NodeJS/matrix-react-sdk/3.59.0-rc.1

Prototype pollution in matrix-react-sdk

Impact

In certain configurations, data sent by remote servers containing special strings in key locations could cause modifications of the Object.prototype, disrupting matrix-react-sdk functionality, causing denial of service and potentially affecting program logic.

(This is part 2, where CVE-2022-36060 / GHSA-2x9c-qwgf-94xr is part 1. Part 2 covers remaining vectors not covered by part 1, found in a codebase audit scheduled after part 1.)

Patches

This is fixed in matrix-react-sdk 3.69.0

Workarounds

None.

References

• Release blog post
• The advisory GHSA-2x9c-qwgf-94xr (CVE-2022-36060) refers to an initial set of vulnerable locations discovered and patched in matrix-react-sdk 3.53.0. We opted not to disclose that advisory while we performed an audit of the codebase and are now disclosing it jointly with this one.

For more information

If you have any questions or comments about this advisory please email us at security at matrix.org.

matrix-react-sdk vulnerable to XSS in Export Chat feature

Description

The Export Chat feature includes certain attacker-controlled elements in the generated document without sufficient escaping, leading to stored XSS.

Impact

Since the Export Chat feature generates a separate document, an attacker can only inject code run from the null origin, restricting the impact.

However, the attacker can still potentially use the XSS to leak message contents. A malicious homeserver is a potential attacker since the affected inputs are controllable server-side.

Patches

This was patched in matrix-react-sdk 3.76.0.

Workarounds

None, other than not using the Export Chat feature.

References

N/A

Matrix SDK for React's URL preview setting for a room is controllable by the homeserver

Impact

A malicious homeserver could manipulate a user's account data to cause the client to enable URL previews in end-to-end encrypted rooms, in which case any URLs in encrypted messages would be sent to the server.

Even if the CVSS score would be 4.1 (AV:N/AC:L/PR:H/UI:N/S:C/C:L/I:N/A:N) the maintainer classifies this as High severity issue.

Patches

This was patched in matrix-react-sdk 3.105.1.

Workarounds

Deployments that trust their homeservers, as well as closed federations of trusted servers, are not affected.

References

N/A.

Malicious homeservers can steal message keys when the matrix-react-sdk user invites another user to a room

Impact

matrix-react-sdk before 3.102.0 allows a malicious homeserver to potentially steal message keys for a room when a user invites another user to that room, via injection of a malicious device controlled by the homeserver. This is possible because matrix-react-sdk before 3.102.0 shared historical message keys on invite.

Patches

matrix-react-sdk 3.102.0 disables sharing message keys on invite by removing calls to the vulnerable functionality.

Workarounds

None.

References

The vulnerability in matrix-react-sdk is caused by calling MatrixClient.sendSharedHistoryKeys in matrix-js-sdk, which is inherently vulnerable to this sort of attack. This matrix-js-sdk vulnerability is tracked as CVE-2024-47080 / GHSA-4jf8-g8wp-cx7c. Given that this functionality is not specific to sharing message keys on invite, is optional, has to be explicitly called by the caller and has been independently patched in matrix-react-sdk by removing the offending calls, we believe it is proper to treat the matrix-react-sdk vulnerability as a separate one, with its own advisory and CVE.

The matrix-org/matrix-react-sdk repository has recently been archived and the project was moved to element-hq/matrix-react-sdk. Given that this happened after the first patched release, no releases of the project on element-hq/matrix-react-sdk were ever vulnerable to this vulnerability.

Patching pull request: https://github.com/matrix-org/matrix-react-sdk/pull/12618.

For more information

If you have any questions or comments about this advisory, please email us at security at security at matrix.org.

HTML injection in search results via plaintext message highlighting

Impact

Plain text messages containing HTML tags are rendered as HTML in the search results. To exploit this, an attacker needs to trick a user into searching for a specific message containing an HTML injection payload.

Cross-site scripting is possible by including resources from recaptcha.net and gstatic.com which are included in the default CSP.

Thanks to Cadence Ember for finding the injection and to S1m for finding possible XSS vectors.

Patches

Version 3.71.0 of the SDK fixes the issue.

Workarounds

Restarting the client will clear the injection.