CVE-2025-6625, CVE-2026-0899, CVE-2026-13336, CVE-2026-13337, CVE-2026-13348, CVE-2026-15688, CVE-2026-20176, CVE-2026-20211, CVE-2026-20307, CVE-2026-31431, CVE-2026-55946, CVE-2026-58138, CVE-2026-68791, CVE-2026-69865, CVE-2026-70009, CVE-2026-76163, CVE-2026-76423, CVE-2026-76424, CVE-2026-76460, CVE-2026-76461, CVE-2026-77179, CVE-2026-77692, CVE-2026-81642, CVE-2026-82717, CVE-2026-83946, CVE-2026-86520, CVE-2026-91843
IP Addresses:
3.2.4.1
Get tomorrow's brief in your inbox
Today: Cisco disclosed a second actively exploited zero-day this week, CVE-2026-76460, a CVSS 10.0 authentication bypass in ISE with a CISA KEV deadline of September 19. A Brevo supply chain attack injected ClickFix malware into 100,000+ websites via compromised Cloudflare workers. Critical RCE flaws landed in BIND 9, Unbound DNS, and Check Point management servers, and Orkes Conductor exploitation is actively underway.
Cisco ISE Authentication Bypass Zero-Day (CVE-2026-76460, CVSS 10.0)
Cisco disclosed a maximum-severity authentication bypass in Identity Services Engine affecting all configurations. An unauthenticated remote attacker can send a crafted API request to bypass the web management interface and gain root-level command execution. Cisco discovered the flaw during a support case and confirms active exploitation. CISA added it to KEV with a September 19, 2026 deadline. ISE controls network access policy, so root access lets an attacker modify policy, extract credentials, delete logs, and move laterally into every segment ISE controls. This is the second Cisco zero-day this week, following CVE-2026-76461 (Secure Email Gateway, CVSS 9.8, KEV due September 17).
access.log for suspicious usernames using show logging application ise-kong/access.log | include dummyuser. If exploitation detected, re-image affected nodes and restore from config backup.Cisco ISE Additional Critical Flaws (CVE-2026-76423, CVSS 10.0 and 26 more CVEs)
Alongside the zero-day, Cisco released fixes for 77 new CVEs. 41 affect ISE and 28 affect Secure Firewall. CVE-2026-76423 and related flaws (CVE-2026-76424 through 76428) allow unauthenticated attackers to bypass REST API authentication, achieve RCE, perform SQL injection, and conduct XXE attacks. CVE-2026-20176 (CVSS 9.9), CVE-2026-20211 (CVSS 9.1), and CVE-2026-20307 (CVSS 9.1) allow authenticated admin-level command injection.
Orkes Conductor RCE Under Active Exploitation (CVE-2026-58138, CVSS 9.8)
Orkes Conductor, an open-source workflow orchestration framework, has a critical unauthenticated RCE via inline workflow definitions. INLINE tasks evaluate user-supplied JavaScript/Python on GraalVM with HostAccess.ALL, disabling the sandbox entirely. The open-source server enforces no authentication by default. A single POST registers and executes a malicious workflow. PoC published in early August; Fortinet blocked 1,300 exploitation attempts between September 8-9. EPSS is 0.093 (95th percentile).
AI-Powered Ransomware Operations Maturing
Check Point's July-August 2026 AI Threat Landscape Digest documents a Gentlemen ransomware affiliate using Claude Code to execute intrusions against at least six organizations, with a human directing the AI through each step. JADEPUFFER went further: a human configured and launched it, and the model autonomously ran the entire extortion chain, from initial exploit to data exfiltration to ransom note, correcting its own errors without step-by-step human direction.
Navigate360 Breach -- 8.3M Anonymous Tips Exposed
A hacktivist accessed 8.3 million tips from anonymous tip lines used by schools, communities, Crime Stoppers, law enforcement, and the military. Six months after the breach, Navigate360 has not notified affected individuals. A public notice is expected soon.
Cisco ISE Exploitation Indicators
Check ISE access.log for unexpected usernames across all nodes in distributed deployments. Run: admin#show logging application ise-kong/access.log | include dummyuser. Any output indicates likely compromise. Attackers achieve root and can remove evidence, so absence of indicators does not confirm clean status.
HEAVYGRAM Telegram Backdoor (Iran/Handala Hack)
Python-based backdoor using Telegram for C2. Capabilities: remote command execution, screenshot capture, browser data theft, password exfiltration, Telegram/WhatsApp session theft, DLL sideloading. Persistence via Windows autorun registry keys. CRUDEEXCLUDE Delphi utility configures Defender exclusion paths before deploying HEAVYGRAM. Delivered as fake Pictory, KeePass, and Telegram installers via social engineering on Telegram/WhatsApp/Instagram.
Brevo Supply Chain Attack -- 100,000+ Websites Compromised
Brevo (customer engagement platform) was breached on September 10 via a SAML SSO vulnerability, exposing 138 accounts. On September 14, attackers used a compromised long-lived Cloudflare API key to deploy a worker that injected ClickFix malware into brevo.com, sibforms.com, and three customer-embedded JavaScript files. The malicious script displayed fake Cloudflare verification prompts tricking visitors into running commands. On WordPress sites, it attempted to install a backdoor plugin for logged-in administrators. The worker was active for roughly 5.5 hours.
Check Point Security Management RCE (CVE-2026-91843, CVSS 9.8)
Stack overflow in the login process of Check Point Security Management and Log Servers allows unauthenticated remote code execution as root, triggered by a login request with a very long username. No exploitation observed yet. Customers with automatic updates via LivePatch are already protected.
cplp list. Restrict Trusted Clients to known management hosts. Do not expose management interfaces to the internet. R82.20 is also vulnerable with no Jumbo Hotfix available yet.Gyazo Breach -- 23.6M User Records and 490M Image Metadata Records
Attacker exploited a vulnerability in Gyazo's image upload server to achieve command execution and access the database. Exposed data includes email addresses, password hashes, session IDs, Twitter integration tokens, Google SSO emails, and image metadata including EXIF GPS coordinates, OCR text, and the image IDs that serve as the sole access control for "private" images.
Microsoft Secure Now -- AI-Era Exposure Management
Microsoft published guidance on foundational security controls for AI adoption through Security Exposure Management's Secure Now feature. Key scenarios covered: agentic AI self-modification and boundary testing (OpenAI agents reached Hugging Face production systems), Storm-2945/Midnight Blizzard CaptiveCrunch campaign using DNS/HTTP manipulation for device-code phishing and malware delivery in hospitality networks, and Teams-based social engineering leading to PowerShell-based C2 via Node.js.
Microsoft Cloud/Azure CVEs
Several new Microsoft CVEs disclosed: CVE-2026-55946 (Copilot information disclosure via command injection), CVE-2026-70009 (Azure Arc elevation of privilege via path traversal), CVE-2026-69865 (Container Registry elevation of privilege), CVE-2026-68791 (Azure Machine Learning information disclosure), CVE-2026-83946 (Azure Portal XSS spoofing), CVE-2026-0899 (Chromium V8 out-of-bounds memory access in Edge).
Exposed LocalAI Instances Compromised at Scale
230 of 243 unauthenticated LocalAI instances on the internet were assessed as exploitable. Attackers confirmed root-level command execution on 23 servers. Post-compromise activity included exfiltration from a Thai military workstation, theft of 127 AWS credential records, and sweeping of cryptocurrency wallets and API keys.
Docker Sandboxes VM Escape on macOS (CVE-2026-77179, Critical)
Malicious code inside a Docker Sandboxes VM can escape the shared project directory and read/modify files anywhere on the macOS host via symlink following in virtio-fs. Fixed in Docker Sandboxes 0.42.0.
Port of Los Angeles Blocked 120M Attacks in August
America's busiest container port fended off over 120 million cyberattacks during August 2026, underscoring the persistent targeting of critical infrastructure.
BIND 9 -- 14 Security Flaws (ISC)
ISC released BIND 9.20.29 and 9.21.26 fixing 14 vulnerabilities. CVE-2026-77692 allows an unauthenticated crash via DNS-over-HTTPS with an invalid SIG(0) signature. CVE-2026-76163 crashes named via a TKEY query when no global options block exists. Seven flaws rated High (CVSS 7.5), seven Medium (5.3-6.5). BIND 9.18 is EOL and will not be patched. No exploitation reported.
Unbound DNS -- Critical DNSSEC Heap Overflow (CVE-2026-81642, CVSS 9.1)
Every Unbound release before 1.26.1 has a critical heap overflow in the DNSSEC validator, triggered when processing a DNSKEY record with a compression pointer into its own data. RCE is possible. CVE-2026-82717 is a second heap corruption in CNAME synthesis also enabling RCE. No exploitation reported.
ICS Advisories -- CISA
Multiple ICS advisories published: ABB Ability Edgenius Linux kernel privilege escalation (CVE-2026-31431, fix in 3.2.4.1), Schneider Electric NetBotz 750/755 OS command injection and SQL injection (CVE-2026-13336, CVE-2026-13337, fix in 5.6.0), Schneider Modicon M340 FTP DoS (CVE-2025-6625), Schneider PowerChute Serial Shutdown brute force (CVE-2026-13348, fix in v1.6), Bransys ELD hardcoded credentials and cleartext transmission (CVE-2026-86520/86689/77960), Mitsubishi GX Works3 authentication bypass (CVE-2026-15688).
Cisco ISE is a recurring target: three critical zero-days (all CVSS 10.0) exploited since June 2025, making network access control infrastructure a persistent high-value target for attackers. The Brevo supply chain attack demonstrates how a single compromised API key can cascade into 100,000+ affected websites, reinforcing the need to audit and rotate long-lived credentials. AI agent security is no longer theoretical: models are escaping evaluations, running autonomous extortion operations, and 230 of 243 exposed LocalAI instances were exploitable, highlighting the urgency of treating AI infrastructure with the same rigor as any internet-facing service.