When Zabbix Agent was installed on Windows into a custom installation directory, the installer did not verify whether the selected directory had secure access permissions. If the target directory allowed unauthorized users to modify its contents, an attacker could place a malicious DLL that could later be loaded by the application, resulting in DLL sideloading. The installer has been hardened to detect potentially unsafe installation directories and now requires explicit user confirmation before proceeding with installation in such locations. This reduces the risk of accidental installation into directories with inappropriate permissions while preserving compatibility with existing deployment scenarios.

An authenticated administrator is able to crash Zabbix server or proxy by creating specifically crafted preprocessing/script item JavaScript scripts, leading to potential denial of service.

The Zabbix API host.get action can be exploited by authenticated users to extract a host's PSK key leading to potential loss of data integrity.

In Zabbix 7.4 the cryptographic key used for signing Frontend sessions has been erroneously written to the database seed. Currently the only known exploitation scenario is for deployments that utilize both - SAML authentication and guest users. In such cases the key can be used to forge valid session cookies, potentially leading to unauthorized access. For other Zabbix deployments this does not have a known impact.

The Item history widget (in Zabbix 7.0+) or the Plain text widget (in Zabbix 6.0) can execute injected JavaScript when HTML display is enabled. This can allow an attacker to perform unauthorized actions depending on which user opens a dashboard containing these widgets. The malicious JavaScript would have to come from a monitored host controlled by the attacker. Note: the Item history widget is a replacement for the Plain text widget since Zabbix 7.0.

Zabbix Agent 2 Docker plugin does not properly sanitize the 'docker.containerinfo' parameters when forwarding them to the Docker daemon. An attacker capable of invoking Agent 2 can read arbitrary files from running Docker containers by injecting them via the Docker archive API.

A low privilege Zabbix user with API access can exploit a blind SQL injection vulnerability in include/classes/api/CApiService.php to execute arbitrary SQL selects via the sortfield parameter. Although query results are not returned directly, an attacker can exfiltrate arbitrary database data through time-based techniques, potentially leading to session identifier disclosure and administrator account compromise.

For performance reasons Zabbix Server/Proxy reuses JavaScript (Duktape) contexts (used in script items, JavaScript reprocessing, Webhooks). This can lead to confidentiality loss where a regular (non-super) Zabbix administrator leaks data for hosts they do not have access to. A fix has been released that makes the built in Zabbix JavaScript objects read-only, but please be advised that usage of global JavaScript variables is not recommended because their content could be leaked. More information <a href='https://www.zabbix.com/documentation/7.4/en/manual/installation/knownissues#preprocessing-global-variables-are-unsafe'>in Zabbix documentation</a>.

An authenticated Zabbix user (User role) with template/host write permissions is able to create objects via the configuration.import API. This can lead to confidentiality loss by creating unauthorized hosts. Note that the User role is normally not sufficient to create and edit templates/hosts even with write permissions.

An authenticated Zabbix user (including Guest) is able to cause disproportionate CPU load on the webserver by sending specially crafted parameters to /imgstore.php, leading to potential denial of service.

An authenticated Zabbix Super Admin can exploit the oauth.authorize action to read arbitrary files from the webserver leading to potential confidentiality loss.

A regular Zabbix user with no permission to the Monitoring -> Problems view is still able to call the problem.view.refresh action and therefore still retrieve a list of active problems.

In Zabbix Agent and Agent 2 on Windows, the OpenSSL configuration file is loaded from a path writable by low-privileged users, allowing malicious modification and potential local privilege escalation by injecting a DLL.

A regular Zabbix user can search other users in their user group via Zabbix API by select fields the user does not have access to view. This allows data-mining some field values the user does not have access to.

The LDAP 'Bind password' value cannot be read after saving, but a Super Admin account can leak it by changing LDAP 'Host' to a rogue LDAP server. To mitigate this, the 'Bind password' value is now reset on 'Host' change.

A Zabbix adminitrator can inject arbitrary SQL during the autoremoval of hosts by inserting malicious SQL in the 'Visible name' field.

Due to a bug in Zabbix API, the hostprototype.get method lists all host prototypes to users that do not have any user groups assigned to them.

Zabbix Agent 2 smartctl plugin does not properly sanitize smart.disk.get parameters, allowing an attacker to inject unexpected arguments into the smartctl command. This can be used to leak the NTLMv2 hash from a Windows system.

Zabbix Agent 2 smartctl plugin does not properly sanitize smart.disk.get parameters, allowing an attacker to inject unexpected arguments into the smartctl command. In Zabbix 5.0 this allows for remote code execution.

Zabbix server is vulnerable to a DoS vulnerability due to uncontrolled resource exhaustion. An attacker can send specially crafted requests to the server, which will cause the server to allocate an excessive amount of memory and perform CPU-intensive decompression operations, ultimately leading to a service crash.

The endpoint /zabbix.php?action=export.valuemaps suffers from a Cross-Site Scripting vulnerability via the backurl parameter. This is caused by the reflection of user-supplied data without appropriate HTML escaping or output encoding. As a result, a JavaScript payload may be injected into the above endpoint causing it to be executed within the context of the victim's browser.

Zabbix API user.get returns all users that share common group with the calling user. This includes media and other information, such as login attempts, etc.

Execution time for an unsuccessful login differs when using a non-existing username compared to using an existing one.

A low privilege (regular) Zabbix user with API access can use SQL injection vulnerability in include/classes/api/CApiService.php to execute arbitrary SQL commands via the groupBy parameter.

A bug in the code allows an attacker to sign a forged zbxsession cookie, which then allows them to sign in with admin permissions.

When exporting media types, the password is exported in the YAML in plain text. This appears to be a best practices type issue and may have no actual impact. The user would need to have permissions to access the media types and therefore would be expected to have access to these passwords.

The researcher is showing that it is possible to leak a small amount of Zabbix Server memory using an out of bounds read in src/libs/zbxmedia/email.c

The researcher is showing that due to the way the SNMP trap log is parsed, an attacker can craft an SNMP trap with additional lines of information and have forged data show in the Zabbix UI. This attack requires SNMP auth to be off and/or the attacker to know the community/auth details. The attack requires an SNMP item to be configured as text on the target host.

In the src/libs/zbxembed/browser.c file, the esbrowserctor method retrieves a heap pointer from the Duktape JavaScript engine. This heap pointer is subsequently utilized by the browserpusherror method in the src/libs/zbxembed/browsererror.c file. A use-after-free bug can occur at this stage if the wd->browser heap pointer is freed by garbage collection.

The HttpRequest object allows to get the HTTP headers from the server's response after sending the request. The problem is that the returned strings are created directly from the data returned by the server and are not correctly encoded for JavaScript. This allows to create internal strings that can be used to access hidden properties of objects.