AI-enabled attacks and supply chains led the day: ransomware crews exploited a Palo Alto GlobalProtect authentication bypass, a suspected North Korean crew backdoored a fake Terraform provider, lab testing showed frontier AI agents meeting every operational technology attack objective, and contractor malware exposed 8.7 million customer records at a Japanese karaoke operator.
🟠 HIGH (3)
Ransomware crews exploit PAN-OS GlobalProtect bug CVE-2026-0257 for unauthorized VPN access Confirmed
ReliaQuest’s October 11 update says several groups, including the Qilin ransomware operation and a newer crew called Settra, use CVE-2026-0257 to open VPN sessions without valid credentials on unpatched PAN-OS GlobalProtect and Prisma Access deployments. Palo Alto Networks disclosed the authentication bypass on May 13 with a CVSS score of 7.8 and has shipped fixes. Because the traffic arrives through the organisation’s own VPN it blends with normal remote work, so patch the portal and gateway, cut unnecessary exposure, and hunt for sessions from unexpected clients before credentials are stolen or ransomware is staged.
Sources CybersecurityNews|Palo Alto Networks|The Cyber Sec Guru
Contractor malware at Nippon Columbia exposes data on 8.7 million karaoke customers in Japan Confirmed
Daiichi Kosho, which runs the Big Echo karaoke chain, says malware on a computer at contractor Nippon Columbia Group, the company that handled its customer data, may have exposed about 8.63 million customer records and 93,000 employee records including names, dates of birth and email addresses. The company reported the incident to Japan’s data protection regulator and says no misuse has been confirmed so far. The case is a reminder to hold data-processing contractors to the same endpoint hardening, least privilege and bulk-export monitoring you apply in house.
Sources BleepingComputer|nippon.com
Suspected TraderTraitor crew hides backdoors in a fake AWS Terraform provider Under investigation
Zscaler ThreatLabz says a campaign it uncovered in July uses a trojanized HashiCorp Terraform provider disguised as an AWS plugin: the Go binary keeps working normally while a Bash loader fetches encrypted payloads hidden in decoy font files from a lookalike domain. The chain drops the FLATROOF and ROOFDECK backdoors on macOS, Linux and Windows, and steals browser credentials, crypto wallet extension data and cloud secrets, targeting crypto, Web3 and cloud engineers whose providers run on workstations and in CI/CD. The researchers attribute the activity to North Korea’s TraderTraitor group with limited confidence and note overlap with the KelpDAO incident; install providers only from trusted sources, verify checksums and rotate cloud and wallet credentials on any machine where the provider ran.
Sources Zscaler ThreatLabz|SOC Prime
🟡 MEDIUM (3)
Commune platform breach leaks about 307,000 member accounts at Japanese companies Confirmed
Tokyo-based Commune Inc. said an intruder abused invite links to private communities, registered as an ordinary member and impersonated administrators, viewing and overwriting member data between October 5 and 7. About 307,000 members are affected, roughly 126,000 of them with email addresses exposed, and customer communities run by Sansan and Panasonic reported leaks from their own spaces. No passwords were confirmed stolen, though Commune reset 144 accounts and told affected members to change any reused password.
Sources TSN Media|ITmedia NEWS|Sansan
Warden Stealer infostealer now harvests AI coding agent tokens and MCP credentials Confirmed
Gen Threat Labs reports that Warden Stealer, a Rust infostealer sold as malware as a service since around May 2026, has quickly become one of the most common stealers in its telemetry and now collects files from AI coding agents: Claude, Codex, Cursor and Grok configuration, MCP settings, access and refresh tokens, and conversation histories, on top of browser data and crypto wallet extensions. Version 1.9 at the end of September made agent collection a standard feature and raised monthly prices. Treat agent configuration directories as credential stores, scope and rotate agent tokens, and alert on unexpected access to those paths.
Sources Gen Digital|CybersecurityNews
Korean regulator had demonstrated an AI-agent attack weeks before the bank breaches Under investigation
Documents given to a National Assembly committee show South Korea’s Financial Security Institute walked about 160 financial-sector security officers, including staff from four of the firms later breached, through an AI-agent attack on a domestic lender at a September 17 seminar: the attacker pointed a DeepSeek-based agent at Oracle WebLogic servers and used a flaw to plant a web shell. The wider campaign hit seven institutions from late September, exposing data on about 40,000 Yegaram Savings Bank customers and 25,000 Shinhan Bank customers. The chief executives of the country’s five largest banks are due to testify at a Financial Supervisory Service audit on October 19.
Sources Aju Press|CrowdStrike
⚪ WATCH (3)
Lab test: frontier AI agents met every autonomous OT attack objective Confirmed
Booz Allen Hamilton’s operational technology lab ran two unnamed frontier models through eight attack scenarios against a simulated multi-vendor plant without giving them source code, engineering documents or OT guidance, and the agents achieved the objective in all eight. One agent moved from perimeter compromise to actions inside the control network in just over 16 minutes, recovered from a failed attack path to take over SCADA operator screens, and found and moved a collaborative robot arm in minutes, with human approval required before exploitation or physical action. The report argues that specialist OT knowledge is no longer a barrier, so segmentation, authenticated control paths and protocol-level monitoring matter more than obscurity.
Sources The Register|CISO Brief
Qilin claims Confipetrol after infostealer harvest of Microsoft 365 credentials Unconfirmed report
The Qilin ransomware group listed Latin American oil services firm Confipetrol as a victim on October 11, according to monitoring that links the intrusion to a Cavalier infostealer infection on employee endpoints. Data cited from HudsonRock covers 121 user accounts, 60 third-party credentials, 1,417 passwords and 46 session cookies. The pattern is familiar: infostealer logs stay a standard route into corporate SaaS, so reset exposed credentials, invalidate active sessions and force fresh authentication for Microsoft 365 and other cloud apps.
Sources QPulse
Zcash developers target quantum and AI key theft with hash-based signatures Confirmed
A Zcash full node developer told Decrypt that hash-based signatures are expected to land in Zcash in January as a hedge against quantum computers and AI-assisted attacks on cryptography, and that a tool for rotating transparent addresses is being rolled out this week. The work sits in a wider push across crypto projects to prepare for the day today’s public key cryptography is no longer safe. No incident is involved: these are forward-looking engineering plans.
Sources Decrypt
Methodology: compiled from vendor advisories, government feeds (CISA KEV, MSRC), security news sources and on-chain/security-firm alerts. Grouping: one incident, one entry, with every source cited. Unconfirmed reports are labeled. Crypto items are incident reporting, not investment advice. Corrections from prior digests are noted at the top when applicable.