Loki Intelligence — Security Briefs · Published
Daily Security Brief: NetScaler SAML Zero-Day, Rejetto HFS and AI Agent Access
Edge and file-sharing servers are again under active attack, with a NetScaler SAML zero-day and a Rejetto HFS session forgery flaw both exploited, while AI agents and AI-generated bug reports are reshaping platform controls and disclosure programs. Teams should patch SAML-enabled NetScaler and HFS servers first, then review which AI agents and apps hold broad data access on managed devices.
Signal 01 · CISA
CISA Adds One Known Exploited Vulnerability to Catalog
CISA's October 4 KEV addition is CVE-2026-88779, a Citrix NetScaler memory buffer vulnerability with evidence of active exploitation. Federal agencies were given until October 7 to mitigate it, and CISA encourages other organizations to prioritize remediation of KEV-listed flaws as part of risk-based vulnerability management.
Why it matters: NetScaler has seen several exploited flaws in quick succession, so validate current build numbers on every appliance rather than relying on the last patch cycle.
Source: CISA
Signal 02 · The Hacker News
Apple Plans Tighter macOS Full Disk Access Controls Over AI Agent Data Access
Apple said it will tighten macOS Full Disk Access so the permission is granted only through explicit user action, warning that some developers use it in ways that expose files, mail, messages and browsing history as AI agents grow more autonomous. No timeline was given; the move follows reports about an agent app reading private messages.
Why it matters: Audit which applications, especially AI agents and assistants, hold Full Disk Access on managed Macs, and remove it wherever deep system access is not strictly required.
Source: The Hacker News
Signal 03 · BleepingComputer
Google halts open-source bug bounty program amid AI spam surge
Google paused new product vulnerability submissions to its Open Source Software Vulnerability Rewards Program after a surge of automated reports, most of them invalid. Supply chain reports and existing submissions are unaffected, alternatives such as the Patch Rewards Program remain open, and Google plans an update in Q1 2027.
Why it matters: Teams running disclosure programs should prepare triage capacity and intake rules for AI-generated reports so valid findings are not lost among low-quality submissions.
Source: BleepingComputer
Signal 04 · The Hacker News
Attackers Target Rejetto HFS Flaw That Enables Admin Session Forgery and RCE
VulnCheck detected exploitation attempts against CVE-2026-61500, a CVSS 9.3 flaw in Rejetto HTTP File Server 3.0.0 through 3.2.0 where a predictable session signing key allows forged administrator sessions and code execution. A fix shipped in version 3.2.1 in July, but a public proof of concept appeared in late September.
Why it matters: Upgrade HFS to 3.2.1 or later, remove internet exposure where file sharing is not required, and review admin configuration on any host that ran vulnerable builds.
Source: The Hacker News
Signal 05 · The Hacker News
New NetScaler Zero-Day Exploited in Targeted Attacks Can Knock SAML Deployments Offline
Citrix patched CVE-2026-88779, a CVSS 8.7 memory overflow in NetScaler ADC and Gateway exploited in targeted attacks against deployments configured as a SAML service provider or identity provider. Fixed builds include 14.1-73.41, 13.1-64.28 and matching FIPS and NDcPP releases, and CISA has added the flaw to KEV.
Why it matters: Search configurations for SAML SP or IdP entries, upgrade those appliances first, and prepare for disruption to identity-dependent remote access until patching is complete.
Source: The Hacker News
Signal 06 · BleepingComputer
Citrix patches NetScaler SAML zero-day exploited in attacks
BleepingComputer details how the NetScaler SAML zero-day surfaced: administrators saw recently patched appliances repeatedly crash and reboot, and one observed crafted login usernames attempting to fetch a payload. Citrix rates the impact as denial of service, but a researcher reported a patched honeypot running downloaded malware, so code execution remains under investigation.
Why it matters: Treat unexplained NetScaler reboots and authentication service crashes as possible compromise, preserve logs before rebuilding, and use Citrix's deny lists only as a stopgap alongside patching.
Source: BleepingComputer
Signal 07 · BleepingComputer
Anthropic asks Claude users to share voice data for AI model training
Claude now asks voice users whether Anthropic may use their audio recordings and voice chat data to improve its models. The choice is optional, controlled by a dedicated Privacy toggle that is independent of chat and Claude Code training preferences, and users can later revoke it and delete the data.
Why it matters: Include voice features in AI acceptable-use and data-handling reviews, and decide centrally whether staff may share audio from work conversations with AI providers.
Source: BleepingComputer
Brief sources
Related briefs
- What Is AI Agent Security Testing?
- AI Red Teaming vs Traditional Pentesting
- MCP Security Testing Checklist
Relevant Loki service: Orvyn — AI agent security private preview.