Rosetta Daily · Jun 18, 2026

Generated automatically · 30 sources scanned · 29 items selected
Window: past ~24h. Direct RSS/HTML feed pulls remain blocked (provenance/sandbox limits); this edition is compiled from WebSearch per-category + RansomLook API. Feed-level timestamps are approximate.

Critical Vulnerabilities

• SAP NetWeaver — SAML Authentication Bypass — SAP, CVE-2026-44748, CVSS 9.9 🔴
  A SAML auth-bypass flaw in SAP's June Security Patch Day. An attacker can forge identity assertions and gain unauthorized access to sensitive data across trust boundaries. Patch immediately on internet-facing SAP estates.
  SOCRadar

• SAP Kernel — Memory Corruption → RCE (unauthenticated) — SAP NetWeaver AS ABAP / ABAP Platform, CVE-2026-27671, CVSS 9.8 🔴
  Unauthenticated attacker sends a crafted RFC request exploiting memory-management logic errors → stack-based buffer overflow → crash, data access, or arbitrary code execution. Restrict RFC exposure and patch.
  SOCRadar

• Windows Kernel — Use-After-Free RCE — Windows 11 / Server, CVE-2026-45657, CVSS 9.8 🔴
  Critical network-reachable RCE in the Windows Kernel (use-after-free); fixed in the June updates. Prioritize deployment to endpoints and servers.
  Windows Forum

• Cisco Catalyst SD-WAN Manager — Path Traversal (in the wild) — Cisco, CVE-2026-20262 🔴🔥⚠️
  Directory/path-traversal added to CISA KEV on 06-15 amid active exploitation. Patch and audit SD-WAN Manager appliances.
  CISA KEV (6/15)

• Apache Flink — Code-Gen Engine SQL Injection → RCE — Apache Flink, CVE-2026-35194 🔴
  High-severity SQL injection in Flink's code-generation engine can expose distributed data-processing clusters to remote code execution. Patch and restrict job-submission surfaces.
  FreeBuf (via CSDN)

In-the-Wild Exploitation (CISA KEV)

• CVE-2026-20262— Cisco Catalyst SD-WAN Manager directory/path traversal (added 6/15).
• CVE-2026-54420— LiteSpeed cPanel plugin symlink-following (added 6/15).
• CVE-2026-35273— Oracle PeopleSoft Enterprise PeopleTools missing authentication for critical function (added 6/12).
• CVE-2026-7473— Arista EOS incomplete comparison / missing factors (added 6/9).
• CVE-2026-11645— Google Chromium V8 out-of-bounds read & write (added 6/9).
  CISA KEV — 6/15· 6/12· 6/9

Vendor Advisories

• Microsoft — June 2026 Patch Tuesday— ~198 CVEs (one of the largest single months): 32 Critical, 166 Important, and 3 publicly disclosed zero-days — CVE-2026-50507 (BitLocker security-feature bypass), CVE-2026-49160 (HTTP.sys DoS), CVE-2026-45586. Notable RCEs: CVE-2026-47652 (Hyper-V), CVE-2026-47288 (Kerberos KDC). Prioritize internet-facing HTTP.sys.
  CyberPress· ZDI Review
• Oracle — June 2026 Critical Patch Update— quarterly CPU released, including the PeopleSoft PeopleTools auth flaw now in KEV (CVE-2026-35273). Patch internet-exposed PeopleSoft.
  Oracle CPU
• SAP — June Security Patch Day— critical SAML bypass (CVE-2026-44748, 9.9) and SAP Kernel memory corruption (CVE-2026-27671, 9.8) lead the set.
  SOCRadar

Web Security Research

• PortSwigger — side-channels as a core primitive— 2025 research crystallized side-channels into mainstream web exploitation; landmark work includes ORM Leaking More Than You Joined For(generic ORM-leak methodology) and Successful Errors(error-based blind SSTI).
  PortSwigger Research
• Top 10 Web Hacking Techniques of 2025— annual community-voted ranking of the most innovative web research of the year.
  PortSwigger Top 10
• GhostLock PoC (Chinese community)— abuses Windows file interfaces to lock files on SMB shares so other users/apps cannot access them; a novel availability-denial primitive.
  FreeBuf (via CSDN)

AI Security

• OWASP GenAI v2.01 — "State of Agentic AI Security and Governance"(June 11) — documents multiple critical agentic flaws, including CVE-2026-2256, a command-injection in ModelScope's MS-Agent whose shell tool fails to sanitize commands.
  arXiv survey
• Prompt injection framed as a permanent, architectural flaw— LLMs have no built-in way to separate trusted commands from untrusted data (both arrive as the same token stream); newer models (GPT-5.x, Gemini, Claude, Grok) remain susceptible. Resource-amplification lets one injected instruction trigger thousands of agent actions.
  TechTimes
• AI-integrated malware enters the mainstream (Mandiant)— PROMPTFLUX and PROMPTSTEAL query LLMs at runtime; PROMPTFLUX (VBScript dropper) uses the Gemini API to rewrite its own source hourly; PROMPTSTEAL attributed to APT28.
  Google Cloud / M-Trends 2026

Threat Intelligence

• Mandiant M-Trends 2026— built on 500,000+ hours of frontline IR. Malware roles: backdoors 36%, downloaders 11%, ransomware 10%, droppers 10%, credential stealers 9%; 714 new malware familiestracked in 2025 (up from 632). Adversary pacing diverges — criminals optimize for speed, espionage groups for persistence.
  Google Cloud
• DragonForce — 'Backdoor.Turn'— custom malware hides command-and-control traffic inside Microsoft Teams relay infrastructure to blend with trusted traffic.
  BleepingComputer
• "The Gentlemen" ransomware— now the second-most-active gang by victim count; aggressive affiliate recruitment (90% ransom cut); claims 478 victims(analysis 06-11).
  BleepingComputer
• ShinyHunters → 7-Eleven— breached a 7-Eleven system holding franchise data; over 600,000 Salesforce records and franchise PII reportedly stolen.
  FreeBuf (via CSDN)
• iRhythm Holdings — healthcare data breach— digital-health firm discloses theft of patient personal/health data from third-party-hosted business apps.
  BleepingComputer

Chinese-Language Community Picks

• Apache Flink high-severity SQL injection (CVE-2026-35194)— injection in the code-generation engine can lead to RCE, threatening distributed data-processing environments.
  FreeBuf (via CSDN)
• ShinyHunters breaches 7-Eleven— over 600,000 Salesforce records and sensitive franchisee data stolen.
  FreeBuf (via CSDN)
• MSHTA widely abused as a LOLBIN— the legacy Windows MSHTA utility has been heavily used by attackers to deliver malicious payloads since late 2025.
  FreeBuf (via CSDN)
• GhostLock PoC— abuses the Windows file interface to lock files on SMB shares, blocking access by other users and applications.
  FreeBuf (via CSDN)

Ransomware Today

RansomLook added 23 posts across 15 groups in the last API window (06-16 08:39 → 06-17 03:46 UTC; window overlaps yesterday's report). Most active: aurora / nightspire / cloak (3 each). ⭐ 4 watchlist hits — qilin (Golfview Developmental Center, healthcare), akira (Insite Architects), inc ransom (framesiprofessional.com, jasperplastics.info). High-value claims: Novo Nordisk (pharma, fulcrumsec), Nintendo (shadowbyt3$), Ralph Lauren (shinyhunters), Sumitomo Electric Bordnetze (automotive, aurora).
→ Full victim table

---

AI Frontier

OpenAI

• Announced the Partner Network(06-14) and new Academycourses (06-12); reported plans to acquire Ona; confidential SEC filing and an Economic Research Exchange. GPT-5.5 remains the default after GPT-5.2 was retired (06-12); frontier models + Codex are now on AWS.
  OpenAI News

Anthropic

• Claude Fable 5 / Mythos 5announced (06-09), alongside Opus 4.8(released 05-28; honesty/reliability focus, 84% on Online-Mind2Web computer-use). Managed Agentscan now run in a customer-controlled sandbox and connect to private MCP servers. Enterprise: TCSand DXCpartnerships for regulated industries (06-12); 20+ legal MCP connectors.
  Anthropic Newsroom

Google DeepMind / AI

• Gemini 3family positioned as the layered stack — Gemini 3 Flash (app default), 3.1 Pro, 3.5 Flash, 3.1 Flash-Lite — plus Deep Thinkreasoning mode and Deep Research Maxfor autonomous research agents.
  Google DeepMind Blog· The Keyword

🛡 = security-relevant

---

Failed Sources / Notes

• Direct RSS/Atom/HTML pulls from sources.yamlremain blocked (workspace web-fetch provenance + sandbox network). This edition uses WebSearchper category + RansomLook API(its URL is in the task message, so fetchable).
• Chinese-community sources (FreeBuf, 先知, 安全客) and HTML AI-security blogs (HiddenLayer, Lakera, Protect AI) are only reachable via WebSearch summaries (US-region search).
• RansomLook ?days=1returned the 06-16→06-17 window (live clock lags the 06-18 report date); ransomware figures overlap the prior day.

---

Sources used: see intel/sources.yaml