Rosetta Intel

Generated automatically · ~22 sources scanned · 23 items selected

Critical Vulnerabilities

PTC Windchill / FlexPLM RCE — PTC Windchill PDMlink & FlexPLM, CVE-2026-12569, CVSS 9.3 🔴 🔥 ⚠️
Improper input validation / deserialization of untrusted data → unauthenticated RCE via a crafted network request. Added to CISA KEV on 6/25 with confirmed active exploitation; attackers drop JSP web shells named with 16 hex chars (hunt for POST /Windchill/login/[0-9a-f]{16}.jsp). Severe enough that German police (BSI) mobilized.
The Hacker News · PTC advisory
DirtyClone — Linux kernel privilege escalation — CVE-2026-43503, CVSS 8.8 🔴 ⚠️
New member of the DirtyFrag family. A local user corrupts file-backed memory via a cloned network packet to gain root. JFrog Security Research published a working exploit walkthrough on 6/25 — expect rapid weaponization.
JFrog / coverage
FFmpeg "PixelSmash" RCE — FFmpeg MagicYUV decoder ⚠️
Malicious video file → remote code execution; the flaw can trigger automatically during thumbnail generation, making it dangerous for any service that auto-processes uploaded media.
coverage
Oracle PeopleSoft Enterprise PeopleTools — missing authentication 🔴 🔥
Missing authentication for a critical function lets an unauthenticated attacker take over PeopleSoft Enterprise PeopleTools. Tracked on CISA KEV.
CISA KEV
Cisco Catalyst SD-WAN Manager — root command injection — CVE-2026-20245 🔥 ⚠️ (carryover, still active)
Authenticated netadmin uploads a crafted file → arbitrary commands as root; attackers push unauthorized config to edge devices. Still being actively exploited across the SD-WAN stack.
Rescana

In-the-Wild Exploitation (CISA KEV)

6/25 batch— PTC Windchill/FlexPLM (CVE-2026-12569) + a Cisco flaw added amid active exploitation / web-shell attacks.
SecurityAffairs· CISA alert
6/23 batch — remediation deadline was 6/26: CVE-2025-67038 (Lantronix EDS5000, CVSS 9.8) + CVE-2026-34908/-34909/-34910 (Ubiquiti UniFi OS). Verify your fleet is patched.
CISA alert

Vendor Advisories

PTC— Active advisory for Windchill/FlexPLM RCE; patch and hunt for JSP web shells, threat activity remains elevated post-patch.
PTC
Fortinet— Patched a critical-severity flaw in FortiSandbox / FortiSandbox Cloud / PaaS this cycle; review FortiGuard PSIRT.
FortiGuard PSIRT
Carryover— Microsoft June Patch Tuesday (206 CVEs / 3 zero-days incl. HTTP/2 Bomb CVE-2026-49160) and Cisco SD-WAN advisories remain the week's patch priorities.
SOCRadar

Web Security Research

Smuggling Requests with Chunked Extensions: A New HTTP Desync Trick(Imperva)
Malformed chunk extensions get the front-end and back-end to disagree on request boundaries, smuggling unauthorized requests past controls — a fresh twist on HTTP/1.1 chunked-encoding desync.
Imperva
XSS Cheat Sheet — 2026 Edition(PortSwigger) — updated reference of vectors/payloads for current browsers.
PortSwigger

AI Security

OWASP State of Agentic AI Security & Governance v2.01(published 6/11)
Catalogs CVEs, vendor advisories, and breach reports across nearly every category of agentic risk; reinforces the "lethal trifecta"(private data + untrusted content + external comms = exfiltration path).
Help Net Security
Prompt injection may be a permanent flaw, not a patchable bug
Argument gaining traction: LLMs have no architectural way to separate trusted commands from untrusted data (same token stream); filtering/least-privilege reduce but don't eliminate risk.
TechTimes
LiteLLM PyPI supply-chain backdoor (recap)— backdoored package sat on PyPI ~3h with ~47k downloads; LiteLLM is the LLM gateway for CrewAI, DSPy, Microsoft GraphRAG and more — large blast radius for agent stacks.
cyberdesserts

Threat Intelligence

Screening Serpens (Iran-nexus)— Unit 42 details AppDomainManager hijacking and new RAT variants targeting tech and defense sectors.
Unit 42
APT41 (Mandiant / Google TAG)— continued campaign hitting global shipping, logistics, media, tech and automotive orgs. (carryover)
Industrial Cyber
M-Trends 2026— attacks 4× faster, with data exfiltration under 1 hour in some cases; treat low-impact/commodity-malware alerts as high-priority precursors.
Google Cloud

Chinese-Language Community Picks

n8n workflow automation platform — three vulnerabilities chain to RCE— CVE-2026-44789~44791; chained together they reach remote code execution. FreeBuf
cPanel fixes three critical vulnerabilities— CVE-2026-29201/2/3, covering privilege escalation, code execution and denial of service. FreeBuf
Apache Flink CVE-2026-35194— SQL injection in the code-generation engine → RCE (carryover, still relevant). FreeBuf

Ransomware Today

RansomLook's 24h window returned 20 new leak-site posts across 15 groups (latest 2026-06-26T15:57). Most active: payload (3 victims posted within one minute — likely a backlog dump), then chaos / inc ransom / nova (2 each). Watchlist hits: akira → Precise Forms, inc ransom → GSP Crop Science + Life Bridges. Notable sensitive targets (not auto-flagged): leaknet → MagMutual (cyber-insurer, claimed 7.3M records), interlock → Clearview Eye Centre and payload → Clínica La Sabana (healthcare), the gentlemen → Atlas Elektronik (German defense electronics), nova → NSW Rural Fire Service and krybit → politur.gob.do (government/emergency).
Full victim table

Bug Bounty

The Bug Bounty deep-dive now updates daily on its own track (per-day disclosures + one deep analysis).
Open the Bug Bounty daily section

AI Frontier

OpenAI

Weighing a 2027 IPO— OpenAI now expects rival Anthropic to go public first; both have filed confidentially with the SEC. OpenAI had eyed a fall listing. (Bloomberg, 6/26)

Anthropic

Claude Tag on Slack (beta)— teams can @Claude in channels to delegate tasks, build context over time, and work async with controlled access to tools/data. Beta for Claude Enterprise & Team. (controlled tool/data access is the security angle)
IPO as soon as Octoberreportedly under consideration.

Google DeepMind / AI

Gemini 3 + Gemini 3 Deep Think— positioned as most powerful agentic/coding model; Pro available in preview across Gemini app, AI Studio, Vertex AI.
Managed Agents (Gemini API, public preview)— autonomous stateful agents run in secure isolated Google-hosted Linux sandboxes(notable from an agent-security/isolation standpoint).

Failed Sources (if any)

None blocking. Per established method, feeds were assembled via WebSearch per category + RansomLook API (direct RSS/Atom WebFetch is blocked by provenance gate / sandbox network). Feed-level timestamps are approximate.

Sources used: see intel/sources.yaml

Rosetta Daily · Jun 27, 2026