Rosetta Daily · Jun 12, 2026

Generated automatically · 31 sources configured (WebSearch aggregation + feeds) · 21 items selected
Window: past 24h (2026-06-11 → 2026-06-12). Friday — the post–Patch-Tuesday fallout week. Today's throughline: CISA's new KEV trio (Chrome V8 / Cisco SD-WAN / Arista EOS) with a June 23 federal deadline, fresh in-the-wild exploitation of ServiceNow and the Langflow AI app-builder, a heavy web-research drop (FROST SSD-timing side channel, HTTP/2 Bomb DoS, an unpatched Windows Search-URI NTLM leak), and LockBit5 still leading the leak sites (Central Romana, Shougang Peru).

In-the-Wild Exploitation (CISA KEV + active)

• 🔴🔥⚠️ Google Chrome V8 CVE-2026-11645 — CVSS 8.8 OOB read/write, the 5th Chrome 0-day of 2026; added to CISA KEV (6/9)
  Out-of-bounds memory access in the V8 JS engine lets a remote attacker run code inside the sandbox via a crafted page. CISA added it alongside two others with a 6/23 federal remediation deadline. Action: force-update Chrome/Chromium and all Edge/Electron derivatives now.
  The Hacker News · Chrome V8 0-day detail · CISA KEV

• 🔥⚠️ Cisco Catalyst SD-WAN Manager CVE-2026-20245 — CVSS 7.8, root command execution; KEV (6/9). Plus a 6th SD-WAN 0-day CVE-2026-20182 (auth bypass, UAT-8616)
  An authenticated local attacker runs arbitrary commands as root by supplying a crafted file. Separately, Cisco patched CVE-2026-20182, an auth-bypass in the SD-WAN Controller/Manager peering mechanism — the sixth SD-WAN zero-day whose exploitation surfaced in 2026, tied to threat actor UAT-8616. Action: patch SD-WAN Manager; restrict management-plane reachability.
  SecurityAffairs · SecurityWeek — 6th SD-WAN 0-day

• 🔥⚠️ Arista EOS CVE-2026-7473 — CVSS 6.9, tunnel-decap flaw exploited; KEV (6/9) — and NO patch is planned
  The switch incorrectly decapsulates and forwards non-configured tunnel traffic whose destination IP matches the decap IP, because it doesn't verify the tunnel protocol type. Arista says it won't patch (a fix risks breaking existing configs) and recommends ACLs on upstream/decap devices. Action: apply upstream ACLs to allow only legitimate tunnel traffic.
  SecurityWeek — no patch planned · CISA alert (6/9)

• 🔴⚠️ Langflow CVE-2026-5027 — CVSS 8.8 unauthenticated RCE in the AI app-builder, actively exploited
  VulnCheck reports in-the-wild exploitation of an unpatched path-traversal in Langflow: POST /api/v2/files doesn't sanitize filename, letting an attacker write files anywhere via ../ and achieve unauthenticated RCE. Langflow is widely used to build LLM apps — AI tooling itself is the entry point. Action: pull internet-facing Langflow behind auth/SSO; hunt for written webshells.
  The Hacker News

• ⚠️ ServiceNow — unauthenticated API endpoint abused to query customer-instance data
  ServiceNow applied a hosted-instance fix on 6/5 for an unauthenticated-access flaw (no CVE yet) and confirmed successful queries of instance tables against a subset of customers. Affected tenants were notified. Action: review ServiceNow audit logs for anomalous table queries; confirm the patched endpoint config.
  The Hacker News

Critical Vulnerabilities

• 🔴 Ivanti Sentry — max-severity (CVSS ~10) unauthenticated root RCE; Fortinet & SAP also ship critical fixes
  Ivanti patched two critical flaws in its Sentry secure-mobile-gateway, including a maximum-severity bug that lets a remote attacker execute code with root. Same day, Fortinet FortiSandbox CVE-2026-25089 (CVSS 9.1) — unauthenticated OS-command injection via the WEB UI — and a batch of SAP critical patches landed. Action: prioritize Sentry and FortiSandbox; both are unauth-RCE on security appliances.
  The Hacker News

• 🔴 "Proto6" — six vulnerabilities in protobuf.js expose Node.js apps to RCE and DoS
  Cyera disclosed six flaws (codenamed Proto6) in protobuf.js: a single malicious schema, descriptor, or crafted payload can trigger crashes, runtime corruption, or code execution. Hits Node.js apps, Google Cloud client libs, messaging frameworks (Baileys), and CI/CD pipelines that deserialize Protobuf. Action: bump protobuf.js; treat untrusted Protobuf input as hostile.
  The Hacker News

• ⚠️ One-character Linux kernel flaw → local root; public exploits available
  A one-character bug in the Linux kernel enables local privilege escalation to root, with working exploits now public. Action: patch kernels on multi-tenant / shell-access hosts as a priority.
  The Hacker News — top stories

Vendor Advisories

• Cisco— SD-WAN Manager CVE-2026-20245(root cmd-exec, KEV) and CVE-2026-20182(auth bypass, 6th SD-WAN 0-day); also Unified CM CVE-2026-20230with public exploit code. SecurityWeek
• Ivanti— Sentry critical patches (max-severity unauth root RCE). THN
• Fortinet— FortiSandbox 5.0.6 / 4.4.9 for CVE-2026-25089(9.1 unauth cmd injection). THN
• Microsoft— record 206-CVEJune Patch Tuesday (39 Critical; top fix Windows Kernel UAF CVE-2026-45657, CVSS 9.8); Defender RoguePlanetPoC now gives SYSTEM on fully-updated Win11/10. THN — 206 flaws· RoguePlanet

Web Security Research

• ⚠️ Unpatched Windows Search-URI flaw leaks NTLMv2 hashes — Microsoft declines to fix
  A crafted search: URI coerces the host to authenticate to an attacker SMB server, leaking the user's NTLMv2 hash for relay. Microsoft says it doesn't meet the servicing bar for Important/Critical, so it stays unpatched. Action: block outbound SMB (445) at the perimeter; filter search:/search-ms: handlers.
  SC Media · CyberSecurityNews

• HTTP/2 "Bomb" — remote DoS against NGINX, Apache, IIS, Envoy and Cloudflare Pingora
  A default-config HTTP/2 abuse pattern exhausts server resources across the major stacks. CVE-2026-49975 (Apache httpd) makes Cookie headers count against LimitRequestFields, closing one bypass. Action: patch httpd; tune HTTP/2 frame/stream and header limits at the edge.
  The Hacker News · GBHackers

• FROST — fingerprinting open sites/apps via OPFS-based SSD timing (DIMVA 2026)
  Graz University researchers show a JavaScript-only side channel that uses the browser's OPFS to time SSD behavior and infer which sites/apps are active — no user interaction, works on Linux and macOS. Action: track browser OPFS-mitigation guidance; relevant to high-assurance/privacy threat models.
  FROST paper (PDF) · Help Net Security

AI Security

• 🔴⚠️🛡 Langflow CVE-2026-5027 exploited in the wild — unauth RCE in an LLM app-builder (also under In-the-Wild above)
  AI-infrastructure software is being directly targeted: unauthenticated path-traversal → RCE on internet-facing Langflow. Action: auth-gate and patch AI build tooling; treat it like any other internet-exposed RCE-class service.
  The Hacker News

• 🛡 "Hacking Salesforce sites with an LLM agent" — autonomous agents weaponized against SaaS
  THN expert analysis details an LLM agent driving end-to-end recon and exploitation of misconfigured Salesforce sites, extending the "first AI-driven intrusion" trend (Sysdig) reported this week. Action: review Salesforce Experience/Site guest-access and object permissions; assume agent-speed enumeration.
  The Hacker News — expert insights

• 🛡 AI voice cloning added to Microsoft Teams social-engineering attacks
  Attackers are layering cloned-voice calls onto Teams-based lures to pressure targets in real time. Action: add out-of-band verification for sensitive Teams requests; brief help-desk and finance teams on voice-clone vishing.
  The Hacker News — expert insights

Threat Intelligence

• 🛡 China-nexus "JDY" botnet resurges to 1,500+ SOHO/IoT devices for recon (KV-botnet / Volt Typhoon lineage)
  Lumen Black Lotus Labs reports JDY — a covert scanner network of compromised routers, firewalls and IoT, first seen as a KV-botnet cluster in 2023 and used by Volt Typhoon — has expanded to 1,500+ devices to fingerprint exposed services at scale. Action: patch/replace EoL SOHO gear; monitor for KV/JDY scanning fingerprints.
  The Hacker News

• 🛡⚠️ Palo Alto CVE-2026-0257 — firewall auth-bypass exploited; campaign bears Chinese-state hallmarks
  Disclosed 5/13 and initially rated medium, PAN-OS CVE-2026-0257 was reassessed critical after Rapid7 saw in-the-wild use; attackers bypass auth to establish a VPN session to the firewall. SecurityWeek ties activity to a China-state-aligned campaign. Action: verify PAN-OS fixed versions; review GlobalProtect/VPN auth logs.
  CyberScoop · SecurityWeek

• "Miasma" worm hits 73 Microsoft GitHub repositories in a supply-chain attack
  A self-propagating worm compromised 73 repos under Microsoft's GitHub org, the latest in a run of CI/CD and dev-platform supply-chain incidents. Action: audit GitHub Actions/workflow tokens and repo write-paths; rotate exposed CI secrets.
  The Hacker News — top stories

---

AI Frontier

OpenAI

• Confidential IPO filing(≈6/8), a week behind Anthropic; last valued ~$852B post-money.
• ChatGPT "Lockdown Mode"ships — restricts tools/connectors that could enable data exfiltration via prompt injection in the browsing/agent surface.

Anthropic

• Claude Fable 5 / Mythos 5 split (GA 6/9):one underlying model shipped as two products separated by safety classifiers. Fable 5is public (routes flagged cyber/bio/chem requests down to Opus 4.8); Mythos 5— same model with cyber safeguards lifted — is restricted to vetted cyber-defenders and critical-infra operators, billed as "the strongest cybersecurity model in the world." Both $10/$50 per M in/out tokens.
• Closed financing at a $965Bvaluation; confidentially filed for IPO.

Google DeepMind / AI

• Gemini 3.5 Proexpected "this month" (no firm date); Gemini 3.5 Flashis now the default in the Gemini app and Search AI Mode.

🛡 = security-relevant

---

Chinese-Language Community Picks

• Shougang Peru named by LockBit5 / steel and mining critical infrastructure as a ransomware target
  The leak site shows LockBit5 naming shougang.com.pe (Shougang Peru, steel and mining)— another case of Chinese-invested critical infrastructure being extorted, with the intrusion estimated to date back to 5/20. Combined with this week's cross-exploitation of Cisco SD-WAN and Palo Alto by both Chinese state-level and financially motivated actors, the exposure surface of Chinese-owned assets and overseas Chinese enterprises is worth reviewing. Action: re-test VPN/SD-WAN exposure and backup isolation at overseas subsidiaries.
  ransomware.live· SecurityWeek — Cisco's sixth SD-WAN zero-day

Ransomware Today

Leak sites stayed busy in the window, with LockBit5 again out front. Fresh named victims (24–48h): Central Romana Corporation (agriculture/sugar, Dominican Republic — LockBit5), Shougang Peru / shougang.com.pe (steel & mining, China-linked parent — LockBit5), Global Schools Foundation (education, Singapore — Fulcrumsec, 6/10), Centre Ellipse (healthcare — Akira, 6/10), and Delano Public Schools (education — LockBit, 6/10). Watchlist hits: groups lockbit / akira / qilin; sectors healthcare, education, manufacturing; geos Singapore, China-linked. → full victim table

---

Failed / Degraded Sources

• RansomLook API(/api/posts?days=1) — returned 403 Forbiddenfrom the sandbox egress proxy; RansomLook RSSlikewise unreachable. Ransomware section reconstructed via WebSearch (ransomware.live / breach trackers). No structured group→victim JSON today.
• Several RSS/Atom feeds not directly fetchable (Apple HT201222, xz.aliyun.com, some vendor Atom) — covered via WebSearch aggregation.

---

Sources used: see intel/sources.yaml