๐จ ACTIVE EXPLOITATION
- UPDATE: Volexity: China-linked UTA0565 used shared exploit kit to hit Chrome and Windows zero-days
Volexity identified a new China-aligned threat group, UTA0565, exploiting a triple zero-day chain in Chrome and Windows using an exploit kit shared across multiple Chinese groups.- Targets included Asian government entities, media organizations, halal restaurant search sites and corporate training organizations.
- UTA0565 exploited CVE-2026-85046 and CVE-2026-87491 (Chromium JavaScript engine RCE) and CVE-2026-85880 (Windows ALPC privilege escalation) on Sept. 3-4, 2026, before patches were available.
- Delivery used fake websites and phishing emails, including spoofed domains impersonating the Center for American Progress and China Digital Times.
- The group deployed a previously undocumented malware family tracked as CLEANGULP.
- The same exploit kit was used by APT31, UNK_LateNight, UNK_DoubleCheck and UNK_QuietRacket, suggesting it was shared and weaponized across Chinese groups.
๐ฅ BREACHES & INCIDENTS
Shai-Hulud Worm Breach Exposes CrowdSec's Private GitHub Repositories
Attackers used the Shai-Hulud npm supply chain worm to steal an OAuth token from a former CrowdSec employee's machine and exfiltrate about 170 private GitHub repositories.- CrowdSec, a French security firm, had roughly 170 private GitHub repositories stolen in a May 22 attack discovered only on Sept. 16 when source code appeared on an underground marketplace.
- The Shai-Hulud worm compromised a former employee's computer via the TanStack npm supply chain attack, yielding a GitHub OAuth token that still had read access to CrowdSec's private repos.
- The exfiltration took only about nine minutes (5:52โ6:01 am UTC), traced to a Toronto IP linked to a BreachForum founder and member 'diencracked'.
- The departed employee's account remained in the GitHub organization after he left; CrowdSec removed it on May 25, three days after the incident.
- CrowdSec says the leak was limited to source code โ no infrastructure, databases, or build pipelines were compromised or altered.
BigCommerce Customer Data Stolen via Compromised Ribon App Credentials
Hackers used a compromised BigCommerce application key held by the Ribon app to steal customer data from hundreds of stores.- Affects merchants using the Ribon and Ribon 1.5 storefront optimization apps (developed by Be A Part Of, a Fastr company) installed on hundreds of BigCommerce stores.
- Attackers compromised Ribon's BigCommerce application key via a Fastr system compromise and used it September 13-17 to access customer data, including names, email addresses, phone numbers, and addresses.
- Data was downloaded 'page by page' until the key was revoked on September 17, a day after Ribon developers learned of the misuse; credentials were also used to inject malicious scripts into a small number of storefronts.
- BigCommerce says its platform was not breached; it uninstalled the Ribon apps from affected stores and began notifying merchants on September 18.
- How Ribon/Fastr was compromised remains unclear, and neither company has publicly acknowledged the incident.
๐ CVEs & KEV
CVE-2026-84412 โ Adobe Campaign Classic โ CVSS 10.0 โ Adobe Campaign Classic Code Injection Enables Remote Arbitrary Code Execution
CVE-2026-90898 โ maximhq Bifrost โ CVSS 9.8 โ Critical Bifrost AI Gateway Flaw Lets Attackers Run Commands Without Credentials
CVE-2026-43641 โ Softaculous Virtualizor โ CVSS 9.3 โ Softaculous Virtualizor OS Command Injection via Billing Module HandlerSoftac...
CVE-2026-43642 โ Softaculous Virtualizor โ CVSS 9.2 โ Softaculous Virtualizor PHP Object Injection via Billing Module HandlerSoftac...
CVE-2026-77243 โ sooperset mcp-atlassian โ CVSS 8.8 โ MCP Atlassian: ENABLED_TOOLS / Toolset authorization bypassMCP Atlassian is a...
CVE-2026-43643 โ Softaculous Virtualizor โ CVSS 8.7 โ Softaculous Virtualizor Authorization Bypass via Billing Module HandlerSoftac...
CVE-2026-77260 โ sooperset mcp-atlassian โ CVSS 8.3 โ MCP Atlassian: Arbitrary local file READ via unconstrained file_path in uploa...
CVE-2026-77271 โ sooperset mcp-atlassian โ CVSS 8.3 โ MCP Atlassian: Incomplete path traversal fix allows intra-CWD module overwrit...
CVE-2026-77251 โ sooperset mcp-atlassian โ CVSS 8.3 โ MCP Atlassian: JIRA_PROJECTS_FILTER / CONFLUENCE_SPACES_FILTER allow forbidde...
CVE-2026-77267 โ sooperset mcp-atlassian โ CVSS 8.3 โ mcp-atlassian has an incomplete SSRF remediationMCP Atlassian is a Model Cont...
CVE-2026-77242 โ sooperset mcp-atlassian โ CVSS 7.5 โ MCP Atlassian: Incomplete fix for CVE-2026-27826: DNS rebinding bypasses SSRF...
CVE-2026-77261 โ sooperset mcp-atlassian โ CVSS 7.1 โ MCP Atlassian: SSRF redirect protection missing for basic-auth and OAuth auth...
CVE-2026-77252 โ sooperset mcp-atlassian โ CVSS 6.5 โ MCP Atlassian: JIRA_PROJECTS_FILTER and CONFLUENCE_SPACES_FILTER can be bypas...
CVE-2026-77250 โ sooperset mcp-atlassian โ CVSS 6.1 โ MCP Atlassian: OAuth fallback token storage writes plaintext access and refre...
๐ต๏ธ RESEARCH & DEEP DIVES
BigDiskBuster Zero-Day PoC Blocks Microsoft Defender Updates by Filling Disk Space
A zero-day PoC tool called BigDiskBuster that stops Microsoft Defender from installing updates by filling the disk was published on GitHub.- Affects Microsoft Defender on all supported Windows versions; no patch, CVE, or Microsoft advisory exists.
- The tool watches Defender's update paths on C:, creates a hidden file filling all free space so platform and signature updates fail, then deletes it and waits for the next attempt.
- It also opens a handle on MRT.exe to block Windows Update from replacing the Malicious Software Removal Tool.
- Author Abdelhamid Naceri, a former Microsoft MSRC researcher dismissed in 2024, says it works like his earlier UnDefend flaw (patched as CVE-2026-45498) but uses a different disk-filling mechanism; his prior tools BlueHammer, RedSun, and UnDefend were exploited in live attacks.
- Naceri self-published the PoC on September 19 without coordinating with Microsoft; no independent researcher has confirmed the behavior.
ClosedQuorum Windows malware uses AI models to autonomously decide attack actions
Cisco Talos has documented ClosedQuorum, a Go-based Windows malware that uses multiple AI models to autonomously decide post-compromise actions.- Targets Windows hosts; first publicly documented implant delegating tactical C2 decisions to a panel of AI models (Gemini, DeepSeek, Qwen, Mistral)
- Acts with no human operator: reconnaissance data feeds a voting system; ties are broken by DeepSeek, then Qwen, Mistral, Gemini
- AI choices are limited to steal (LSASS dumping, Chrome/Edge/Firefox credential theft, crypto wallet extraction), inject (shellcode via process hollowing or Early Bird APC), persist, and move (no handler in analyzed build)
- Stolen data is exfiltrated to operators via a Discord webhook; analyzed binary contained placeholder API credentials and a dummy webhook
- Discovered via Cisco Talos' CAIRN toolkit; developer linked to 2025 criminal forum postings on carding; no confirmed in-the-wild deployment
Sources ๐ Source: blog.talosintelligence.com ยท ๐ Coverage: bleepingcomputer.com ยท ๐ via BleepingComputer
๐ ADVISORIES
- ๐ Source for Vendor-signed UEFI Shell apps allow Secure Boot bypass via memory-modify commands โ kb.cert.org