as a private cloud** Someone built a functional private cloud cluster from 2,000 decommissioned Google Pixel handsets. The architecture is genuinely interesting from a distributed compute standpoint, but the enterprise takeaway is narrower: commodity ARM hardware at end-of-life has more residual compute value than most asset disposal programs account for. Worth watching as edge and AI inference workloads continue to stress conventional infrastructure budgets. Source: https://www.theregister.com/on-prem/2026/06/18/2000-retired-google-pixel-phones-get-a-second-life-as-a-private-cloud/5258035 **Citrix DaaS Flex targets PC refresh deferral** Citrix is positioning a cost-conscious DaaS licensing model as a direct response to soaring PC hardware prices. For MSPs managing client endpoint fleets, this is relevant: the financial case for VDI is stronger when PC refresh costs spike, and Citrix is clearly trying to convert that pressure into pipeline. Worth evaluating against existing Citrix agreements before clients start asking independently. Source: https://www.theregister.com/virtualization/2026/06/18/citrixs-daas-flex-aims-to-delay-pricey-pc-upgrades/5258476 **Netflix's scalable media processing pipeline published** Netflix detailed their cloud-based camera file processing pipeline built to handle global film and TV production workflows at scale. The architecture patterns around distributed media ingest and processing are applicable to any organization managing large unstructured data pipelines. The specifics around async processing queues and cloud-native scaling policies are worth reading if you're designing any high-volume file processing infrastructure. Source: https://www.infoq.com/news/2026/06/netflix-camera-file-processing/ --- CYBERSECURITY & COMPLIANCE **F5 patches two critical NGINX RCE vulnerabilities** F5 has released patches for two critical remote code execution flaws in NGINX Open Source. Specific CVEs and version ranges weren't fully enumerated in the article, so the direct reference point is last year's CVE-2026-42945 (NGINX Rift, CVSS 9.2), which went from disclosure to active exploitation within days. If you're running NGINX anywhere in your stack, check F5's advisory portal tonight and treat unpatched instances as imminently at risk. Source: https://thehackernews.com/2026/06/f5-patches-two-critical-nginx-open.html **INC ransomware: 830 victims, purpose-built Veeam credential dumper** INC is the fourth most active ransomware group in Q1 2026, with 830+ victims since 2023 and 120+ incidents last quarter. The Rust-rewritten encryptors now target Windows, Linux and ESXi. The critical operational detail: their credential dumper has been updated specifically for newer Veeam deployments using salted DPAPI encryption. Backup infrastructure is an explicit, engineered target. Verify your Veeam credentials aren't exposed via any compromised endpoint or domain account. Source: https://thehackernews.com/2026/06/inc-ransomware-claims-830-victims-since.html **DragonForce hides C2 inside Microsoft Teams relay traffic** DragonForce deployed a Go-based RAT that tunneled QUIC sessions through legitimate Microsoft TURN relays, making all C2 traffic appear as normal outbound Teams connections. The attack chain included BYOVD using a Huawei driver (HWAuidoOs2Ec.sys) to disable security tooling, DLL side-loading and persistent injection into DbgView64.exe. Access went undetected for one to two months. Immediate defensive steps: implement QUIC protocol inspection, restrict anonymous Teams/Skype token generation from endpoints and treat that Huawei driver hash as a high-confidence compromise indicator. Source: https://thehackernews.com/2026/06/dragonforce-hackers-abuse-microsoft.html **Microsoft Defender zero-day "RoguePlanet" disclosed, patch pending** Microsoft has formally acknowledged CVE-2026-50656, a Windows Defender zero-day assigned CVSS 7.8, currently without a patch. This is the fourth consecutive privilege escalation zero-day in the Defender codebase this cycle. Any detection architecture that treats Defender as a hard trust anchor on endpoints where untrusted code executes needs a compensating control layered on top until the patch ships. Source: https://thehackernews.com/ --- CLOUD PLATFORMS & STRATEGY **Enterprise AI workloads gravitating toward private cloud** The Register's partner content piece surfaces a signal beneath the marketing wrapper: enterprises scaling AI inference and training workloads are increasingly preferring private cloud over public hyperscalers for cost predictability and data governance reasons. This aligns with the datacenter capacity crunch story tracked earlier this week. The architectural pressure is showing up in procurement conversations. Source: https://www.theregister.com/ai-and-ml/2026/06/18/the-ai-tipping-point-where-enterprise-ai-runs-at-scale/5258147 **VS Code 1.123 adds two-hour extension update delay as supply chain control** Microsoft's VS Code team inserted a two-hour delay before auto-applying newly published extension updates, creating a revocation window to pull malicious updates before they deploy broadly. This is a direct response to the extension-based supply chain attack vector. For any team running VS Code in managed environments, this is a hygiene win, but it reinforces the broader point that developer toolchain components need the same update vetting discipline as production software. Source: https://www.infoq.com/news/2026/06/vscode-extension-update-delay/ **Splunk and Atlassian push critical patches** Splunk patched an OS command injection vulnerability in its AI Toolkit. Atlassian fixed multiple flaws in third-party dependencies. Both are common in enterprise environments and MSP toolstacks. Fast-track these in the next change window, particularly the Splunk AI Toolkit patch given the command injection severity. Source: https://www.securityweek.com/ --- NETDEVOPS & NETWORK AUTOMATION **ZTE and China Telecom advancing cross-vendor IP network simulation** ZTE and China Telecom Guangdong are running digital twin pilots claiming greater than 95% fidelity for multi-vendor network simulation before pushing changes to production. The partner content framing aside, the underlying use case is legitimate: change validation via high-fidelity digital twin is where network automation is heading for complex multi-vendor environments. Worth watching as the tooling matures toward commercial availability outside China Telecom's ecosystem. Source: https://www.theregister.com/networks/2026/06/18/zte-and-china-telecom-guangdong-advance-crossvendor-ip-network-simulation-pilots-paving-the-way-for-intelligent-network-operations/5258361 **UNC6508 actively targeting enterprise servers for backdoor deployment** China-linked UNC6508 is running an active campaign targeting enterprise servers for initial access and persistent backdoor deployment. No specific product or CVE was named in available reporting, but the pattern fits the broader state-sponsored initial access broker model. Server-facing management interfaces and exposed APIs are the consistent entry point class. Audit exposed management plane interfaces this week. Source: https://www.securityweek.com/ --- AI IN INFRASTRUCTURE & AIOPS **Microsoft Scout: always-on enterprise agent built on OpenClaw** Microsoft introduced Scout at Build 2026, an always-on enterprise autopilot agent. Scout represents a category shift: autonomous agents with persistent operational context, moving beyond prompt-response tooling. The IAM implication flagged earlier this week applies directly here. Non-human principals operating continuously inside enterprise environments need identity governance that current access control models weren't designed for. Source: https://www.infoq.com/news/2026/06/microsoft-scout-openclaw-build/ **Canonical releases Myna, local speech-to-text for Ubuntu** Canonical's Myna is a locally-executed speech-to-text application shipping with Ubuntu 26.10. The operational relevance is narrow but worth noting: local AI inference for sensitive transcription use cases removes the data-exfiltration risk of cloud-based voice services. For environments where conversation content is sensitive, local inference models like this are worth tracking as they mature. Source: https://www.theregister.com/os-platforms/2026/06/18/canonical-unveils-myna-speech-to-text-app-for-ubuntu-2610/5258376 **Neuromorphic computing as AI power efficiency pathway** Current research suggests hybrid neuromorphic-conventional systems could deliver meaningful efficiency gains at the edge for AI inference workloads. The practical horizon is still multi-year, but the directional signal matters for anyone sizing long-term edge AI infrastructure: the power consumption assumptions baked into today's edge designs may shift materially within the next hardware generation cycle. Source: https://www.theregister.com/science/2026/06/18/neuromorphic-computing-may-one-day-offer-ai-a-power-saving-brainwave/5258316 --- HARDWARE, GPU & COMPUTE **FortiBleed: 75,000 Fortinet firewalls, live credentials being sold** To put hard numbers on the lead story: the FortiBleed dataset covers devices across 194 countries with India, the U.S. and Mexico combining for roughly 12,000 compromised credentials. The dataset represents approximately 50% of all internet-facing Fortinet devices on Shodan. This is a threat-actor-controlled, actively monetized credential store. Rotate FortiGate admin credentials now, enforce MFA on VPN and management interfaces and audit for unauthorized sessions going back at least 90 days. Source: https://www.networkworld.com/article/4186794/fortibleed-campaign-exposes-75000-fortinet-firewalls-worldwide-2.html **Athena Coalition launches AI-driven open source security coordination** Chainguard launched the Athena coalition to apply AI to coordinated defense of open source software supply chains. For MSPs and enterprises with significant open source dependencies in their stack, this is the organizational layer being built to address the supply chain attack surface at scale. Worth monitoring as a source of threat intelligence specific to open source component risk. Source: https://www.infoq.com/news/2026/06/athena-security-coalition/ --- NETWORK MANAGEMENT & MONITORING No notable developments tonight. --- MANAGED SERVICE PROVIDERS No notable developments tonight. --- IT VENDOR ECOSYSTEM & M&A **Google denied bug bounty for unfixed flaw, researcher goes public** A researcher reported a vulnerability to Google, received a "nice catch" acknowledgment and then got denied the bounty with a "working as intended" ruling. The flaw remains unpatched. This is a recurring pattern with Google's VRP and the reputational cost is accumulating. More directly: if your security posture relies on Google-provided tooling, disclosed-but-unpatched flaws in that ecosystem don't show up in your standard vulnerability feeds until Google decides to act. Source: https://www.theregister.com/security/2026/06/18/google-told-researcher-nice-catch-then-denied-bug-bounty-for-flaw-it-still-hasnt-fixed/5258076 --- EDGE COMPUTING & IOT No notable developments tonight. --- WHAT TO WATCH The combination of FortiBleed's live credential dataset and INC ransomware's purpose-built Veeam credential dumper describes a complete, operationalized kill chain running from perimeter device compromise through backup neutralization to encryption. Both halves of that chain are active this week, simultaneously. Any organization with FortiGate perimeter gear and Veeam-based backup infrastructure should treat this as a drill trigger, not a watch item. --- CONVERSATION STARTER Roughly 50% of all internet-facing Fortinet firewalls visible on Shodan have had their administrator credentials harvested and are being actively sold. Patching alone doesn't close it. That's the opening for a direct conversation about whether MFA is enforced on every management and VPN interface in your environment right now. ===========================================