cybr.cx Daily Digest — September 09, 2026

Critical Vulnerabilities

⚠️ Actively exploited — CVE-2026-86218 | N-able N-central | No CVSS in NVD feed
N-able's N-central RMM platform has a static code injection flaw enabling pre-authentication remote code execution. This is as bad as it gets for MSPs: N-central agents sit on thousands of managed endpoints, meaning a single compromised server is a skeleton key to every client network downstream. CISA's patch deadline was September 11 — if you're running N-central and haven't patched, treat this as an active incident.

⚠️ Actively exploited — CVE-2026-75650 | Adobe Commerce / Magento | No CVSS in NVD feed
A template engine injection flaw in Adobe Commerce and Magento Open Source is seeing active exploitation, giving unauthenticated attackers arbitrary code execution on e-commerce storefronts. With end-of-summer retail traffic elevated, threat actors are clearly timing campaigns for maximum impact. CISA's deadline passed September 11 — patch immediately and audit for indicators of webshell deployment.

⚠️ Actively exploited — CVE-2026-85046 | Google Chromium V8 | No CVSS in NVD feed
A type confusion vulnerability in Chromium's V8 JavaScript engine allows a remote attacker to execute arbitrary code inside the sandbox via a crafted HTML page. Every Chromium-based browser is affected — Chrome, Edge, and others. Drive-by exploitation is trivially achievable; force-push browser updates across your fleet now and consider restricting access to unmanaged browsing on sensitive systems.

⚠️ Actively exploited — CVE-2026-85880 | Microsoft Windows ALPC | No CVSS in NVD feed
A heap-based buffer overflow in Windows Advanced Local Procedure Call (ALPC) allows local privilege escalation to SYSTEM. Combined with any initial-access vector, this is a reliable post-exploitation path to full machine compromise. CISA deadline is September 22, but prioritise deployment given active in-the-wild use.

⚠️ Actively exploited — CVE-2026-81963 | Microsoft Windows Update Stack | No CVSS in NVD feed
A link-following vulnerability in the Windows Update Stack enables local privilege escalation to SYSTEM. Pairing this with CVE-2026-85880 gives attackers two independent local escalation paths on unpatched Windows systems — the risk is compounded in environments where standard users have any foothold. Patch deadline is September 22.

⚠️ Actively exploited — CVE-2026-82329 | JFrog Artifactory | No CVSS in NVD feed
Under default configuration, JFrog Artifactory allows an unauthenticated network-adjacent attacker to obtain administrative privileges. Artifactory commonly sits at the heart of software supply chains; admin access means poisoning build artefacts, exfiltrating signing keys, and lateral movement into CI/CD pipelines. CISA's remediation deadline was September 5 — if you haven't acted, audit your instance logs immediately.

⚠️ Actively exploited — CVE-2026-83548 / CVE-2026-83549 | SonicWall SMA1000 | No CVSS in NVD feed
Two actively exploited vulnerabilities in SonicWall SMA1000: an unauthenticated SSRF (CVE-2026-83548) enabling access to sensitive internal functionality, and an authenticated OS command injection (CVE-2026-83549) delivering full RCE as administrator. SonicWall edge appliances are perennial ransomware gang favourites — both CISA deadlines passed September 5.

⚠️ Actively exploited — CVE-2026-49869 | Kestra OSS | No CVSS in NVD feed
Kestra's open-source workflow orchestration platform has an OS command injection flaw allowing unauthenticated remote attackers to create and execute arbitrary workflows without credentials. Orchestration platforms have broad internal network access by design, making this a high-value pivot point. Remediation deadline was September 5.

⚠️ Actively exploited — CVE-2026-9586 | Sangoma Switchvox | No CVSS in NVD feed
A single unauthenticated SQL injection request against Sangoma Switchvox gives attackers arbitrary SQL execution against the backend PostgreSQL database, including a path to OS-level RCE. VoIP infrastructure is consistently under-patched relative to its network exposure. CISA deadline passed September 5.

⚠️ Actively exploited — CVE-2026-59822 | BerriAI LiteLLM | No CVSS in NVD feed
An improper authentication flaw in LiteLLM's MCP Streamable HTTP endpoint allows any unauthenticated attacker to establish a fully authenticated MCP session using an arbitrary Bearer token. As LiteLLM proxies are increasingly deployed as AI gateway infrastructure, this represents a direct route to hijacking LLM API sessions and their associated credentials. Patch deadline is September 16.

⚠️ Actively exploited — CVE-2026-48710 | Kludex Starlette | No CVSS in NVD feed
An HTTP request smuggling flaw in the Starlette ASGI framework allows attackers to inject paths into the host component, potentially bypassing authentication middleware that trusts the reconstructed URL. Given Starlette underpins FastAPI and a wide range of Python web services, the blast radius is substantial. Patch deadline is September 16.

CVE-2026-86712 | SiYuan (before 3.8.2) | CVSS 8.8
SiYuan's desktop note-taking app (Electron-based) trusts a writable text/siyuan clipboard MIME type and skips sanitisation in its paste handler. An attacker can craft a malicious web page that writes to the clipboard; if the user pastes into SiYuan, injected scripts execute with full Node.js access through the Electron main process — effectively native code execution. Update to 3.8.2 immediately and treat clipboard content from untrusted sources as an attack vector.

CVE-2026-86730 / CVE-2026-86732 | Craft CMS (before 5.10.12) | CVSS 8.8 (both)
Two separate RCE paths in Craft CMS affect authenticated control-panel users. CVE-2026-86730 abuses Yii2 behavior attachment via unsanitised field-layout JSON strings; CVE-2026-86732 exploits the element-index endpoint's criteria parameter to instantiate arbitrary PHP classes through a gadget chain involving yii\rbac\PhpManager. Authenticated-only scope limits exposure, but compromised editor accounts are common. Patch to 5.10.12 and audit recent field-layout changes.

CVE-2026-62744 | Microsoft Windows Media Foundation | CVSS 8.8
A heap-based buffer overflow in Windows Media Foundation is exploitable over a network, leading to code execution. Media parsing libraries processing untrusted content remain a reliable attack surface; ensure this month's Windows updates are applied, and review any services that process media files on behalf of external users.

CVE-2026-62895 | Azure Arc | CVSS 8.8
A permissive cross-domain policy in Azure Arc allows an unauthenticated network attacker to elevate privileges. Azure Arc is often used to manage on-premises and multi-cloud resources from a single control plane, making privilege escalation here significant for hybrid environments. Apply the available patch and review Arc-connected resource policies.

CVE-2026-65772 | Microsoft Dynamics 365 | CVSS 8.8
Deserialization of untrusted data in Dynamics 365 allows an authenticated attacker to achieve remote code execution. Dynamics deployments often hold sensitive CRM and financial data; treat any authenticated user as a potential threat and apply patches promptly.

CVE-2026-66814 / CVE-2026-66818 | Microsoft SQL Server | CVSS 8.8 (both)
Two privilege escalation flaws in SQL Server — one via insufficient access control granularity, one via improper privilege management — allow authenticated attackers to elevate privileges over the network. Database servers are high-value lateral movement targets; patch and audit SQL Server role assignments.

Headline News

$320 Million Drained from Bitcoin's Liquid Network in Disputed "White Hat" Operation

Attackers drained approximately $320 million in Bitcoin from the Liquid Network sidechain before returning the majority of funds and retaining roughly $47 million — claiming a white-hat security disclosure bounty. The technical mechanism exploited weaknesses in Liquid's federated multi-signature peg architecture, which relies on a federation of functionaries to validate cross-chain transfers. The self-declared "white hat" framing raises serious questions for practitioners: the line between responsible disclosure and profit-motivated exploitation blurs considerably when the "researcher" keeps $47 million. For DeFi and sidechain operators, the incident is a stark reminder that federated bridge designs carry systemic risk, and that peg security assumptions deserve adversarial scrutiny at least as rigorous as the underlying chain itself.

The Security Debt Clock Is Ticking: One Year to Fix It Everywhere

A widely circulated technical essay arguing that the industry has approximately one year to address foundational security failures before a confluence of threats — maturing offensive AI tooling, quantum cryptography timelines, and increasingly sophisticated supply chain attacks — makes remediation prohibitively costly has sparked significant discussion among practitioners. The piece frames the problem not as any single vulnerability class but as structural: too much critical software runs on dependency stacks that have never been audited, with security properties that have never been formally verified. The argument carries particular weight given the KEV entries in today's digest — orchestration platforms, RMM tools, and ASGI frameworks all represent exactly the kind of foundational infrastructure described. Whether or not the one-year thesis holds precisely, the directional warning is sound: organisations that treat security debt as a backlog item rather than an active liability are accumulating risk faster than they can remediate it.

Schrödinger's Feed

IonQ has published a detailed fault-tolerant resource estimate showing that 256-bit elliptic curve cryptography — the algorithm family securing Bitcoin wallets, TLS certificates, and a significant portion of the internet's authentication infrastructure — could be broken in 25.7 days using 19,397 physical qubits in their optimised "Walking Cat" architecture with qLDPC error correction codes. This isn't a theoretical upper bound; it's a concrete engineering roadmap. The timeline to hardware capable of realising this estimate remains contested, but the fact that the resource requirements are now this precisely specified changes the conversation from "if" to "when and how fast." Practitioners managing long-lived secrets, certificate authorities, or any system relying on ECDSA should treat PQC migration as an active project, not a future consideration — the planning horizon is shorter than most roadmaps acknowledge.

/dev/random

Meta's new personal AI agent "Muse" launched this week with a feature set that includes reading your messages, browsing on your behalf, and taking actions across apps — which is either a productivity revolution or a perfectly engineered exfiltration endpoint, depending on your threat model. The genuinely interesting security detail: Muse uses a persistent "memory" layer that summarises and retains context across sessions, meaning prompt injection attacks don't just affect a single conversation — they potentially poison the agent's long-term behavioural context. Someone is already working on a malicious webpage that, when visited by Muse during an agentic browsing task, rewrites the agent's memory to believe its user has different preferences. We're calling it now: "memory poisoning" is going to be a vulnerability class that appears in a KEV feed before 2027.