cybr.cx Daily Digest — October 05, 2026

Critical Vulnerabilities

⚠️ Actively exploited — CVE-2026-88771 | Citrix NetScaler ADC & Gateway
An improper input validation flaw allows unauthenticated attackers to execute arbitrary commands on Citrix NetScaler ADC and Gateway. With CISA's remediation deadline already passed (September 30), any unpatched instance should be treated as compromised. Patch or isolate immediately.

⚠️ Actively exploited — CVE-2026-88772 | Citrix NetScaler ADC & Gateway
A memory buffer bounds violation in the same NetScaler product line enables remote code execution or denial of service. This runs alongside CVE-2026-88771 in active exploitation campaigns — the two likely form a combined attack surface. No further excuses for delay here.

⚠️ Actively exploited — CVE-2026-88779 | Citrix NetScaler ADC & Gateway
A third NetScaler buffer vulnerability, added to the KEV on October 4 with a due date of October 7 — that's 48 hours from now. This one is DoS-focused but the pattern of three simultaneous NetScaler KEV entries strongly suggests coordinated threat actor activity against this platform. Treat the entire NetScaler estate as high-priority.

⚠️ Actively exploited — CVE-2026-104286 | Fortinet FortiMail
A path traversal combined with NULL byte injection allows unauthenticated attackers to write arbitrary files to the underlying system via crafted HTTP/HTTPS requests — effectively unauthenticated remote file write. FortiMail is a perimeter mail gateway; exploitation here can lead to full system compromise with no prior access required. Patch now.

⚠️ Actively exploited — CVE-2026-76504 | Cisco Catalyst SD-WAN Manager
A hex encoding mishandling flaw lets unauthenticated remote attackers access the SD-WAN Manager with full admin privileges by crafting specific URI-encoded HTTP requests. Admin-level access to SD-WAN infrastructure is as bad as it sounds — network-wide visibility and configuration control are on the table.

⚠️ Actively exploited — CVE-2026-102489 + CVE-2026-102490 | Zammad (Zammad GmbH)
These two chain together: CVE-2026-102489 is a session fixation flaw enabling remote code execution as the zammad user, while CVE-2026-102490 escalates that to root via improper privilege management. The CISA due date was today — if your organisation runs the Zammad helpdesk platform and hasn't patched, assume exploitation is occurring.

⚠️ Actively exploited — CVE-2026-86950 | Apple iOS, macOS, iPadOS
An out-of-bounds write in the CoreGraphics framework can lead to arbitrary code execution across Apple's major platforms. Given Apple's ubiquity in enterprise environments, especially among executive and developer populations, this deserves immediate attention. Ensure endpoint management policies are enforcing the latest OS updates.

CVE-2026-105210 | ZITADEL (3.x / 4.x) — CVSS 8.2
ZITADEL's hosted Login V1 UI allows second-factor enrollment before any primary factor is verified. An attacker who knows only a victim's login name can enroll attacker-controlled TOTP, SMS OTP, email OTP, or U2F factors — and overwrite the victim's verified phone number — locking them out entirely. Update to 3.4.15 or 4.17.1 respectively.

CVE-2026-105211 | ZITADEL before 4.17.1 — CVSS 8.1
An authentication bypass in Login V2 exposes OTP codes via the returnCode delivery mechanism in server-action responses. Anyone knowing a victim's login name can harvest both OTP-Email and OTP-SMS codes directly from server responses, achieving MFA-authenticated sessions — including administrator accounts. This one is particularly nasty for SaaS environments using ZITADEL as their identity provider.

CVE-2026-105212 | ZITADEL (3.x / 4.x) — CVSS 7.5
Both Login V1 and V2 accept passkey and authenticator enrollment on identify-only sessions before primary factor verification. An unauthenticated attacker with just a username can register their own authenticator and subsequently log in as the target user. Patch to 3.4.14 / 4.16.2.

CVE-2026-105123 | WCMS (vincent-peugnet/wcms) through 3.18.0 — CVSS 8.8
Authenticated editors can achieve remote code execution by uploading .php files via an unvalidated path parameter in the media upload API, using encoded ../ sequences to write outside the media directory entirely. A separate DELETE endpoint allows arbitrary file deletion. If WCMS is exposed to untrusted authenticated users — or if an editor account is compromised — this is a full server takeover.

CVE-2026-105086 & CVE-2026-105089 | WWBN AVideo 12.4–29.2.0 — CVSS 8.7
Two stored XSS vulnerabilities in the AVideo open-source video platform: one exploiting a double-decode flaw in the safeString() sanitiser via video titles, the other via malicious trailer URLs rendered unescaped in YouPHPFlix2 templates. Both require only upload permission and affect trending, gallery, embed, playlist, and channel pages — anywhere victims browse content.

Headline News

ShinyHunters Member "Rey" Detained in Jordan, Cooperating with FBI
A suspected member of the ShinyHunters cybercrime group, operating under the alias "Rey," has been detained in Jordan and is reportedly cooperating with FBI investigators to identify and locate other members of the group. ShinyHunters has been linked to some of the largest data theft and extortion operations of recent years, targeting cloud storage misconfiguration and credential stuffing at scale across numerous enterprise and consumer platforms. The cooperation could prove significant: threat groups of this structure tend to rely on operational trust between pseudonymous actors, and a cooperating insider dramatically raises the risk of internal exposure for remaining members. For defenders, this is a reminder that extortion group infrastructure and TTPs should be actively tracked — arrests often trigger retaliatory activity or hasty operational pivots by remaining members.

C2PA's Timestamp Mechanism Found Exploitable for Content Authentication Bypass
A detailed technical walkthrough has demonstrated that C2PA — the Coalition for Content Provenance and Authenticity standard increasingly used to authenticate AI-generated and photographic media — contains a weakness in how it handles time assertions that can be exploited to manipulate content provenance records. The attack leverages the fact that timestamp validation in C2PA manifests is not as tightly bound to the signed content as its trust model implies, allowing an attacker to craft or modify content in ways that survive authenticity checks. This matters considerably: C2PA has been positioned as a key technical defence against AI-generated disinformation and deepfakes, and exploitability undermines the trust guarantees the standard is built on. Practitioners working on media integrity pipelines or integrating C2PA into content verification workflows should review the technical details before placing production-grade trust in timestamp assertions alone.

Schrödinger's Feed

The quantum data feed came up empty today — fitting, in a way: the qubit both fired and didn't. Rather than speculate, we'll note that the silence itself is informative. Post-quantum cryptography timelines are no longer theoretical — NIST's PQC standards are finalised and migration planning is actively underway at major financial institutions and government agencies. If your organisation hasn't begun a cryptographic inventory to identify RSA and ECC dependencies, the absence of dramatic quantum headlines this week is not a reason to delay — it's a reminder that the transition window is finite and the work is slow. Practitioners should be tracking CISA's PQC migration guidance and nudging vendor conversations on algorithm support timelines.

/dev/random

A tool quietly appeared on GitHub this week designed to remove and disable Apple's macOS 27 on-device AI model components — local inference engines, background daemons, and associated ML assets — giving users who'd rather not have a multi-gigabyte neural network quietly resident on their filesystem a clean exit. The tool strips out the AI stack without touching core OS functionality, which is either a power-user's dream or a corporate endpoint hardening script waiting to happen. It's a neat inversion of the usual direction of travel: instead of adding security tooling, you're removing vendor-installed AI tooling for security (or just philosophical) reasons. Somewhere, an MDM administrator is already asking whether this can be deployed via configuration profile.