Rosetta Intel

Generated automatically · ~22 sources scanned · 20 items selected

Critical Vulnerabilities

Ubiquiti UniFi OS — three actively-exploited flaws — Ubiquiti, CVE-2026-34908 / CVE-2026-34909 / CVE-2026-34910 🔴 🔥
CISA added all three to KEV on 6/23 citing active exploitation of internet-exposed UniFi OS consoles (UDM/Cloud Gateway/Network Application). Chainable toward unauthenticated access on edge appliances. Patch UniFi OS now and restrict the management UI from the internet.
CISA alert
Cisco Unified Communications Manager — SSRF — Cisco, CVE-2026-20230, CVSS 8.6 🔴 🔥
Unauthenticated SSRF via improper HTTP-request validation; can write files to the OS and escalate toward root. No workarounds. In CISA KEV (added 6/25); federal patch deadline was 6/28 — treat UC infra as exposed and confirm patched. (carryover, still hot)
CISA alert
PTC Windchill / FlexPLM — RCE (improper input validation) — PTC, CVE-2026-12569, CVSS 9.3 🔴 🔥 ⚠️
PTC confirmed continued reports of heightened threat activity despite the patch. In CISA KEV (added 6/25). After patching, still hunt for the 16-hex JSP webshell (POST /Windchill/login/[0-9a-f]{16}.jsp). (carryover, still hot)
The Hacker News
Lantronix EDS5000 — OS command injection — Lantronix, CVE-2025-67038 🔴 🔥
Serial-to-IP device server flaw added to KEV on 6/23; injection on an OT/edge gateway is high-impact for industrial networks. Inventory device servers and apply vendor firmware.
CISA alert
n8n "Ni8mare" — unauthenticated RCE / full takeover — n8n, CVE-2026-21858, CVSS 10.0 ⚠️
Content-Type confusion in webhook/file-handling lets an unauthenticated attacker read the DB + config, forge an admin session and reach RCE; ~100k self-hosted instances exposed. Patched in 1.121.0 — defenders are re-warning as internet-exposed instances remain unpatched. Upgrade or restrict public webhook/form endpoints. (reference — first disclosed Jan 2026, renewed exposure warning)
SecurityWeek · Cyera

In-the-Wild Exploitation (CISA KEV)

Ubiquiti UniFi OS CVE-2026-34908/34909/34910— added 6/23; patch consoles and pull the admin UI off the internet.
CISA alert
Cisco CUCM SSRF CVE-2026-20230— federal deadline 6/28 passed; confirm Unified CM / CM-SME are patched (no workaround).
CISA alert
PTC Windchill/FlexPLM CVE-2026-12569— exploitation continuing past the patch; hunt for webshells.
The Hacker News

Vendor Advisories

Microsoft— June Patch Tuesday recap stands: 206 CVEs, ~39 Critical, six 0-days (five publicly disclosed, one actively exploited) — the largest single Patch Tuesday in the program's 23-year history. Privilege-escalation dominated (~65 EoP). (carryover)
Cisco Talos· Security Affairs
Ubiquiti— UniFi OS fixes for the three KEV CVEs; update consoles immediately and audit remote-access exposure.
CISA alert
Cisco— CUCM SSRF (CVE-2026-20230, CVSS 8.6), no workarounds; patch immediately.
Cisco Security Advisories

Web Security Research

HTTP/1.1 desync endgame(PortSwigger) — desync research keeps escalating; timeout-based detection is now heavily WAF-fingerprinted, pushing researchers to malformed-chunk and structural techniques. Core reference for request-smuggling defense. (reference)
PortSwigger Research
Novel SAML authentication bypasses(PortSwigger) — new SAML exploitation techniques enabling complete auth bypass; revisit IdP/SP signature-validation and assertion handling. (reference)
PortSwigger Research

AI Security

Prompt injection: still OWASP's #1 LLM risk— 2026 reporting puts prompt injection in ~73% of production AI deployments with attacks reportedly +340% YoY. Architectural, not a patchable bug: LLMs can't separate trusted commands from untrusted data in one token stream. Filtering + least-privilege reduce, don't eliminate. (carryover)
Kunal Ganglani· GetAstra
MCP expands the agent attack surface— the rise of Model Context Protocol and tool-using agents introduces tool poisoningand credential theft; direct injection that redirects tool calls is now realistic on any agent with outbound capability. Scope tool permissions and treat all tool I/O as untrusted.
Sombra
Jailbreak-as-a-service economy— jailbreak/guardrail-bypass frameworks selling for $50–200/monthon dark-web forums; system-prompt extraction and roleplay attacks remain effective against deployed guardrails.
Group-IB

Threat Intelligence

Unit 42 2026 Global IR Report— 750+ major incidents across 50+ countries in 2025; attacks ~4× faster(exfiltration in <1 hr in some cases), 65% of initial access is identity-based, and supply-chain risk now centers on abuse of trusted SaaS/vendor connectivity. AI is a force-multiplier compressing the access-to-impact lifecycle.
RH-ISAC summary· Unit 42
Screening Serpens (Iran-nexus APT)(Unit 42) — AppDomainManager hijacking + new RAT variants targeting aerospace, defense manufacturing and telecom, expanded into Western Europe. (carryover)
Unit 42
ShinyHunters extortion wave— access + data-theft + extortion (not encryption) linked to 100+ orgsvia the Oracle PeopleSoft 0day (CVE-2026-35273); Kodak among confirmed victims. (carryover)
Malwarebytes

Chinese-Language Community Picks

FreeBuf / Anquanke— this cycle's Chinese-language source pages are JS-rendered, so only headlines were captured via search and the article bodies could not be fully retrieved. Worth watching for domestic reproduction and remediation write-ups on the KEV entries above (Cisco CUCM, PTC Windchill, UniFi OS) and on n8n Ni8mare.
FreeBuf

Ransomware Today

RansomLook's 24h API has not advanced — same 6 posts / 5 groups as 06-29, latest discovered still 2026-06-28T12:52Z (feed stall, treat as no fresh data today). Watchlist hits unchanged: play → Kuhnline and play → J&J Gaming. Manually-flagged sensitive carryovers: redact → Hologic (medical diagnostics, US) and redact → FCCI Insurance Group (insurance, US).
Full victim table

Bug Bounty

Bug Bounty has its own daily deep-dive (themed recent disclosures + one analysis per day).
Open the Bug Bounty daily section

AI Frontier

OpenAI

gpt-oss open-weight models— OpenAI's open-source release pairs a 120Bmodel (runs on a single GPU, near o3/o4-mini on several benchmarks) with a lighter 20Bthat runs locally on high-end laptops, intensifying the open-weight race.
Model arms race— new releases now land roughly every 2 days across OpenAI/Anthropic/Google/xAI; IPO timing remains a watch item. (carryover)

Anthropic

Claude Fable 5 shipped(6/9) — first public model in the new "Mythos" family; strong on software engineering, knowledge work and visual benchmarks; 1M-token context.
Project Glasswing expanded— ~150 more orgs granted Claude Mythos preview access for cyber-defense; the program reports helping early partners find 10,000+ high-severity vulnerabilities.

Google DeepMind / AI

Gemini 3.5 Pro nearing release— promised "next month" at I/O; Flash already GA (default in the Gemini app and AI Mode in Search). Positioned for agentic/coding.
AI coding push— Microsoft and Google moving directly against Anthropic and OpenAI in coding-focused models.

🛡 = security-relevant

Failed / Degraded Sources

RansomLook API— days=1window stalled at 2026-06-28T12:52Z (identical payload to 06-29); no fresh ransomware posts today.
FreeBuf / 安全客 / 先知— Chinese feeds are JS-rendered; item titles captured via search, full bodies not fully scrapable.
Several feeds not fetched directly (provenance/sandbox limits) — assembled via WebSearch per category.

Sources used: see intel/sources.yaml

Rosetta Daily · Jun 30, 2026