`CVE-2026-107858`

Severity	7.2 High CVSS 4.0
Weakness	Missing authentication (CWE-306)
Published	23 Oct 2026
Updated	2 Nov 2026

Published 22 days ago. NVD often adds the list of affected products days or weeks after a CVE is published, and the other sources we check lag too. The scope and status here may change.

Description

Missing Origin Validation in WebSockets (CWE-1385) and Permissive Cross-domain Policy with Untrusted Domains (CWE-942) in the development web server (ADK Web) in Google Agent Development Kit (ADK) for Java (com.google.adk:google-adk-dev versions 0.1.0 before 1.13.0 and com.google.adk:google-adk-maven-plugin versions 0.3.0 before 1.13.0) allow a remote attacker to control a locally running agent and read its data via a malicious web page visited by a developer while the server is running. The /run_live WebSocket endpoint accepts connections from any origin by default (hardcoded via setAllowedOrigins("*") in versions 0.1.0 through 1.5.0, and defaulting to * via adk.web.cors.origins in versions 1.6.0 through 1.12.0), and since version 0.4.0 the REST API returns Access-Control-Allow-Origin: * by default. Because the server has no authentication and requests originate from the developer's browser (so binding to localhost does not prevent exploitation), a cross-origin page can send messages to the agent (triggering registered tools), stream agent responses, and list, read, create, or delete sessions and artifacts. In versions 1.6.0 through 1.12.0, users can mitigate the issue by setting the adk.web.cors.origins property to an explicit allowlist of trusted origins (such as http://localhost:8000). The vulnerability is fixed in version 1.13.0, which enforces a same-origin default for CORS and the /run_live WebSocket endpoint, and rejects DNS-rebinding requests over loopback.

CWE-942CWE-1385

Exploitation status

Proof of concept only

A proof of concept exists, but no reviewed exploit catalogue lists this CVE.

CISA KEV not listed
VulnCheck KEV not listed
CISA Vulnrichment exploitation: poc CISA recorded that an exploit has been demonstrated. Automatable: no, technical impact: partial.
Indexed PoC none indexed
EPSS 0.36% chance in 30 days

Waratek protection

No exploit published

Within Waratek scope. In the sources we check, nobody has published exploit code for this CVE or reported it being exploited.

A public code fix exists: google/adk-java 1ae5078.

Protection none
Action unverified
Review automated Upstream data
Record active

Decided by	Exploit code for this candidate, published in a reviewed catalogue or a GitHub repository with 5 or more stars, or in use per CISA or VulnCheck; not only a CISA verdict, a forecast, or nothing at all
Finding	This CVE is not listed in CISA KEV or VulnCheck KEV, CISA Vulnrichment records no active exploitation, no reviewed exploit catalogue carries an entry for it, and no GitHub PoC repository for it has 5 or more stars. Nobody has published exploit code worth tracing, so it is not a candidate for a rule ahead of the ones where somebody has; an upstream fix path does exist, so the remediation half of the candidacy still holds. This is a statement about what has been published, not about whether the defect is reachable or serious.
Candidate fix	https://github.com/google/adk-java/commit/1ae5078072e13e101318e32833f1db323ef084e1 : a commit: the change itself

Decided by

Exploit code for this candidate, published in a reviewed catalogue or a GitHub repository with 5 or more stars, or in use per CISA or VulnCheck; not only a CISA verdict, a forecast, or nothing at all

Finding

This CVE is not listed in CISA KEV or VulnCheck KEV, CISA Vulnrichment records no active exploitation, no reviewed exploit catalogue carries an entry for it, and no GitHub PoC repository for it has 5 or more stars. Nobody has published exploit code worth tracing, so it is not a candidate for a rule ahead of the ones where somebody has; an upstream fix path does exist, so the remediation half of the candidacy still holds. This is a statement about what has been published, not about whether the defect is reachable or serious.

Candidate fix

https://github.com/google/adk-java/commit/1ae5078072e13e101318e32833f1db323ef084e1 : a commit: the change itself

References

URL	Tags
https://github.com/google/adk-java/commit/1ae5078072e13e101318e32833f1db323ef084e1	commit
https://github.com/google/adk-java/commit/5667d382ab5078a0d6e8f7184713a6bab767e004
https://github.com/google/adk-java/commit/c2b087cbaf44f44d20dc946d8f2391506a2fce3d
https://github.com/google/adk-java/issues/1257
https://github.com/google/adk-java/pull/1222
https://github.com/google/adk-java/releases/tag/v1.13.0
https://github.com/google/adk-java/releases/tag/v1.6.0

URL

Timeline

Published	2026-10-23	By the CVE Program.
NVD record modified	2026-11-02	NVD's own last-modified date for this record.

Published

2026-10-23

By the CVE Program.

NVD record modified

2026-11-02

NVD's own last-modified date for this record.

How VRT decided (18 of 24 rules ran)

18 of 24 rules ran. The rest didn't run because an earlier rule had already settled the CVE.

# Rule Outcome Status Detail

identity: Does this identifier describe a vulnerability at all? the first rule that matches decides

1 rejected no match CVE Program withdrawal of this identifier

scope: Is it a vulnerability Waratek protects? the first rule that matches decides

2 armr-patch-file no match An ARMR patch file on disk for this CVE

3 armr-secure-rule-file no match An ARMR security-rule template on disk for this CVE

4 manual-classification no match A researcher's recorded decision for this specific CVE

5 manual classification no match Manual classification of this CVE in the legacy dataset

6 non-java-upstream no match A root cause in non-Java code: an OS kernel, a C library, or a bundled runtime

7 oracle-component no match The defect being in a component of a tracked product the agent does not reach: JavaFX, the Java SE installer, the JInitiator browser plug-in, the native WebLogic proxy plug-in, or a native part of Oracle Application Server or Collaboration Suite (HTTP Server, Web Cache, Internet Directory, mail and calendar servers). Oracle gives these no CPE of their own and names them only in the description.

8 java-client-side no match Oracle's own risk-matrix note saying the CVE is reached only through untrusted code in the Java sandbox (applets, Web Start) or only through a client installation of Java, so a server running its own trusted code is not exposed

9 osv-maven no match An OSV.dev record of this CVE affecting a package published to Maven Central. When every such package is one business applications do not use (Minecraft mods and launchers, Fastjson) and no tracked product is named by CPE, it settles the CVE as Outside Waratek scope instead

10 cna-maven-package matched In scope CNA declares the affected component is a Maven artifact

11 known-cpe-product not reached An NVD CPE naming a product ARMR tracks

12 cna-cpe-product not reached A CPE the CNA published naming a tracked product. Oracle publishes these on advisory day, months before NVD.

13 oracle-advisory-family not reached A tracked product family named in Oracle's own risk matrix. CPE cannot carry a suite name, only a module.

14 cna-named-product not reached A tracked product named in words by the CVE's own CNA, on a record with no CPE yet. Apache and Eclipse publish these on advisory day, before NVD adds CPE data.

15 non-deployable-product not reached The only affected product being a sealed appliance image, where Java runs but no -javaagent can be attached

16 unmatched-product not reached No tracked product matched above, and none present at all

disposition: What do we do about one that is? every rule runs

17 patch-hint matched Mitigation candidate References include candidate fix URL (commit tier): https://github.com/google/adk-java/commit/1ae5078072e13e101318e32833f1db323ef084e1

18 open-source-maven skipped The affected package being open source on Maven, so the fix can be read

19 poc-derivable skipped An indexed exploit clearing the evidence bar to reverse-engineer, where there is no source to read

20 no-exploit skipped Nothing to build a rule from yet: no published exploit, no CISA record of exploitation and no source-level change to read

21 secure-rule-match no match A vulnerability class already blocked by an ARMR security rule

22 exploited-no-fix-or-poc skipped Somebody exploiting this CVE (a CISA or VulnCheck KEV listing, or a CISA verdict of active exploitation), with no fix, source code or PoC to build a rule from

23 below-action-threshold no match Severity is HIGH — only MEDIUM/LOW are deprioritized

24 no-exploit-published decided it matched No exploit published No exploit code has been published, and an upstream fix path does exist, so the remediation half of the candidacy still holds.

NVD ↗ · CVE.org ↗