The Cyber Express reports three AI-assisted attack campaigns targeting Japan

What happened

The Cyber Express reports that researchers reconstructed three separate attack infrastructures targeting Japanese web systems. The researchers worked from files the attackers left in open directories. According to the outlet, all three are separate from earlier reports on ARTEX, an autonomous AI penetration-testing system (The Cyber Express).

The outlet also reports two warnings. Japan’s National Cybersecurity Office alerted businesses on October 9. The chair of the ruling party’s cybersecurity headquarters said cyberspace was in a “state of emergency” (The Cyber Express).

The Cyber Express says that without victim-side logs, the findings cannot confirm whether each attempt succeeded, how much damage was done, or who the attackers are. Victims are not named.

According to the report:

• Case A (October): The attackers abused CyberStrikeAI, an open-source agent-based penetration-testing tool. It ran 44 agents and produced Japanese forms and fake login pages.
  • At least 19 organizations received identical reports, including at least 16 in Japan and others in Taiwan and Hong Kong.
  • Many entries read “no session, no web shell, no exfiltration”.
  • The deepest intrusion reportedly began at an unauthenticated administration interface on a legacy system.
• Case B (about 24 days in August and September): A human operator used an LLM for planning and tooling.
  • The operator scanned at least 6,000 Japanese IP addresses and probed AI assets in South Korea.
  • Some tools reportedly sought to hijack other people’s AI services.
  • One company reportedly suffered full control through FTP, a database, administrator rights, a web shell and SSH, with backdoors restored after cleanup.
  • At another victim, a GIS server left on default credentials reportedly exposed records, Windows credentials and email.
• Case C (January to October): An operator ran at least six campaigns against at least 1,300 WebLogic candidates and 35 VPN entry points, mainly in Japan and South Korea.
  • Targeted devices included Cisco SD-WAN, Citrix NetScaler, F5 BIG-IP and Yamaha RTX1210.
  • On the RTX1210, leftover credentials in configuration files reportedly gave direct access to corporate networks.

The outlet says AI increased scale, speed and localization. It did not change the types of vulnerabilities exploited or the entry points. No AI-found zero-day was exploited in these cases (The Cyber Express).

The following details are not confirmed: whether the three operations share an operator, their origin, specific CVE IDs and the full victim list.

Am I affected?

You may be exposed if you run any of the system types the report names:

• Internet-facing legacy systems with unauthenticated admin interfaces.
• Servers left on default credentials, such as the GIS server described.
• WebLogic.
• Cisco SD-WAN, Citrix NetScaler, F5 BIG-IP or Yamaha RTX1210 devices.
• Exposed AI or GPU services.

Specific vulnerable versions are not confirmed.

What to do now

1. Inventory public-facing systems and update or isolate legacy ones. This is the recommendation the report gives (The Cyber Express).
2. The following steps are based on the entry points the report describes:
   • Check for admin interfaces that need no authentication and for default credentials, and close them.
   • Review Yamaha RTX1210 and other edge device configuration files for leftover credentials, and rotate any you find.
3. If you suspect a compromise, check whether web shells or backdoors return after cleanup, as described in Case B.

Sources

• The Cyber Express: Three Cyberattack Campaigns Target Japan, Exposing AI’s Growing Role