A flaw was found in the pipe lookup plugin of ansible. Arbitrary commands can be run, when the pipe lookup plugin uses subprocess.Popen() with shell=True, by overwriting ansible facts and the variable is not escaped by quote plugin. An attacker could take advantage and run arbitrary commands by overwriting the ansible facts.

Ansible before 1.5.5 sets 0644 permissions for sources.list, which might allow local users to obtain sensitive credential information in opportunistic circumstances by reading a file that uses the "deb http://user:pass@server:port/" format.

The vault subsystem in Ansible before 1.5.5 does not set the umask before creation or modification of a vault file, which allows local users to obtain sensitive key information by reading a file.

The safe_eval function in Ansible before 1.5.4 does not properly restrict the code subset, which allows remote attackers to execute arbitrary code via crafted instructions.

The safe_eval function in Ansible before 1.6.4 does not properly restrict the code subset, which allows remote attackers to execute arbitrary code via crafted instructions. NOTE: this vulnerability exists because of an incomplete fix for CVE-2014-4657.

Ansible before 1.5.5 constructs filenames containing user and password fields on the basis of deb lines in sources.list, which might allow local users to obtain sensitive credential information in opportunistic circumstances by leveraging existence of a file that uses the "deb http://user:pass@server:port/" format.

Multiple argument injection vulnerabilities in Ansible before 1.6.7 allow remote attackers to execute arbitrary code by leveraging access to an Ansible managed host and providing a crafted fact, as demonstrated by a fact with (1) a trailing " src=" clause, (2) a trailing " temp=" clause, or (3) a trailing " validate=" clause accompanied by a shell command.

Ansible before 1.6.7 does not prevent inventory data with "{{" and "lookup" substrings, and does not prevent remote data with "{{" substrings, which allows remote attackers to execute arbitrary code via (1) crafted lookup('pipe') calls or (2) crafted Jinja2 data.

Ansible prior to 1.5.4 mishandles the evaluation of some strings.

Ansible, versions 2.9.x before 2.9.1, 2.8.x before 2.8.7 and Ansible versions 2.7.x before 2.7.15, is not respecting the flag no_log set it to True when Sumologic and Splunk callback plugins are used send tasks results events to collectors. This would discloses and collects any sensitive data.

A flaw was found in Ansible Tower, versions 3.6.x before 3.6.2 and 3.5.x before 3.5.4, when /websocket is requested and the password contains the '#' character. This request would cause a socket error in RabbitMQ when parsing the password and an HTTP error code 500 and partial password disclose will occur in plaintext. An attacker could easily guess some predictable passwords or brute force the password.

A flaw was found in Ansible Tower, versions 3.6.x before 3.6.2, where files in '/var/backup/tower' are left world-readable. These files include both the SECRET_KEY and the database backup. Any user with access to the Tower server, and knowledge of when a backup is run, could retrieve every credential stored in Tower. Access to data is the highest threat with this vulnerability.

A flaw was found in Ansible Tower, versions 3.6.x before 3.6.2 and 3.5.x before 3.5.3, where enabling RabbitMQ manager by setting it with '-e rabbitmq_enable_manager=true' exposes the RabbitMQ management interface publicly, as expected. If the default admin user is still active, an attacker could guess the password and gain access to the system.

ansible before versions 2.8.6, 2.7.14, 2.6.20 is vulnerable to a None

A vulnerability was found in Ansible Tower before 3.6.1 where an attacker with low privilege could retrieve usernames and passwords credentials from the new RHSM saved in plain text into the database at '/api/v2/config' when applying the Ansible Tower license.

A flaw was found in ansible 2.8.0 before 2.8.4. Fields managing sensitive data should be set as such by no_log feature. Some of these fields in GCP modules are not set properly. service_account_contents() which is common class for all gcp modules is not setting no_log to True. Any sensitive data managed by that function would be leak as an output when running ansible playbooks.

ansible-playbook -k and ansible cli tools, all versions 2.8.x before 2.8.4, all 2.7.x before 2.7.13 and all 2.6.x before 2.6.19, prompt passwords by expanding them from templates as they could contain special characters. Passwords should be wrapped to prevent templates trigger and exposing them.

A vulnerability was found in Ansible engine 2.x up to 2.8 and Ansible tower 3.x up to 3.5. When a module has an argument_spec with sub parameters marked as no_log, passing an invalid parameter name to the module will cause the task to fail before the no_log options in the sub parameters are processed. As a result, data in the sub parameter fields will not be masked and will be displayed if Ansible is run with increased verbosity and present in the module invocation arguments for the task.

In Ansible, all Ansible Engine versions up to ansible-engine 2.8.5, ansible-engine 2.7.13, ansible-engine 2.6.19, were logging at the DEBUG level which lead to a disclosure of credentials if a plugin used a library that logged credentials at the DEBUG level. This flaw does not affect Ansible modules, as those are executed in a separate process.

core.py in Mitogen before 0.2.8 has a typo that drops the unidirectional-routing protection mechanism in the case of a child that is initiated by another child. The Ansible extension is unaffected. NOTE: the vendor disputes this issue because it is exploitable only in conjunction with hypothetical other factors, i.e., an affected use case within a library caller, and a bug in the message receiver policy code that led to reliance on this extra protection mechanism