Apache Tomcat CVE-2026-34486: AI-Powered Attacks from China

Step by step
How the attack unfolded
Click a phase for details – or let the animation play through.
PHASE 1/5 · Reconnaissance

AI-driven autonomous enumeration automatically scans for vulnerable Apache Tomcat servers.

T1595 – Active Scanning T1595.002 – Vulnerability Scanning
  • Chinese-speaking actor knaithe / KnYuan (China-nexus)
  • AI model DeepSeek, orchestrated via 'Hermes Agent'
  • Over 540,000 publicly reachable Tomcat instances worldwide (Oct 2025)
  • Seven vulnerabilities targeted in total
PHASE 2/5 · Initial Access

Attackers bypass the EncryptInterceptor via the regression flaw CVE-2026-34486.

T1190 – Exploit Public-Facing Application
  • CVE-2026-34486 (CVSS 7.5, High) – regression of CVE-2026-29146
  • 'Missing Encryption of Sensitive Data' in cluster communication
  • Affected: Tomcat before 11.0.21, 10.1.54, 9.0.117
  • First exploitation from 2026-04-24 against servers in Taiwan
PHASE 3/5 · Execution

Java deserialization is abused to achieve remote code execution (RCE).

T1059 – Command and Scripting Interpreter T1203 – Exploitation for Client Execution
  • Technique: Java deserialization
  • CommonsCollections6 gadget using the 'ysoserial' tool
  • Result: Remote Code Execution (RCE)
  • Manual exploitation complements the automated AI phase
PHASE 4/5 · Persistence & Remote Access

Reverse shells and the SNOWLIGHT malware establish persistent remote access.

T1105 – Ingress Tool Transfer T1071 – Application Layer Protocol
  • Reverse shells on nine compromised Tomcat servers
  • SNOWLIGHT malware family (tracked by Google GTIG since 2024)
  • Loaders linked to access brokers UNC5174 and UNC6586
  • Payload retrieval via curl/wget (e.g. path '/slt', IP 130.94.30.168)
PHASE 5/5 · Impact

Attackers gain full server control with a high risk of data exfiltration.

T1486 – Data Encrypted for Impact
  • Full compromise of the server
  • Very high risk of unauthorized access to personal data
  • GDPR notification (Art. 33/34) generally mandatory
  • Kill-switch/payload files such as '/tmp/log_de.log' or 'c35ca3ebtcp'
Short & clear answers
Frequently asked questions about this incident
Am I affected by the Apache Tomcat vulnerability CVE-2026-34486?
All Apache Tomcat installations in versions prior to 11.0.21, 10.1.54, and 9.0.117 are affected, especially those using the EncryptInterceptor for cluster communication. Check your version via the Tomcat Manager dashboard or by running version.sh (Linux/macOS) or version.bat (Windows) in the bin directory. Since Tomcat is often set up by service providers and runs in the background, you should also ask your hosting provider.
What do I need to do right now to protect my server?
Immediately update to Apache Tomcat 11.0.21, 10.1.54, or 9.0.117 – this is the only truly effective measure. If an immediate update isn't possible, disable the cluster functionality if you don't strictly need it. Additionally, isolate your Tomcat servers from the public internet and restrict access via firewall to trusted IP addresses.
How can I tell if my server has already been hacked?
Examine your server logs for suspicious curl or wget requests (e.g. to "/slt") or connections to suspicious IP addresses such as 130.94.30.168. Also check for unexpected files like /tmp/log_de.log or c35ca3ebtcp, which are used by the attackers. If you find such indicators, assume a compromise and immediately bring in an IT security expert.
Do I have to report this incident under GDPR?
If attackers gained access to personal data through this vulnerability, you must report the breach to the competent supervisory authority under Article 33 GDPR, ideally within 72 hours. Because attackers can install reverse shells and gain full server control (RCE), there is a very high risk of data exfiltration, making a report generally mandatory. Affected users may also need to be informed under Article 34 GDPR.
Why is this attack particularly dangerous?
According to Palo Alto Networks Unit 42, this is an AI-enabled autonomous hacking campaign by a Chinese-speaking threat actor who uses an AI model (DeepSeek) to automatically scan for and attack vulnerable systems. This means attacks run semi-automatically and at large scale. Patches have been available since April 2026 – anyone who hasn't updated has been vulnerable for months.
More security news
You might also be interested in
Critical Elementor Pro Flaw: Attackers Can Fully Take Over WordPress Sites
A file upload bug in Elementor Pro up to 4.2.1 allows RCE without login. Version 4.2.2 fixes the flaw – update now!
Pods Plugin: Critical Flaw Enables Admin Takeover, No Login
A critical flaw in the WordPress plugin Pods lets attackers overwrite admin passwords without login. Over 100,000 sites are affected.
miniOrange SAML SSO: Critical Bypass Turns Attackers Into Admins
Two critical auth bypasses in the miniOrange SAML SSO plugin allow forged SAML assertions—up to full WordPress admin access.