CVE-2026-105785: Joplin Server: reset-token reuse enables account takeover

MEDIUMPoC AVAILABLECISA: TRACK*

Published October 5, 2026

CISO Take

Joplin Server before 3.7.2 stores CSRF, account-confirmation, email-change, and password-reset tokens without a purpose, so any of them is accepted by the public password-reset endpoint. An attacker who obtains a victim's CSRF or confirmation token through a separate disclosure channel can reset the victim's password, and doing so also deletes the victim's existing sessions and API applications. This is a medium-severity issue (CVSS 4.8) because it requires a prior token leak (high attack complexity); EPSS is low at 0.00188 (the top-92% percentile label is inconsistent with that raw value). There is no KEV listing, no public exploit, and no Nuclei template. Joplin is a note-taking app, not an AI/ML framework, so the AI-specific exposure is limited to self-hosted instances that hold notes used as AI knowledge sources or that integrate with AI tooling via API applications. Upgrade Joplin Server to 3.7.2 or later and rotate credentials on any instance where tokens may have leaked.

Sources:NVDEPSSGitHub AdvisoryATLAS

What is the risk?

Medium-to-low risk. Exploitation requires network access plus a separately obtained CSRF or confirmation token, which is why attack complexity is high. Confidentiality and integrity impact are low (C:L/I:L) and availability is unaffected. No public exploit, no KEV listing, and a very low EPSS (0.00188) indicate low near-term likelihood. Risk rises for internet-exposed Joplin Server instances that have token-leaking components such as logs, referrers, or proxies. The AI/ML relevance is tangential: Joplin is classified here as an AI-adjacent tool and is not an ML component.

How does the attack unfold?

Token Acquisition

The attacker obtains a victim's CSRF or confirmation token through a separate disclosure channel such as logs or referrers.

AML.T0055

Exploitation

The attacker submits the token to the public password-reset endpoint, which accepts it because tokens carry no purpose.

AML.T0049

Account Takeover

The password is replaced and the attacker logs in as the victim.

AML.T0012

Impact

Existing sessions and API applications are deleted, locking out the user and exposing or altering stored notes.

AML.T0036

Token Acquisition

The attacker obtains a victim's CSRF or confirmation token through a separate disclosure channel such as logs or referrers.

AML.T0055

Exploitation

The attacker submits the token to the public password-reset endpoint, which accepts it because tokens carry no purpose.

AML.T0049

Account Takeover

The password is replaced and the attacker logs in as the victim.

AML.T0012

Impact

Existing sessions and API applications are deleted, locking out the user and exposing or altering stored notes.

AML.T0036

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
joplin	—	—	No patch

Do you use joplin? You're affected.

How severe is it?

CVSS 3.1

4.8 / 10

EPSS

0.2%

chance of exploitation in 30 days

Higher than 14% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

Exploit Available

Exploitation: MEDIUM

Sophistication

Moderate

Exploitation Confidence

medium

○CISA SSVC: Public PoC

○Public PoC indexed (trickest/cve)

Composite signal derived from CISA KEV, VulnCheck KEV, CISA SSVC, EPSS, Metasploit, Exploit-DB, trickest/cve, Nuclei templates, and inthewild.io exploitation reports.

What is the attack surface?

AVNetwork

ACHigh

PRNone

UINone

SUnchanged

CLow

ILow

ANone

What should I do?

1 step

1) Upgrade Joplin Server to 3.7.2 or later (fix in commit 7766eef, PR 16274). 2) Until patched, restrict access to the server's password-reset and confirmation endpoints, and put the instance behind a VPN or an authenticating reverse proxy. 3) Review logs, referrer headers, and proxy captures for leakage of CSRF or confirmation tokens, and purge them. 4) Audit for unexpected password resets or sudden session and API-application deletions, and force a credential rotation and re-authorization of integrations for affected users. 5) Enable MFA at the proxy or SSO layer if available. 6) Re-validate any AI or RAG pipelines that ingest Joplin notes if an account compromise is suspected.

What does CISA's SSVC say?

DecisionTrack*

Exploitationpoc

AutomatableNo

Technical Impactpartial

Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.

How is it classified?

Auth BypassData LeakageAPIPluginAML.T0012- Valid Accounts AML.T0049- Exploit Public-Facing Application

Which compliance frameworks are affected?

This CVE is relevant to:

ISO 42001

A.7.4-Quality of data for AI systems

NIST AI RMF

MANAGE 2.3-Responses to previously unknown risks are managed

OWASP LLM Top 10

LLM03-Supply Chain

Frequently Asked Questions

What is CVE-2026-105785?

Is CVE-2026-105785 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2026-105785, increasing the risk of exploitation.

How to fix CVE-2026-105785?

1) Upgrade Joplin Server to 3.7.2 or later (fix in commit 7766eef, PR 16274). 2) Until patched, restrict access to the server's password-reset and confirmation endpoints, and put the instance behind a VPN or an authenticating reverse proxy. 3) Review logs, referrer headers, and proxy captures for leakage of CSRF or confirmation tokens, and purge them. 4) Audit for unexpected password resets or sudden session and API-application deletions, and force a credential rotation and re-authorization of integrations for affected users. 5) Enable MFA at the proxy or SSO layer if available. 6) Re-validate any AI or RAG pipelines that ingest Joplin notes if an account compromise is suspected.

What systems are affected by CVE-2026-105785?

This vulnerability affects the following AI/ML architecture patterns: Self-hosted note-taking servers, Personal knowledge bases feeding RAG pipelines, API-integrated AI assistants.

What is the CVSS score for CVE-2026-105785?

CVE-2026-105785 has a CVSS v3.1 base score of 4.8 (MEDIUM). The EPSS exploitation probability is 0.23%.

What is the AI security impact?

Affected AI Architectures

Self-hosted note-taking serversPersonal knowledge bases feeding RAG pipelinesAPI-integrated AI assistants

MITRE ATLAS Techniques

AML.T0012Valid Accounts

AML.T0049Exploit Public-Facing Application

Compliance Controls Affected

ISO 42001: A.7.4

NIST AI RMF: MANAGE 2.3

OWASP LLM Top 10: LLM03

What are the technical details?

Original Advisory

Joplin is an open source note-taking and to-do application that organises notes and lists into notebooks. Prior to Joplin Server 3.7.2, packages/server/src/models/TokenModel.ts stores CSRF, account-confirmation, email-change, and password-reset tokens without a purpose, and packages/server/src/models/UserModel.ts allows UserModel.resetPassword to accept any token returned by TokenModel.userFromToken. An attacker who obtains a victim's CSRF or confirmation token through a separate disclosure channel can submit it to the public password-reset endpoint, replace the victim's password, and cause the existing sessions and API applications to be deleted. This issue is fixed in Joplin Server 3.7.2.

Exploitation Scenario

An attacker obtains a victim's CSRF or account-confirmation token through a separate leak, for example a misconfigured proxy log or an exposed referrer. They submit that token to the public password-reset endpoint of a self-hosted Joplin Server, set a new password, and log in as the victim. The victim's sessions and API applications are wiped, locking them out. The attacker reads the notes and can edit them. If those notes serve as a knowledge base for an internal RAG assistant or agent, the attacker could plant misleading content there.

Weaknesses (CWE)

CWE-620Unverified Password ChangePrimary CWE-640Weak Password Recovery Mechanism for Forgotten PasswordPrimary CWE-620Unverified Password Change CWE-640Weak Password Recovery Mechanism for Forgotten Password

CWE-620 — Unverified Password Change: When setting a new password for a user, the product does not require knowledge of the original password, or using another form of authentication.

[Architecture and Design] When prompting for a password change, force the user to provide the original password in addition to the new password.
[Architecture and Design] Do not use "forgotten password" functionality. But if you must, ensure that you are only providing information to the actual user, e.g. by using an email address or challenge question that the legitimate user already provided in the past; do not allow the current user to change this identity information until the correct password has been provided.

Source: MITRE CWE corpus.