Versions of Puppet Agent prior to 1.6.0 included a version of the Puppet Execution Protocol (PXP) agent that passed environment variables through to Puppet runs. This could allow unauthorized code to be loaded. This bug was first introduced in Puppet Agent 1.3.0.

The pxp-agent component in Puppet Enterprise 2015.3.x before 2015.3.3 and Puppet Agent 1.3.x before 1.3.6 does not properly validate server certificates, which might allow remote attackers to spoof brokers and execute arbitrary commands via a crafted certificate.

A flaw was discovered in Puppet Agent and Puppet Server that may result in a leak of HTTP credentials when following HTTP redirects to a different host. This is similar to CVE-2018-1000007

Puppet Server before 2.3.2 and Ruby puppetmaster in Puppet 4.x before 4.4.2 and in Puppet Agent before 1.4.2 might allow remote attackers to bypass intended auth.conf access restrictions by leveraging incorrect URL decoding.

The (1) template and (2) inlinetemplate functions in the master server in Puppet before 2.6.18, 2.7.x before 2.7.21, and 3.1.x before 3.1.1, and Puppet Enterprise before 1.2.7 and 2.7.x before 2.7.2 allows remote authenticated users to execute arbitrary code via a crafted catalog request.

A flaw was discovered in Puppet DB, this flaw results in an escalation of privileges which allows the user to delete tables via an SQL query.

For certificates that utilize the auto-renew feature in Puppet Server, a flaw exists which prevents the certificates from being revoked.

Puppet 2.7.x before 2.7.21 and 3.1.x before 3.1.1, when running Ruby 1.9.3 or later, allows remote attackers to execute arbitrary code via vectors related to "serialized attributes."

Puppet 2.7.x before 2.7.22 and 3.2.x before 3.2.2, and Puppet Enterprise before 2.8.2, deserializes untrusted YAML, which allows remote attackers to instantiate arbitrary Ruby classes and execute arbitrary code via a crafted REST API call.

Prior to version 0.3.0, chloride's use of net-ssh resulted in host fingerprints for previously unknown hosts getting added to the user's knownhosts file without confirmation. In version 0.3.0 this is updated so that the user's knownhosts file is not updated by chloride.

Puppet Enterprise 2015.3.3 and 2016.x before 2016.4.0, and Puppet Agent 1.3.6 through 1.7.0 allow remote attackers to bypass a host whitelist protection mechanism and execute arbitrary code on Puppet nodes via vectors related to command validation, aka "Puppet Execution Protocol (PXP) Command Whitelist Validation Vulnerability."

Puppet before 2.6.18, 2.7.x before 2.7.21, and 3.1.x before 3.1.1, and Puppet Enterprise before 1.2.7 and 2.7.x before 2.7.2, when listening for incoming connections is enabled and allowing access to the "run" REST endpoint is allowed, allows remote authenticated users to execute arbitrary code via a crafted HTTP request.

Utilization of a module presented a security risk by allowing the deserialization of untrusted/user supplied data. This is resolved in the Puppet Agent 7.4.0 release.

Puppet 2.6.x before 2.6.18 and Puppet Enterprise 1.2.x before 1.2.7 allows remote authenticated users to execute arbitrary code on the puppet master, or an agent with puppet kick enabled, via a crafted request for a report.

The default vhost configuration file in Puppet before 3.6.2 does not include the SSLCARevocationCheck directive, which might allow remote attackers to obtain sensitive information via a revoked certificate when a Puppet master runs with Apache 2.4.

A flaw was discovered in Puppet Agent where the agent may silently ignore Augeas settings or may be vulnerable to a Denial of Service condition prior to the first 'pluginsync'.

Puppet 2.7.x before 2.7.5, 2.6.x before 2.6.11, and 0.25.x allows local users to overwrite arbitrary files via a symlink attack on the .k5login file.

Puppet 2.7.x before 2.7.5, 2.6.x before 2.6.11, and 0.25.x allows local users to modify the permissions of arbitrary files via a symlink attack on the SSH authorizedkeys file.

Puppet 2.7.x before 2.7.5, 2.6.x before 2.6.11, and 0.25.x, when running in --edit mode, uses a predictable file name, which allows local users to run arbitrary Puppet code or trick a user into editing arbitrary files.

Last updated 24 July 2024

Puppet 2.6.x before 2.6.15 and 2.7.x before 2.7.13, and Puppet Enterprise (PE) Users 1.0, 1.1, 1.2.x, 2.0.x, and 2.5.x before 2.5.1 allows remote authenticated users with agent SSL keys and file-creation permissions on the puppet master to execute arbitrary commands by creating a file whose full pathname contains shell metacharacters, then performing a filebucket request.

Puppet 2.6.0 through 2.6.3 does not properly restrict access to node resources, which allows remote authenticated Puppet nodes to read or modify the resources of other nodes via unspecified vectors.

Unspecified vulnerability in Puppet 2.7.x before 2.7.23 and 3.2.x before 3.2.4, and Puppet Enterprise 2.8.x before 2.8.3 and 3.0.x before 3.0.1, allows remote attackers to execute arbitrary Ruby programs from the master via the resourcetype service. NOTE: this vulnerability can only be exploited utilizing unspecified "local file system access" to the Puppet Master.

Puppet 2.7.x before 2.7.21 and 3.1.x before 3.1.1, and Puppet Enterprise 2.7.x before 2.7.2, does not properly negotiate the SSL protocol between client and master, which allows remote attackers to conduct SSLv2 downgrade attacks against SSLv3 sessions via unspecified vectors.

Puppet Labs Puppet Enterprise before 2.8.0 does not use a "randomized secret" in the CAS client config file (casclientconfig.yml) when upgrading from older 1.2.x or 2.0.x versions, which allows remote attackers to obtain console access via a crafted cookie.

Directory traversal vulnerability in Puppet 2.6.x before 2.6.10 and 2.7.x before 2.7.4 allows remote attackers to write X.509 Certificate Signing Request (CSR) to arbitrary locations via (1) a double-encoded key parameter in the URI in 2.7.x, (2) the CN in the Subject of a CSR in 2.6 and 0.25.

Puppet before 2.6.18, 2.7.x before 2.7.21, and 3.1.x before 3.1.1, and Puppet Enterprise before 1.2.7 and 2.7.x before 2.7.2 allows remote authenticated users with a valid certificate and private key to read arbitrary catalogs or poison the master's cache via unspecified vectors.

A flaw was discovered in bolt-server and ace where running a task with sensitive parameters results in those sensitive parameters being logged when they should not be. This issue only affects SSH/WinRM nodes (inventory service nodes).

Description of problem: I noticed that puppetmasterd does not initialize its supplementary groups, which may lead to allow puppetmasterd to access files, it should not. E.g. if it is started with "service puppetmaster start", it still has access to all files that allow read access for the supplementary groups of root, e.g. raw disk devices.

I filed an upstream bug report including patches (one needs still to be tested) here: http://projects.reductivelabs.com/issues/show/1806

Version-Release number of selected component (if applicable): puppet-0.24.6-1.fc10 puupet-0.24.6-1.el5

How reproducible: always

Steps to Reproduce: 1. # service puppetmaster start 2. # cat /proc/$(ps --User puppet -o pid | tail -n 1)/status | grep Group Actual results: The output matches "id -G root".

Expected results: The output should match "id -G puppet", i.e. the process should run with the supplementary groups of puppet.

The default supplementary groups of root include the group disk, which e.g. provides raw read access on disk devices.

Additional info: I am not sure, whether this really classifies as a security vulnerability, because https://fedoraproject.org/wiki/Security/Classifications only mentions code execution and denial of service, but imho unwanted access to restricted information is a security vulnerability, too.

A flaw was divered in Puppet Enterprise and other Puppet products where sensitive plan parameters may be logged