CVE-2026-34486: Apache Tomcat EncryptInterceptor Bypass (CVSS 7.5) Added to CISA KEV — Patch to 11.0.21 Now

Apache Tomcat is running somewhere in your infrastructure. That statement is true for the overwhelming majority of enterprise and cloud environments — Tomcat remains one of the most widely deployed Java servlet containers in the world, underpinning everything from banking portals to e-commerce platforms to internal enterprise applications. Which makes CVE-2026-34486, added to CISA's Known Exploited Vulnerabilities Catalog on August 4, 2026, a vulnerability that belongs at the top of your patch queue this week.

The flaw has a CVSS v3.1 score of 7.5 (High) and is classified as a Missing Encryption of Sensitive Data vulnerability. But the practical risk exceeds what the classification suggests: under the right conditions, successful exploitation creates a pathway to unauthenticated remote code execution through Tomcat's clustering deserialization layer.

What Is the EncryptInterceptor and Why Does It Matter?

Apache Tomcat's clustering stack allows multiple Tomcat instances to share session state across nodes in a cluster — a standard pattern in load-balanced, high-availability deployments. The EncryptInterceptor is a component in this clustering stack that adds pre-shared key encryption to messages sent between cluster nodes. Its purpose is to ensure that cluster communication cannot be intercepted and tampered with, particularly in environments where cluster traffic traverses shared or untrusted network segments.

CVE-2026-34486 arises from an error in the fix for CVE-2026-29146, a prior vulnerability in the same clustering stack. The remediation for CVE-2026-29146 accidentally changed the EncryptInterceptor's failure mode from fail-closed to fail-open. In the corrected design, if encryption cannot be applied, the interceptor should reject the message. In the broken builds, the interceptor silently passes the unencrypted message downstream. Attacker-controlled content that should have been blocked by the encryption check can instead continue through the interceptor chain and reach the deserialization layer.

Deserialization of untrusted data is a well-understood critical risk in Java applications. When an attacker can deliver arbitrary content to a deserialization endpoint — as this vulnerability allows in clustering deployments — the result is typically remote code execution under the privileges of the Tomcat process.

Affected Versions

Exactly three Tomcat builds are vulnerable: 11.0.20, 10.1.53, and 9.0.116. These are the versions that included the flawed fix for CVE-2026-29146. All earlier and all later versions are outside this vulnerability's scope.

If your Tomcat deployment is not one of these three exact versions, you are not affected by CVE-2026-34486. However, if you cannot immediately confirm your Tomcat version across your environment — which is a common situation in organisations with sprawling application estates — treat the environment as potentially vulnerable until you have verified.

Remediation

Patches are available and straightforward. Upgrade to Apache Tomcat 11.0.21, 10.1.54, or 9.0.117 depending on your current branch. Apache's security advisory for CVE-2026-34486 was published alongside the patched releases and is available on the Apache Tomcat security pages for each supported version.

CISA's KEV listing carries a remediation deadline for federal agencies, but the presence of confirmed active exploitation means every organisation — not just federal entities — should treat this as an urgent patching priority. CISA's guidance is unambiguous: CVE-2026-34486 must be patched.

Who Is Exploiting This and How?

Active exploitation of CVE-2026-34486 has been attributed to a Chinese-speaking threat actor operating under the aliases knaithe and KnYuan, based in Zhuhai, China. What makes this campaign operationally significant beyond the vulnerability itself is the tooling being used: the attacker is leveraging DeepSeek via the Hermes Agent framework as an autonomous offensive operator. Researchers from StrigaAI identified this as an AI-enabled autonomous hacking campaign in which the agent conducts reconnaissance, identifies vulnerable Tomcat instances, and delivers exploitation payloads without continuous human direction.

This is not an isolated incident. Earlier this week, CISA also added CVE-2026-9198 (IBM Langflow Code Injection) and CVE-2026-18556 (N-able N-central Authentication Bypass) to the KEV catalog in the same alert — both involving infrastructure-level vulnerabilities being exploited at scale. The convergence of AI-assisted tooling with widely deployed middleware like Tomcat means that the window between vulnerability publication and mass exploitation continues to shrink.

Attack Conditions and Scope

CVE-2026-34486 only affects Tomcat deployments where the clustering stack is active and EncryptInterceptor is configured. Tomcat instances running as standalone application servers without cluster configuration are not exposed through this specific attack path.

However, two important caveats apply. First, cluster configuration is often applied at a infrastructure level and developers may not know their applications are running in a cluster mode. Confirming your deployment topology requires checking the server.xml and cluster configuration files, not just the application code. Second, active exploitation has been confirmed — meaning attackers are already scanning for and successfully targeting vulnerable instances. Clusters where you assumed nobody would look are not safe.

Recommended Actions

Start by identifying all Apache Tomcat instances in your environment, including those bundled within application servers, embedded in vendor products, or deployed via containerised workloads. Version discovery should cover production, staging, development, and any customer-facing environments. Tools that cannot enumerate bundled Tomcat instances in fat JARs or WAR archives frequently miss a significant percentage of exposed endpoints.

Once you have identified instances running 11.0.20, 10.1.53, or 9.0.116, prioritise upgrades based on exposure: internet-facing clustered deployments first, internal clustered deployments second. For deployments where immediate upgrading is not operationally feasible, disabling the EncryptInterceptor (and with it, all cluster encryption) removes the specific vulnerable code path — but this must be paired with network-level controls to restrict cluster traffic to trusted cluster nodes only.

After patching, review cluster traffic logs for anomalous inter-node communication patterns from the period before the patch was applied. Given the active exploitation of this vulnerability, organisations running affected builds should assume potential exposure and conduct an investigation rather than assuming no compromise occurred.

Detection Guidance

In environments running vulnerable builds, look for unusual cluster communication patterns: traffic arriving at clustering ports from unexpected source IPs, deserialization exceptions in Tomcat logs that do not correspond to normal application behaviour, and new or modified files in Tomcat's web application deployment directories following inter-node traffic spikes. If your SIEM has Apache Tomcat cluster port visibility (typically port 4000 or the configured cluster port), alert on connections from sources outside the known cluster node set.

Alastor Pulse's continuous asset inventory maintains version-level visibility across your application stack, flagging new CISA KEV additions against your deployed software within hours of catalog updates. If CVE-2026-34486 appears on Alastor Pulse before your patch cycle can respond, the finding includes affected host counts, exposure severity context, and remediation guidance linked directly to the affected assets. Reach out to see how Alastor Pulse handles vulnerability-to-asset correlation at scale.

The Broader Context: AI-Exploited Vulnerabilities Are Changing Patch Timelines

The threat actor behind CVE-2026-34486's active exploitation is notable not just because of what they targeted, but because of how they operated. Using an AI agent framework to autonomously identify, target, and exploit vulnerable Tomcat instances at scale means that the dwell time between vulnerability publication and widespread exploitation is now measured in hours for high-value targets, not days or weeks.

This changes the calculus for patch prioritisation. CISA KEV additions are no longer an indicator that exploitation has started — by the time a vulnerability reaches KEV, autonomous exploitation at scale is already underway. For any software in the Tomcat tier of criticality, patch timelines should be measured in hours for the affected builds, not the next maintenance window.

The response to AI-assisted attacks is not purely defensive patching — it also requires AI-assisted detection. Enforster AI's continuous scanning stack includes detection signatures updated within hours of confirmed KEV additions, ensuring that exploitation attempts against patched and unpatched instances are surfaced in real time. Visit Enforster AI to learn more.

CVE-2026-34486 is a reminder that a CVSS 7.5 score understates real-world risk when the vulnerable component sits at the centre of your application cluster's trust boundary — patch to 11.0.21, 10.1.54, or 9.0.117 today.

For a comprehensive vulnerability assessment of your Apache Tomcat deployments, including version discovery across containerised and on-premise environments, contact [email protected].