Rosetta Intel

Generated automatically · 31 sources configured (WebSearch aggregation + feeds) · 18 items selected
Window: past 24h (2026-06-12 → 2026-06-13). Saturday — the tail end of the post–Patch-Tuesday week. Today's throughline: the June 23 federal deadline for the CISA KEV trio (Chrome V8 / Cisco SD-WAN / Arista EOS) now drives the action list, with Cisco's unpatched SD-WAN Manager 0-day (CVE-2026-20245) still the standout — no fix available. Microsoft's record 206-CVE / 3-zero-day June Patch Tuesday is now being weaponized (Defender RoguePlanet PoC). On the leak sites, DragonForce and Qilin drove a busy Friday with ShinyHunters dumping Nexstar and Ralph Lauren data. AI-security stays hot: prompt injection is OWASP's #1 LLM threat of 2026, now with confirmed in-the-wild indirect-injection campaigns.

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

🔥⚠️ Cisco Catalyst SD-WAN Manager CVE-2026-20245 — root command execution, actively exploited, NO patch available
An attacker with netadmin privileges uploads a crafted file to inject commands and run them as root; Cisco has observed real exploitation pushing config changes to edge devices. Discovered by Google Mandiant. There is still no patch or mitigation — Cisco only advises applying the CVE-2026-20182 fixes (released 5/14) to limit the prerequisite path. On CISA KEV with a 6/23 federal deadline. Action: lock down management-plane reachability; hunt for unexpected edge-config pushes.
The Hacker News · Help Net Security · Cisco advisory
🔴🔥⚠️ Google Chrome V8 CVE-2026-11645 — CVSS 8.8 OOB read/write, 5th Chrome 0-day of 2026; CISA KEV (6/9)
Out-of-bounds memory access in the V8 engine lets a remote attacker execute code inside the sandbox via a crafted page. Patched in Chrome 149.0.7827.102/.103; same 6/23 KEV deadline. Action: confirm Chrome/Chromium auto-update has landed across the fleet, plus Edge/Electron derivatives.
The Hacker News · Help Net Security
🔥⚠️ Arista EOS CVE-2026-7473 — tunnel-decap flaw exploited; KEV (6/9), no patch planned
The switch decapsulates and forwards non-configured tunnel traffic whose destination IP matches the decap IP because it doesn't verify the tunnel protocol type. Arista won't patch (a fix risks breaking configs) and recommends upstream ACLs. Action: apply upstream/decap-device ACLs to allow only legitimate tunnel traffic.
SecurityWeek · CISA alert (6/9)
⚠️ Android Framework CVE-2025-48595 — elevation-of-privilege, under active targeted exploitation
Google's June 2026 Android security update patches an EoP flaw in the Android Framework component confirmed to be under limited, targeted exploitation. Action: push the June Android patch level to managed devices; prioritize high-risk users.
CyberInsider

Critical Vulnerabilities

🔴⚠️ Windows Kernel CVE-2026-45657 — CVSS 9.8 use-after-free RCE, the top fix in June Patch Tuesday
A use-after-free in the Windows Kernel can yield remote, unauthenticated code execution at SYSTEM. Microsoft rated exploitation "less likely," but it headlines a record 206-CVE / 39-Critical release. Companion critical RCEs cluster in HTTP.sys (CVE-2026-47291), DHCP Client (CVE-2026-44815), Remote Desktop Client (11 CVEs) and Hyper-V (guest→host escape). Action: prioritize Kernel, HTTP.sys, DHCP, RDP and Hyper-V hosts this weekend.
The Hacker News — 206 flaws · BleepingComputer
⚠️ Microsoft Defender "RoguePlanet" LPE PoC released post-patch — SYSTEM on fully-updated Win11/10
Researcher "Nightmare Eclipse" published a working exploit (RoguePlanet) for a race condition in Microsoft Defender (tied to CVE-2026-41091) shortly after Patch Tuesday, enabling local privilege escalation to SYSTEM. Action: confirm June updates applied; watch for Defender-targeting LPE behavior on endpoints.
Cybernews · SecurityWeek
⚠️ Windows BitLocker bypasses CVE-2026-45585 / CVE-2026-50507 ("bitskrieg") + BitUnlocker downgrade attack
June Patch Tuesday fixed publicly-disclosed BitLocker bypasses, including a believed-zero-day ("bitskrieg"). Separately, the BitUnlocker technique exploits the gap between patch deployment and certificate revocation to crack encrypted volumes on patched Windows 11 with physical access in ~5 minutes. Action: ensure pre-boot auth (PIN/TPM+PIN); track certificate-revocation timing for full mitigation.
Cybersecurity News

Vendor Advisories

Cisco— SD-WAN Manager CVE-2026-20245(root cmd-exec, KEV, no patch) plus auth-bypass CVE-2026-20182(5th/6th SD-WAN 0-days of 2026). The Hacker News
Microsoft— record 206-CVEJune Patch Tuesday (39 Critical; 3 publicly-disclosed zero-days: CTFMON CVE-2026-45586, BitLocker CVE-2026-45585& CVE-2026-50507). BleepingComputer
Google— Chrome 149.0.7827.102/.103 for V8 0-day CVE-2026-11645; June Android patch level for actively-exploited CVE-2025-48595. Help Net Security
Adobe— June Patch Tuesday critical batch shipped alongside Microsoft (per Qualys review). Qualys

Web Security Research

Zero Day Initiative — June 2026 Security Update Review breaks down the record Microsoft month, flagging HTTP.sys, DHCP and Kernel as the must-patch RCE cluster and noting the unusually high Critical count. Action: use ZDI's prioritization to sequence weekend patching.
ZDI
PortSwigger Research — "Top 10 Web Hacking Techniques of 2025" nominations open, the annual community vote surfacing the year's most important web research (recent contenders: CSS/HTML account-takeover on major email providers, new request-smuggling/desync variants, SAML auth-bypass, cache-poisoning via URL-parsing discrepancies). Action: review the shortlist as a web-AppSec reading list; nominate notable 2025 work.
PortSwigger Research

AI Security

🔴⚠️🛡 Prompt injection is OWASP's #1 LLM threat of 2026 — attacks up 340% YoY, first large-scale in-the-wild indirect campaigns confirmed
OWASP's 2026 LLM report keeps prompt injection at #1; Munich Re's cyber report named it a "major attack vector." In March 2026 Unit 42 documented the first large-scale indirect prompt-injection attacks in the wild (ad-review evasion, system-prompt leakage on live commercial platforms). Root cause is architectural: LLMs process instructions and data in the same channel. Action: treat all model-ingested external content (web, PDF, email, docs) as untrusted; gate agent tool/connector permissions.
Securance · Radware
🛡 Indirect injection + shadow AI + RAG are 2026's compounding LLM risks
Security analyses converge on indirect injection (malicious instructions hidden in retrieved/RAG content), unsanctioned "shadow AI" usage, and over-permissioned agents as the dominant enterprise LLM exposures. Action: inventory AI usage; constrain RAG sources and agent scopes; log model I/O for review.
SombraInc — LLM Security Risks 2026

Threat Intelligence

🛡 Unit 42 — Iran-nexus "Screening Serpens" (APT) expands espionage via AppDomainManager hijacking + new RAT variants
Unit 42 details ongoing 2026 campaigns by the Iran-aligned group targeting aerospace, defense manufacturing and telecom across the Middle East and Western Europe, using AppDomainManager hijacking and fresh RAT tooling. Action: hunt for AppDomainManager-hijack TTPs; review defense/aerospace supply-chain access.
Unit 42
🛡 Mandiant M-Trends 2026 / Unit 42 IR — adversaries 4× faster, sub-1-hour data exfiltration
Mandiant's M-Trends 2026 (500k IR hours) and Unit 42's Global IR report both highlight dramatically compressed attacker dwell time, with exfiltration in under an hour in some intrusions. Action: assume minutes-to-hours, not days; tune detection/response for speed, not just coverage.
M-Trends 2026 · Unit 42

AI Frontier

OpenAI

GPT-5.5 Instantis now the default ChatGPT model (replacing GPT-5.3 Instant) — sharper accuracy, better web-search use, richer context for Plus/Pro. Memory-source visibility/controls rolling out to Go & Free (6/9).
Confidential IPO filing(~6/8), a week behind Anthropic.

Anthropic

Claude Fable 5 / Mythos 5 (GA 6/9):one model split into two products by a safety-classifier layer. Fable 5is public (routes flagged cyber/bio requests down to Opus 4.8); Mythos 5— same model with cyber safeguards lifted — is restricted to vetted parties via Project Glasswing. The broad Fable release is enabled by new safeguards blocking high-risk cyber/bio responses.
Filed for IPO (6/1); ~$965Bvaluation, ~$47B annualized revenue.

Google DeepMind / AI

Gemini 3.5 Flashlaunched — beats Gemini 3.1 Pro on key benchmarks, leads multimodal, tuned for long-horizon agentic/coding tasks at <½ the cost. Ultra subscription cut $250→$200/mo; new $100/mo Developer tier.

🛡 = security-relevant

Chinese-Language Community Picks

Pwn2Own Berlin capacity crisis + "retaliatory disclosure": as reported by FreeBuf and others, Pwn2Own Berlin hit a hard registration cap for the first time in 19 years, triggering a capacity crisis; some excluded researchers launched "retaliatory disclosure", publishing details of the vulnerabilities they had prepared for the contest. Action: watch this batch of unplanned disclosures for usable exploits.
FreeBuf vulnerabilities
cPanel patches three high-severity vulnerabilities, CVE-2026-29201/2/3(privilege escalation, code execution, denial of service); upgrade immediately — the blast radius covers a large number of domestic shared-hosting and control-panel environments. Action: upgrade cPanel/WHM as soon as possible.
FreeBuf

Ransomware Today

A busy Friday on the leak sites, led by DragonForce and Qilin. Fresh named victims (24h): Cheoy Lee Shipyards (shipbuilding/maritime — watchlist: shipping), Areco Steel and Astec Valves & Fittings (manufacturing — DragonForce, watchlist: manufacturing), Bitek System Inc. (Qilin — watchlist group), Çekok Gıda (agriculture, Turkey — DragonForce), plus contracting/real-estate targets. Separately, ShinyHunters ("Trinity of Chaos") dumped data claimed from Nexstar (>1M Salesforce records) and Ralph Lauren (>220GB PII). Watchlist hits: group qilin; sectors manufacturing, shipping. → full victim table

Failed / Degraded Sources

RansomLook API(/api/posts?days=1) and RSS— not directly fetchable from the sandbox (provenance/egress restrictions). Ransomware section reconstructed via WebSearch (ransomware.live / breach trackers); counts are a lower bound, not a full leak-site snapshot.
Direct feed fetch (WebFetch) blocked by provenance rules this run — all categories sourced via WebSearch aggregation. Per-feed entry timestamps approximate; several items reflect the 6/9–6/12 Patch-Tuesday window still in scope.

Sources used: see intel/sources.yaml

Rosetta Daily · Jun 13, 2026