Security Monitor 2026-10-11 - Weekend Edition

Informational security guidance for awareness and prioritization. Not certification, legal advice, or a guarantee of protection.

Executive Summary

CISA added five old but actively exploited flaws in ProFTPD, Apache Struts, ONLYOFFICE Docs, Strapi, and ISC BIND to its Known Exploited Vulnerabilities (KEV) catalog. They're tied to Flax Typhoon, a China-linked group working through contractor Integrity Technology Group, in a joint advisory from seven countries. The FBI and Justice Department seized seven domains and disrupted the group's MicroScan scanner and FishHub phishing platform. Federal agencies had until October 11th to fix them. If you run any of the five, patch now and hunt for SoftEther VPN persistence and Exchange password spraying.

ASOS says an attacker impersonated a trusted contact of one employee, got that employee's login, and used it to reach third-party platforms, including the app notification system that pushed a message to customers on October 6th. A group calling itself Xuanye claims it took data from ASOS's Snowflake instance and used an agentic marketing tool to pull customer records. Neither claim has been confirmed. Treat marketing and notification platforms as critical systems.

In Japan, five companies disclosed breaches covering up to about 37 million customer records, two of them through a ransomware attack on SoftBank's cloud subsidiary IDC Frontier.

Microsoft CEO Satya Nadella published an essay arguing that companies should treat AI models as insider risks and build in containment, including an emergency brake an authorized person can pull. It's framing, not an incident, but it's a major platform vendor saying the agent-permission problem out loud.

Headline Developments

1. CISA adds five exploited flaws to KEV under a seven-country Flax Typhoon advisory HIGH

Source: Security Affairs - U.S. CISA adds ProFTPD, ONLYOFFICE Docs, Strapi, Apache Struts, and ISC BIND flaws to its KEV catalog (October 11th). Also The Hacker News - Flax Typhoon Exploits Five Flaws as CISA Sets October 11 Deadline, The Hacker News - FBI Seizes 7 Domains, Disrupts Flax Typhoon Tools, Byteiota, and Innovation Network Design.
- The five: CVE-2015-3306 (ProFTPD mod_copy, CVSS 10.0, unauthenticated file read and write through SITE CPFR / SITE CPTO), CVE-2016-3081 (Apache Struts, 8.1, OGNL code execution through the method: prefix when Dynamic Method Invocation is on), CVE-2021-3199 (ONLYOFFICE Docs, 9.8, path traversal through an image upload parameter with JWT enabled), CVE-2023-22894 (Strapi, 7.2, admin-panel query filters expose password hashes and reset tokens), and CVE-2015-5477 (ISC BIND, 7.5, one malformed TKEY query crashes named).
- The joint advisory comes from Australia, Canada, Japan, New Zealand, Spain, the United Kingdom, and the United States. The campaign used eight flaws in total. The other three were already in KEV: Shellshock variant CVE-2014-6278, Pulse Connect Secure CVE-2019-11510, and GitLab CVE-2021-22205.
- Once inside, operators harvested credentials with cross-site scripting, password-sprayed Exchange and Microsoft 365, read mailboxes with scripts, and installed SoftEther VPN for persistent access.
- U.S. authorities seized seven domains and disrupted MicroScan (a scanner with more than 1,300 exploit and recon scripts) and FishHub (a spear-phishing platform). Secondary coverage of the advisory says the group pairs AI-assisted scanning with botnets and hands-on exploitation.
- Under BOD 22-01, federal civilian agencies had until October 11th. Private organizations have no mandatory deadline, but these are confirmed-exploited, long-patched bugs.

Why this matters: The rating is HIGH because exploitation is confirmed by seven governments and every fix has existed for years. The risk is the forgotten install: ONLYOFFICE behind a collaboration suite, Struts 2.3 in an old Java app, BIND or ProFTPD on an appliance or vendor VM nobody patches. The group isn't writing new exploits. It's running a large pre-built kit against the internet edge and keeping whatever answers.

2. ASOS breach: one employee's login led to customer data and a hijacked push notification HIGH

Source: Dark Reading - ASOS Breach Reveals the Risks in Customer-Facing SaaS (October 9th). Also CX Today (October 10th) and Sky News (secondary coverage).
- On October 6th, a group calling itself Xuanye Group announced the breach and sent an unauthorized push notification to ASOS app users. ASOS has about 17 million customers.
- ASOS's update says an attacker impersonated a trusted contact of one employee, obtained that employee's login, and used the access to reach some of the company's third-party platforms, including the mobile notification system.
- ASOS says names, contact details, and non-personal account information were taken, and that it doesn't believe payment cards or passwords were affected. The BBC, as cited by Dark Reading, reported addresses, phone numbers, emails, dates of birth, customer IDs, and search histories in the data.
- The attackers claimed they had "fully compromised" ASOS's Snowflake instance and used "Simon AI," an agentic marketing platform that runs inside cloud data platforms, to reach customer data. Neither ASOS nor independent researchers have confirmed this. Sky News reports the data firm named in the message has responded.
- The UK's NCSC told ASOS customers to assume they're affected and to be careful with links in messages that look like they come from ASOS.

Why this matters: The rating is HIGH because the breach and data exposure are confirmed by ASOS. The entry point wasn't a code flaw. It was one employee's login and broad access on a marketing platform where one account can message every customer with no second approval. If the agentic-tool claim holds up, it's another case of an agent's data access becoming the attacker's data access. Either way, patching doesn't fix this one: MFA, scoped roles, and alerts on bulk sends do.

3. Japan: five firms disclose breaches covering up to about 37 million records MEDIUM

Source: The Japan Times - More Japanese firms hit by large-scale data breaches (October 10th).
- JR East (about 6.09 million accounts) and JR Kyushu (about 1.3 million) both traced their exposure to a ransomware attack on IDC Frontier, SoftBank's cloud services subsidiary.
- Bookoff Group Holdings reported unauthorized access at the subsidiary that manages membership data (up to 6.43 million customers). Daiichikosho, which runs the Big Echo karaoke chain, cited unauthorized access to a system managed by an external contractor (about 8.72 million). Adventure Inc., which runs Skyticket, reported an attack on its servers and booking management system (about 14.64 million sets of data).
- At JR East the exposed data is email addresses. At the other three it's names, dates of birth, and phone numbers. Adventure says card and passport numbers weren't leaked. None of the companies has found evidence the data has been misused.

Why this matters: The rating is MEDIUM because these are "may have been compromised" disclosures with no operator patch to apply. The IDC Frontier detail is the lesson: one ransomware hit at a cloud provider turned into breach notices for two rail operators. When customer data sits with a cloud host or contractor, their incident becomes yours.

4. Satya Nadella: treat AI models as insider risks and build in an emergency brake CONTEXT

Source: sn scratchpad - Models as Insider Risks in the Super Intelligence Era (Satya Nadella, October 10th). Also Windows Forum, Ccleaks, and Dynamite News (October 11th).
- Nadella argues that because model behavior can't be traced to specific code the way traditional software can, companies should treat frontier and open-weight models like insider risks: capable, useful, and able to make mistakes or be compromised.
- He says responsibility can't be outsourced to the model provider, and that no single model should control both a system's behavior and the evidence used to judge it.
- Under his containment principle, an authorized person should always be able to pause or shut down a model, even mid-task. He also calls for incident disclosure that explains which controls failed.
- Microsoft's own agent guidance lists agent sprawl, over-privileged agents, tool misuse, misconfiguration, prompt injection, and data leakage as risks. CX Today reports Microsoft Execution Containers, which enforce file and network limits on agents at runtime, are now generally available.

Why this matters: This is framing, not an incident: no CVE and no breach. It matters because the CEO of a major agent platform is now saying in public what security teams have been saying for months: the agent's permissions are the risk, and the controls have to sit outside the model.

Continuity Footnotes

• AhsayCBS CVE-2026-105133 / CVE-2026-105134 - covered October 10th. Still no patch in today's sources; keep the management interface off the internet.
• GhostAction GitHub workflows - covered October 10th; nothing new today. Remove any security-audit.yml added since October 7th and rotate what it could see.
• LMCache CVE-2026-107204 and related cache flaws - covered October 10th. Today's sources don't confirm an LMCache fix.
• Anthropic cuts live internet for internal evaluations - covered October 10th; nothing new today.
• Zenity AgentCorruption, tensorlake 0.5.144, Warden Stealer, PCI SSC AI guidance - covered October 8th through 10th; nothing new today.
• KVM guest-to-host escape - covered October 7th; still waiting on a CVE and patch.

Pattern Analysis

Recommended Actions

Immediate (this week):
1. Find and patch every ProFTPD, Apache Struts, ONLYOFFICE Docs, Strapi, and ISC BIND install, including ones inside appliances and vendor VMs. ProFTPD 1.3.5a or later (or unload mod_copy), Struts 2.3.28.1 / 2.3.24.3 / 2.3.20.3 or later (or set struts.enable.DynamicMethodInvocation to false), ONLYOFFICE Document Server 5.6.3 or later, Strapi 4.8.0 or later (then rotate admin credentials), and a supported BIND 9.18 or 9.20 release.
2. Hunt for Flax Typhoon activity. Look for SoftEther VPN on servers that shouldn't run a VPN, Exchange and Microsoft 365 password-spray patterns, SITE CPFR / SITE CPTO in FTP logs, new PHP or JSP files in web roots, and method: or OGNL strings in web logs. Block the seized domains listed in the advisory.
3. If you ever ran a vulnerable Pulse Connect Secure build (CVE-2019-11510), rotate those credentials even if the appliance has since been patched.
4. Lock down marketing and notification platforms. Enforce MFA, scope who can message the full customer base, and alert on bulk sends.

This month:
5. Review which employees and agents can reach customer-data platforms (CRM, marketing automation, data warehouses), and remove access nobody needs.
6. Add an emergency stop to every production agent. Document what each agent can touch, who can pause it, and how.
7. Check breach-notification terms with your cloud hosts and contractors that hold customer PII.

Ongoing:
8. Track AhsayCBS and LMCache patch status and deploy as soon as fixes ship.
9. Watch KEV additions weekly, especially old bugs in software you didn't know you were running.

Relevant Risk Summary

Sources

• Security Affairs - U.S. CISA adds ProFTPD, ONLYOFFICE Docs, Strapi, Apache Struts, and ISC BIND flaws to its Known Exploited Vulnerabilities catalog
• The Hacker News - Flax Typhoon Exploits Five Flaws as CISA Sets October 11 Deadline for Federal Agencies
• The Hacker News - FBI Seizes 7 Domains, Disrupts Flax Typhoon Tools Used in Critical Infrastructure Intrusions
• Byteiota - Flax Typhoon Is Exploiting These 5 CVEs
• Innovation Network Design - Flax Typhoon Exploits Five Old Flaws as CISA Sets Weekend Deadline
• Dark Reading - ASOS Breach Reveals the Risks in Customer-Facing SaaS
• CX Today - What ASOS, AI Agents and 400+ Vulnerabilities Reveal About CX Risk
• Sky News - 'Snowflake' firm in ASOS hacking message issues response
• The Japan Times - More Japanese firms hit by large-scale data breaches
• sn scratchpad - Models as Insider Risks in the Super Intelligence Era
• Windows Forum - Satya Nadella Calls for AI Emergency Brakes and Independent Agent Controls
• Ccleaks - Satya Nadella Says Companies Should Treat AI Models as Insider Risks
• Dynamite News - Satya Nadella Says AI Needs an Emergency Brake

Informational security guidance. Not certification.