cybr.cx — Daily Intelligence Digest

Wednesday, October 21, 2026

Critical Vulnerabilities

⚠️ Actively exploited — CVE-2026-106931 | EG Secure Solutions SiteGuard Server Edition | CVSS: N/A (KEV)
An OS command injection flaw in SiteGuard Server Edition allows unauthenticated attackers to execute arbitrary commands with root or SYSTEM privileges. CISA's patch deadline has already passed — if this is in your environment, treat it as compromised until patched.

⚠️ Actively exploited — CVE-2026-96652, CVE-2026-96654, CVE-2026-96655, CVE-2026-96656 | Plex Media Server | CVSS: N/A (KEV)
Four chained vulnerabilities in Plex Media Server are being actively exploited together. The chain combines two SSRF flaws (96652, 96655), a URI scheme injection that hijacks plugin calls (96654), and an arbitrary file write that executes on load (96656). Any authenticated — or in one case, any X-Plex-Token holder — can initiate the chain. Plex instances exposed to untrusted networks should be isolated or taken offline immediately.

⚠️ Actively exploited — CVE-2026-110936 | Siemens Software Center | CVSS: N/A (KEV)
An uncontrolled search path vulnerability allows a local authenticated attacker to drop a crafted DLL into a directory Siemens Software Center searches at runtime, resulting in arbitrary code execution. Particularly concerning in OT/ICS environments where Siemens tooling is ubiquitous and lateral movement dwell time is high.

⚠️ Actively exploited — CVE-2026-102764 | BeyondTrust Remote Support | CVSS: N/A (KEV)
Authentication bypass via spoofing in BeyondTrust Remote Support allows remote attackers to perform file operations on the appliance without credentials. PAM infrastructure being actively targeted is a high-severity operational risk — assume privileged session data may be at risk.

CVE-2026-105949 | LinkedIn Cruise Control (Apache Kafka) | CVSS: 8.8 — HIGH
Cruise Control's Jetty security handler registers constraints only on exact path specs and never calls setDenyUncoveredHttpMethods(true). Combined with Path.of(pathInfo).getFileName() stripping leading path segments, a request like POST /kafkacruisecontrol/x/rebalance bypasses authentication entirely. Anyone exposing Cruise Control's API surface — even internally — should firewall it immediately.

CVE-2026-100134 | Adobe Acrobat for Edge | CVSS: 8.8 — HIGH
An out-of-bounds write in Acrobat for Edge allows arbitrary code execution in the current user's context when a malicious PDF is opened. User interaction is required, making phishing delivery the most likely vector. Disable the Acrobat Edge extension or restrict PDF handling until a patch is available.

CVE-2026-53995 | MLflow ≤ 3.17.1 | CVSS: 8.2 — HIGH
Unauthenticated path traversal in MLflow's FileStore backend allows attackers to submit crafted trace IDs containing ../ sequences to the /api/3.0/mlflow/traces endpoint, writing arbitrary files to the filesystem. ML infrastructure increasingly sits adjacent to sensitive model weights and training data — this deserves urgent attention in data science pipelines.

CVE-2026-109440 | LangAlpha ≤ 2026.09.07 | CVSS: 8.1 — HIGH
An indirect prompt injection vulnerability allows attackers to embed instructions in web content retrieved by LangAlpha agents. Because Bash and ExecuteCode tools have no allowlist and no human-in-the-loop approval gate, malicious instructions execute directly in the workspace sandbox and can exfiltrate secrets from _internal/.vault_secrets.json. Agentic AI pipelines with unrestricted tool access remain a systemic risk.

CVE-2026-113613 | FlexGet < 3.20.7 | CVSS: 8.1 — HIGH
A zip-slip style path traversal in FlexGet's decompress plugin lets attackers deliver malicious archives via a feed containing ../ sequences, writing Python files directly into the plugins directory for code execution. Update to 3.20.7 and audit feed sources.

CVE-2026-113394 | DrayTek Vigor2960 1.5.1.6 | CVSS: 7.4 — HIGH
Path traversal via the setSWMProfile argument in /cgi-bin/mainfunction.cgi is remotely exploitable and has a public exploit. The vendor has addressed this in firmware 1.5.1.7. Given DrayTek's historical targeting by threat actors, patch or segment these devices immediately.

Headline News

Half a Million AfD Emails Hit the Public Domain

Over 420,000 internal emails belonging to Germany's Alternative für Deutschland party were leaked and published, offering an unprecedented window into the internal communications of a major European political organisation. The breach — attributed to an external intrusion rather than an insider — exposed private correspondence, strategic discussions, and sensitive personal data across a broad cross-section of party figures. For security practitioners, the incident underscores the persistent exposure risk of email as an archival system: years of communications, once exfiltrated, become a permanently exploitable intelligence asset. The scale and searchability of the leak will make it a reference case for how political organisations handle — and fail to protect — long-term email retention. Organisations in sensitive sectors should treat email archives as high-value targets equivalent to databases, with equivalent access controls and monitoring.

A Novel Ransomware Variant Is Confounding Identification Efforts

A previously unidentified ransomware strain encrypting files with a .locked extension, a CHL3 marker, and a distinctive 93-byte footer is circulating and resisting classification by current tooling. Victims are reporting widespread encryption across multiple drives on Windows 11 systems, with no confirmed decryptor available. The unique file footer signature suggests a custom or heavily modified encryption implementation, complicating the usual identification workflow through tools like ID Ransomware. Practitioners running incident response should capture and preserve the 93-byte footer and ransom note verbatim — these artefacts will be critical to any eventual cryptographic analysis. Until attribution is established, treat this as a novel family and do not assume existing decryption tools will apply.

Suspicious OT/ICS Framework Appears on PyPI

A package named industrialxpl surfaced on PyPI advertising itself as an "OT/ICS/SCADA lab framework" with over 1,190 modules, built-in Mirai botnet and C2 support, and described as "Python-First, Mirai/C2 native ops." The framing as a legitimate lab tool is a thin cover for what reads as an operational capability targeting industrial control systems. Security teams should check Python environments in OT-adjacent networks for this package, and organisations with PyPI mirrors or internal package proxies should blocklist it. The incident is a reminder that threat actors increasingly use open package registries as distribution vectors, particularly for tooling dressed up as research or testing frameworks.

Schrödinger's Feed

Post-Quantum Security for Institutional Custody Moves Forward

Project Eleven, a post-quantum security company focused on digital assets, has announced a development partnership with Relm Insurance to explore post-quantum security frameworks for institutional custody products. The collaboration is specifically examining how cryptographic agility can be built into custody infrastructure before cryptographically relevant quantum computers arrive — a forward-looking posture that most financial institutions have yet to adopt. China's accelerating investment in fault-tolerant neutral-atom quantum hardware (two well-funded startups reaching significant scale within months) adds urgency to the timeline question. Practitioners managing long-lived financial or identity infrastructure should be actively tracking PQC migration readiness — "harvest now, decrypt later" attacks mean the threat window is already open.

/dev/random

Facebook Pages Have Discovered Paywalled Outbound Links

Meta has quietly introduced a limit of two outbound links per month for Facebook Pages that don't hold a paid subscription — a policy that, if enforced broadly, would effectively make organic link-sharing from public pages a premium feature. The mechanism appears to be enforced at the post-submission layer rather than via UI warnings, meaning page admins may not know they've hit the cap until posts silently fail or get suppressed. It's a novel form of platform-level content control that sits somewhere between a paywall and a shadow restriction — security researchers and threat intel teams who use Pages for public disclosure or community communication should take note. At this rate, the most reliable way to share a CVE advisory on Facebook will be to print it on a hat and photograph it.