AhnLab Reviews October Breaches and AI-Agent Attacks on Peripheral Systems
Summary
AhnLab ASEC's October 2026 review says reported breaches in South Korea and Japan often involved exposed APIs, auxiliary systems and third-party services. It documents agentic-AI tools in attacks, notes some links remain unconfirmed, and recommends stronger perimeter1
Key points
- As of Oct. 14, South Korean financial authorities had identified 16 affected institutions and about 100,000 leaked personal records; causes of individual incidents were mostly still under investigation.
- Reported South Korean and Japanese breach points often included app APIs, reservation and management systems, logs, cloud files and outsourced services rather than core systems.
- ARTEX traces were reported in Korean financial-sector and Japanese-targeting activity. The article says the actors' identities and nationalities were unconfirmed and does not establish that the cases shared operators.
- Japan's NICT described App-to-Backend (A2B) attacks involving analysis of public apps to identify backend servers and APIs; it cited SSRF, RCE, SQL injection and IDOR/BOLA, but did not link specific companies to A2B attacks.
- The article summarizes unauthenticated RCE disclosures affecting PHP and GitLab, and RCE in Node.js. It reports PHP2Shell attack attempts before official disclosure and recommends patching and reviewing logs.
- In an AhnLab test, ARTEX identified a web-server upload function and ran a web shell without human intervention; with exposed ActiveMQ, it exploited CVE-2023-46604 and executed commands on four internal hosts.
- Recommendations include inventorying exposed assets, strengthening API authorization and MFA, reviewing third-party access, patching internet-facing software promptly, monitoring peripheral systems and deleting expired data.
Article Details
- Event Type
- A wave of reported data breaches in South Korea and Japan, alongside critical vulnerability disclosures and reported use of AI-agent tools in cyberattacks.
- Impact
- South Korean authorities had identified 16 affected financial institutions and approximately 100,000 leaked personal records as of October 14; an insurance-sector leak was reported on October 26. Separately, police said a suspect group used ARTEX and an AI coding agent to steal information from about 90 Korean companies, involving approximately 8.2 million records. Japanese cases included reported or potential leaks ranging from hundreds of thousands to tens of millions of records. The article also describes service disruption at a Japanese cloud provider affecting 495 clients and multiple serious software vulnerabilities, including unauthenticated remote code execution flaws. Some incident causes and the scope of damage remained under investigation.
MITRE ATT&CK
T1021.001 · Remote Desktop ProtocolAhnLab ASEC reported recovering attack-operation data that included internal RDP relaying.T1046 · Network Service DiscoveryAhnLab ASEC reported recovering attack-operation data that included internal network scan results.T1078 · Valid AccountsMicrosoft reported that Storm-0940 used credentials associated with a US IT managed services provider's cloud backup app to attempt logins to Microsoft 365 tenants.T1110 · Brute ForceMicrosoft reported that Storm-0940 conducted brute-force password attacks, generating up to 2,900 failed login attempts per minute.T1190 · Exploit Public-Facing ApplicationThe article describes attacks exploiting vulnerabilities and misconfigurations in publicly exposed web services and app backend APIs; ARTEX also exploited a known vulnerability in an exposed Apache ActiveMQ test environment.T1505.003 · Web ShellWeb shells were reportedly installed during AI-agent attack activity and in the article's ARTEX test, which executed a program as a web shell.
CVE
CVE-2023-46604In an environment where Apache ActiveMQ was exposed, ARTEX gained initial access via a known vulnerability (CVE-2023-46604) and proceeded to perform remote command execution on four internal hosts.CVE-2026-109618CVE-2026-109618, CVE-2026-109619 (CVSS 9.2) CVE-2026-109619CVE-2026-109618, CVE-2026-109619 (CVSS 9.2) CVE-2026-110264CVE-2026-110264 (CVSS 10.0) CVE-2026-111497CVE-2026-94300 in BIND was reported by a researcher at OpenAI, and the official Node.Js announcement credited CVE-2026-111497 to OpenAI Codex Security.CVE-2026-115560Cryptocurrency Cryptomining CVE-2026-109618 CVE-2026-109619 CVE-2026-110264 CVE-2026-111497 CVE-2026-115560 CVE-2026-19630 CVE-2026-91403 CVE-2026-96325 CVE-2026-97206 DarkWeb data-breach DeepSeek DoSCVE-2026-19630CVE-2026-19630 and 3 others CVE-2026-72898abusing administrative APIs through app analysis, exploiting an SQL injection vulnerability (CVE-2026-72898) in the BI tool Metabase, and installing web shells. However, since neither organization identifiedCVE-2026-75650and Claude—as well as traces of a publicly available PoC for an e-commerce platform vulnerability (CVE-2026-75650) exploitable without authentication and embedded IRC-based remote control code.CVE-2026-91403CVE-2026-91403 (CVSS 9.1) CVE-2026-94300Denial of Service, GSS-TSIG memory corruption (CVE-2026-94300, potential remote code execution) CVE-2026-96325CVE-2026-109619 CVE-2026-110264 CVE-2026-111497 CVE-2026-115560 CVE-2026-19630 CVE-2026-91403 CVE-2026-96325 CVE-2026-97206 DarkWeb data-breach DeepSeek DoS e-commerce Education 미국 Finance France generative-AICVE-2026-97206CVE-2026-110264 CVE-2026-111497 CVE-2026-115560 CVE-2026-19630 CVE-2026-91403 CVE-2026-96325 CVE-2026-97206 DarkWeb data-breach DeepSeek DoS e-commerce Education 미국 Finance France generative-AI GitLab GLM
Threat Actors
Vendors
Products
Apache ActiveMQIn an environment where Apache ActiveMQ was exposed, ARTEX gained initial access via a known vulnerability (CVE-2023-46604) and proceeded to perform remote command execution on four internal hosts.Claudeverified by our company contained configurations for various LLMs—including DeepSeek, GLM, Kimi, and Claude—as well as traces of a publicly available PoC for an e-commerce platform vulnerability (CVE-2026-75650)DeepSeekAnalysis indicates that the threat actors connected and used several large language models (LLMs), such as DeepSeek, GLM, and Grok, with ARTEX.GLMAnalysis indicates that the threat actors connected and used several large language models (LLMs), such as DeepSeek, GLM, and Grok, with ARTEX.GrokAnalysis indicates that the threat actors connected and used several large language models (LLMs), such as DeepSeek, GLM, and Grok, with ARTEX.Kimiversions verified by our company contained configurations for various LLMs—including DeepSeek, GLM, Kimi, and Claude—as well as traces of a publicly available PoC for an e-commerce platform vulnerabilityMetabaseAPIs through app analysis, exploiting an SQL injection vulnerability (CVE-2026-72898) in the BI tool Metabase, and installing web shells. However, since neither organization identified specific companies thatMicrosoft 365 Apps for Enterpriseof a US IT managed services provider’s cloud backup app to attempt logins to more than 2,250 Microsoft 365 tenants; in some cases, they gained privileged access to the cloud within three hours of beingNode.jswere disclosed one after another in widely used software such as web backend runtimes (e.G., PHP and Node.Js) and GitLab, and there were also numerous reports that AI agents, such as ARTEX, were used in actualPHPexecution were disclosed one after another in widely used software such as web backend runtimes (e.G., PHP and Node.Js) and GitLab, and there were also numerous reports that AI agents, such as ARTEX, were used inUnboundUnbound 1.22.0–1.26.1
Tools
ARTEXruntimes (e.G., PHP and Node.Js) and GitLab, and there were also numerous reports that AI agents, such as ARTEX, were used in actual breach operations.CairnAn investigation of exposed Cairn management screens revealed more than 1,100 attack projects in the largest clusterOpenAI Codex SecurityCVE-2026-94300 in BIND was reported by a researcher at OpenAI, and the official Node.Js announcement credited CVE-2026-111497 to OpenAI Codex Security.
Countries
ArgentinaA2B-Attack Agentic-AI AI-agent Apache-ActiveMQ API-abuse Argentina Arrest ARTEX autonomous-pentest banking BIND BOLA Brotli CAIRN China Claude Cloud credential-abuse Cryptocurrency Cryptomining CVE-2026-109618ChinaChina-linked group Storm-0940 attacks hundreds of organizations worldwide using ARTEXJapanThe large-scale data breaches disclosed in quick succession in South Korea and Japan in October 2026 mostly originated not from internal core systems, but from peripheral entry points such as externally exposed webSouth KoreaThe large-scale data breaches disclosed in quick succession in South Korea and Japan in October 2026 mostly originated not from internal core systems, but from peripheral entry points such as externally exposed webUnited StatesDuring the same period, breaches involving outsourced services and publicly exposed websites, as well as large-scale automated attacks using AI agents, were reported in the US, Europe, and South America.
Industries
Accounting and ERP softwareEnterprise accounting and ERP software providers E-commerceOutsourced e-commerce platform Educationenterprise accounting software providers, large churches, coffee franchises, universities, and education districts—the points of compromise were consistently access points outside core business systems, suchGovernmentfrom several million to tens of millions of records were made during the same period, and Japanese government agencies identified a method involving the analysis of publicly available apps to attack backendHealthcareIncludes breaches at at least 15 Argentine healthcare institutionsInsurancesavings banks, finance companies, mutual financial institutions, virtual asset service providers, and insurance companies—while data breaches were also confirmed in the non-financial sector, including corporateRetailN-day Node.js Node2Shell open-weight-model Password-spraying PHP PHP2Shell PKI Ransomware RCE Religion Retail SaaS Securities Software South-Korea sql-injection SSRF Storm-0940 supply-chain-attack third-party-risk