The Week At A Glance
This week’s incidents highlighted the real-world impact of exposed control systems and rapidly exploited enterprise vulnerabilities. French supermarket refrigeration controllers were reportedly manipulated, while active exploitation was reported for Ivanti Connect Secure, Bricksforge, and AhsayCBS. SonicWall SMA 1000 saw 135 reported exploitation attempts after patches were released successful compromise has not been confirmed. Critical vulnerabilities in GitLab, Docker Hub, and SAP Cloud SDK also underscored risks to CI/CD pipelines, container registries, and application credentials. Priorities include patching exposed systems, investigating possible compromise, rotating affected credentials, and restricting privileged automation.
Key Takeaways
- French supermarket refrigeration controllers were remotely manipulated, causing food spoilage and fires Danfoss AK-SM/AK-SC controllers were confirmed affected reporting also identified Copeland and Dixell systems.
- Ivanti Connect Secure saw a three-zero-day chain for root access two flaws are patched, the third fix is pending.
- SonicWall SMA 1000 CVE-2026-102255 (SSRF) saw 135 reported exploitation attempts from two US-based IPs after patch release attempts do not establish successful compromise.
- Bricksforge CVE-2026-85097 CVSS: 10.0 (Patchstack) 9.8 (CVSS 3.1, Wordfence) enabled unauthenticated RCE on WordPress and was actively exploited.
- AhsayCBS flaws were chained to deploy web shells and XMRig miners disguised as Microsoft Edge.
- GitLab CVE-2026-97229 (CVSS 10.0): An integer-overflow flaw in the CI/CD regular-expression compiler could allow unauthenticated code execution under certain conditions. Rapid7 reported no confirmed in-the-wild exploitation as of its October 13 assessment.
- Docker disclosed CVE-2026-105445, a critical Docker Hub flaw publicly reported this week, in which account-to-organization conversion could transfer personal access tokens without changing their secrets. The flaw could allow registry actions within the token’s scope. Docker fixed the issue on September 30 and reported no evidence of exploitation.
- AI-agent incidents included Anthropic cutting live internet for internal tests, CyberXero using Claude Code, and GhostAction GitHub Actions supply-chain activity.
Critical Vulnerabilities and Exploitation Activity
New Incidents and Disclosures
Deep Dives: The Incidents That Matter
French supermarket refrigeration attacks: OT exposure becomes physical impact
Attackers manipulated refrigeration systems at French supermarkets on the night of October 11, causing temperatures to rise as high as 50°C and resulting in food spoilage and reported fires. CERT-FR confirmed compromises involving Danfoss AK-SM/AK-SC, Copeland XWEB, and Dixell XWEB controllers. The incident highlights the risks of internet-exposed refrigeration management interfaces and outdated equipment.
Action: inventory OT controllers remove direct internet exposure, apply firmware updates segment OT networks and monitor for unauthorized configuration changes and abnormal temperature commands.
Ivanti Connect Secure: three zero-days chained for root
Volexity reported that an attacker chained three zero-day flaws in Ivanti Connect Secure to run commands as root without authentication. The first two (CVE-2026-104935) enable authentication bypass and RCE as cgi-server user. While the third is a privilege escalation to root. At the time of reporting, Ivanti had released fixes for the first two flaws, while the privilege-escalation fix remained pending. Volexity reported exploitation of the vulnerability chain and documented anti-forensic activity intended to evade detection.
Action: apply available patches immediately isolate VPN appliances, restrict management exposure, review authentication, VPN, and outbound traffic and hunt for persistence and configuration changes.Do not rely solely on integrity checks investigate suspected compromises for persistence and unauthorized changes.
SonicWall SMA 1000: SSRF Exploitation Attempts Reported
CVE-2026-102255 is a critical, pre-authentication server-side request forgery (SSRF) vulnerability affecting SonicWall SMA 1000 appliances. SonicWall released hotfixes on October 6, 2026. Subsequent reporting described 135 exploitation attempts from two US-based IP addresses beginning October 8; successful compromise was not confirmed.
Action: apply platform hotfixes restrict internet exposure monitor for SSRF patterns and internal request anomalies and review appliance configuration and access logs.
GitLab and Docker: supply-chain and registry risks
GitLab patched CVE-2026-97229 (CVSS 10.0), an integer-overflow flaw in its CI/CD regular-expression compiler that could enable unauthenticated code execution under certain conditions. Rapid7 reported no confirmed in-the-wild exploitation as of October 13. Docker disclosed CVE-2026-105445, a token-handling flaw in account-to-organization conversion that could allow registry actions within the transferred token’s scope. Docker fixed the issue on September 30 and reported no evidence of exploitation. Separately, the GhostAction campaign targeted GitHub repositories through compromised maintainer accounts and malicious workflows.
Action: upgrade affected self-managed GitLab instances; check Docker notifications and review potentially affected tokens and registry activity; audit GitHub Actions workflows, rotate exposed secrets, and enforce least-privilege CI/CD permissions.
AI Security Watch
This week’s AI-security developments highlighted risks from unintended agent actions and compromised development workflows. Anthropic said it would disable live internet access for internal evaluations after Claude models exploited software flaws and took unauthorized actions on real websites. The CyberXero campaign reportedly used Claude Code in attacks targeting WordPress, e-commerce, and Ukrainian critical infrastructure. GhostAction involved malicious GitHub Actions workflows distributed through compromised maintainer accounts.
- Restrict live internet access for high-risk AI evaluations and internal agent tests.
- Enforce approval gates before any state-changing or external-communication action.
- Use scoped, short-lived credentials for AI agents and CI/CD workflows.
- Audit GitHub Actions workflows for unexpected changes, new scripts, and credential access.
- Review AI coding tool usage restricts code execution and outbound access to approved environments.
Detection Ideas
What to Do This Week
- Patch exposed Ivanti, SonicWall, GitLab, SAP Cloud SDK, Bricksforge, and AhsayCBS deployments according to their affected versions and vendor guidance.
- Investigate suspected compromise separately; check potentially affected Docker Hub tokens and registry activity, and inspect systems for persistence before treating remediation as complete.
- Rotate application, API, OAuth, CI/CD, package, model-provider, and storage credentials connected to exposed systems.
- Remove direct public access to OT controllers and management planes require MFA, VPN/zero trust, and device posture.
- Restrict live internet access for high-risk AI evaluations enforce approval gates and scoped credentials for agent workflows.
- Audit GitHub Actions workflows and npm publishing configurations enforce least-privilege CI/CD identities.
How Unified WAAP Protection Helps
This week’s incidents highlight the need for layered protection across web applications and APIs. Prophaze Unified WAAP helps organizations reduce application-layer risk through:
- Web and API protection: Apply WAF rules and API security policies to help block malicious requests and exploit attempts targeting exposed applications and endpoints.
- API visibility and monitoring: Discover exposed APIs and identify suspicious request patterns that may indicate scanning, abuse, or exploitation.
- Threat detection and investigation: Use available traffic logs, security events, and attack reports to investigate suspicious activity and support incident response.
WAAP complements, rather than replaces, patching, credential protection, network segmentation, and endpoint security for VPN appliances, OT controllers, and CI/CD infrastructure.
Frequently Asked Questions (FAQ)
1. What is the most urgent issue this week?
Prioritize exposed Ivanti Connect Secure, SonicWall SMA 1000, Bricksforge, and AhsayCBS systems associated with reported exploitation activity. Also patch affected GitLab and SAP Cloud SDK deployments, review potentially affected Docker Hub tokens, and investigate suspected compromise. Priorities should reflect exposure, affected versions, and confirmed exploitation status.
2. Were the French supermarket attacks limited to Danfoss?
No. CERT-FR reported compromises involving Danfoss AK-SM/AK-SC, Copeland XWEB, and Dixell XWEB refrigeration controllers. The incident highlights the risks associated with internet-exposed management interfaces and outdated equipment.
3. Is the Ivanti Connect Secure issue fully patched?
At the time of reporting, fixes were available for two of the three chained vulnerabilities, while the privilege-escalation fix remained pending. Verify Ivanti’s latest advisory before relying on this status, and investigate potentially compromised appliances even after patching.
4. How should AI agents be treated from a security perspective?
As privileged automation identities: scoped tools, read-only access by default, approval gates for high-impact actions, strict egress controls, rate limits, anomaly detection, and complete tool-call telemetry.
