Syntribos is an open source automated API security testing tool that is maintained by members of the OpenStack Security Project.
Given a simple configuration file and an example HTTP request, syntribos can replace any API URL, URL parameter, HTTP header and request body field with a given set of strings. Syntribos iterates through each position in the request automatically. Syntribos aims to automatically detect common security defects such as SQL injection, LDAP injection, buffer overflow, etc. In addition, syntribos can be used to help identify new security defects by automated fuzzing.
Syntribos has the capability to test any API, but is designed with OpenStack applications in mind.
With syntribos, you can initiate automated testing of any API with minimal configuration effort. Syntribos is ideal for testing the OpenStack API as it will help you in automatically downloading a set of templates of some of the bigger OpenStack projects like nova, neutron, keystone, etc.
A short list of tests that can be run using syntribos is given below:
- Buffer Overflow
- Command Injection
- CORS Wildcard
- Integer Overflow
- LDAP Injection
- SQL Injection
- String Validation
- XML External Entity
- Cross Site Scripting (XSS)
- Regex Denial of Service (ReDoS)
- JSON Parser Depth Limit
- User Defined
Buffer overflow attacks, in the context of a web application, force an application to handle more data than it can hold in a buffer. In syntribos, a buffer overflow test is attempted by injecting a large string into the body of an HTTP request.
Command injection attacks are done by injecting arbitrary commands in an attempt to execute these commands on a remote system. In syntribos, this is achieved by injecting a set of strings that have been proven as successful executors of injection attacks.
CORS wildcard tests are used to verify if a web server allows cross-domain resource sharing from any external URL (wild carding of Access-Control-Allow-Origin header), rather than a white list of URLs.
Integer overflow tests in syntribos attempt to inject numeric values that the remote application may fail to represent within its storage. For example, injecting a 64 bit number into a 32 bit integer type.
Syntribos attempts LDAP injection attacks by injecting LDAP statements into HTTP requests; if an application fails to properly sanitize the request content, it may be possible to execute arbitrary commands.
SQL injection attacks are one of the most common web application attacks. If the user input is not properly sanitized, it is fairly easy to execute SQL queries that may result in an attacker reading sensitive information or gaining control of the SQL server. In syntribos, an application is tested for SQL injection vulnerabilities by injecting SQL strings into the HTTP request.
Some string patterns are not sanitized effectively by the input validator and may cause the application to crash. String validation attacks in syntribos try to exploit this by inputting characters that may cause string validation vulnerabilities. For example, special unicode characters, emojis, etc.
XML external entity attacks target the web application's XML parser. If an XML parser allows processing of external entities referenced in an XML document then an attacker might be able to cause a denial of service, or leakage of information, etc. Syntribos tries to inject a few malicious strings into an XML body while sending requests to an application in an attempt to obtain an appropriate response.
XSS attacks inject malicious JavaScript into a web application. Syntribos tries to find potential XSS issues by injecting string containing "script" and other HTML tags into request fields.
ReDoS attacks attempt to produce a denial of service by providing a regular expression that takes a very long time to evaluate. This can cause the regex engine to backtrack indefinitely, which can slow down some parsers or even cause a processing halt. The attack exploits the fact that most regular expression implementations have an exponential time worst case complexity.
There is a possibility that the JSON parser will reach depth limit and crash, resulting in a successful overflow of the JSON parsers depth limit, leading to a DoS vulnerability. Syntribos tries to check for this, and raises an issue if the parser crashes.
This test gives users the ability to fuzz using user defined fuzz data and provides an option to look for failure strings provided by the user. The fuzz data needs to be provided using the config option [user_defined].
Example:
[user_defined] payload=<payload_file> failure_strings=<[list_of_failure_strings] # optional
Other than these built-in tests, you can extend syntribos by writing
your own custom tests. To do this, download the source code and look at
the tests in the syntribos/tests
directory. The CORS test may be an easy
one to emulate. In the same way, you can also add different extensions
to the tests. To see how extensions can be written please see the
syntribos/extensions
directory.
Details
- Documentation
- Free software: Apache license
- Launchpad project
- Blueprints
- Bugs
- Source code
Syntribos has been developed primarily in Linux and Mac environments and would work on most Unix and Linux based Operating Systems. At this point, we are not supporting Windows, but this may change in the future.
Syntribos can be installed directly from pypi with pip.
pip install syntribos
For the latest changes, install syntribos from source with pip.
Clone the repository:
$ git clone https://github.com/openstack/syntribos.git
Change directory into the repository clone and install with pip:
$ cd syntribos $ pip install .
Once syntribos is installed, you must initialize the syntribos environment.
This can be done manually, or with the init
command.
$ syntribos init
Note
By default, syntribos init
fetches a set of default payload files
from a remote repository
maintained by our development team. These payload files are necessary for
our fuzz tests to run. To disable this behavior, run syntribos with the
--no_downloads
flag. Payload files can also be fetched by running
syntribos download --payloads
at any time.
To specify a custom root for syntribos to be installed in,
specify the --custom_root
flag. This will skip
prompts for information from the terminal, which can be handy for
Jenkins jobs and other situations where user input cannot be retrieved.
If you've already run the init
command but want to start over with a fresh
environment, you can specify the --force
flag to overwrite existing files.
The --custom_root
and --force
flags can be combined to
overwrite files in a custom install root.
Note: if you install syntribos to a custom install root, you must supply the
--custom_root
flag when running syntribos.
Example:
$ syntribos --custom_root /your/custom/path init --force $ syntribos --custom_root /your/custom/path run
All configuration files should have a [syntribos]
section.
Add other sections depending on what extensions you are using
and what you are testing. For example, if you are using the
built-in identity extension, you would need the [user]
section. The sections [logging]
and [remote]
are optional.
The basic structure of a syntribos configuration file is given below:
[syntribos] # # End point URLs and versions of the services to be tested. # endpoint=http://localhost:5000 # Set payload and templates path templates=<location_of_templates_dir/file> payloads=<location_of_payloads_dir> [user] # # User credentials and endpoint URL to get an AUTH_TOKEN # This section is only needed if you are using the identity extension. # endpoint= username=<yourusername> password=<yourpassword> [remote] # # Optional, to define remote URI and cache_dir explicitly # templates_uri=<URI to a tar file of set of templates> payloads_uri=<URI to a tar file of set of payloads> cache_dir=<a local path to save the downloaded files> [logging] log_dir=<location_to_save_debug_logs>
The endpoint URL specified in the [syntribos]
section is the endpoint URL
tested by syntribos. The endpoint URL in the [user]
section is used to
get an AUTH_TOKEN. To test any project, update the endpoint URL under
[syntribos]
to point to the API and also modify the user
credentials if needed.
Payload and template files can be downloaded remotely in syntribos.
In the config file under the [syntribos]
section, if the templates
and payloads
options are not set, by default syntribos will
download all the latest payloads and the templates for a few OpenStack
projects.
To specify a URI to download custom templates and payloads
from, use the [remotes]
section in the config file.
Available options under [remotes]
are cache_dir
, templates_uri
,
payloads_uri
, and enable_cache
. The enable_cache
option is
True
by default; set to False
to disable caching of remote
content while syntribos is running. If the cache_dir
set to a path,
syntribos will attempt to use that as a base directory to save downloaded
template and payload files.
The advantage of using these options are that you will be able to get the latest payloads from the official repository and if you are using syntribos to test OpenStack projects, then, in most cases you could directly use the well defined templates available with this option.
This option also helps to easily manage different versions of templates remotely, without the need to maintain a set of different versions offline.
A sample config file is given in examples/configs/keystone.conf
.
Copy this file to a location of your choice (the default file path for the
configuration file is: ~/.syntribos/syntribos.conf
) and update the
necessary fields, such as user credentials, log, template directory, etc.
$ vi examples/configs/keystone.conf [syntribos] # # As keystone is being tested in the example, enter your # # keystone auth endpoint url. endpoint=http://localhost:5000 # Set payload and templates path templates=<location_of_templates_dir/file> payloads=<location_of_payloads_dir> [user] # # User credentials # endpoint=http://localhost:5000 username=<yourusername> password=<yourpassword> # Optional, only needed if Keystone V3 API is used #user_id=<youruserid> # Optional, api version if required #version=v2.0 # Optional, for getting scoped tokens #user_id=<alt_userid> # If user id is not known # For V3 API #domain_name=<name_of_the_domain> #project_name=<name_of_the_project> # For Keystone V2 API #tenant_name=<name_of_the_project> #[alt_user] # # Optional, Used for cross auth tests (-t AUTH) # #endpoint=http://localhost:5000 #username=<alt_username> #password=<alt_password> # Optional, for getting scoped tokens #user_id=<alt_userid> # If user id is not known # For V3 API #domain_name=<name_of_the_domain> #project_name=<name_of_the_project> # For Keystone V2 API #tenant_name=<name_of_the_project> [remote] # # Optional, Used to specify URLs of templates and payloads # #cache_dir=<a local path to save the downloaded files> #templates_uri=https://github.com/your_project/templates.tar #payloads_uri=https://github.com/your_project/payloads.tar # To disable caching of these remote contents, set the following variable to False #enable_caching=True [logging] # # Logger options go here # log_dir=<location_to_store_log_files> # Optional, compresses http_request_content, # if you don't want this, set this option to False. http_request_compression=True
Below are the set of commands that can be specified while using syntribos:
init
This command sets up the syntribos environment after installation. Running this command creates the necessary folders for templates, payloads, and logs; as well a sample configuration file.
$ syntribos init
To learn more about
syntribos init
, see the installation instructions here.run
This command runs syntribos with the given config options.
$ syntribos --config-file keystone.conf -t SQL run
dry_run
This command ensures that the template files given for this run parse successfully and without errors. It then runs a debug test which sends no requests of its own.
$ syntribos --config-file keystone.conf dry_run
Note
If any external calls referenced inside the template file do make requests, the parser will still make those requests even for a dry run.
list_tests
This command will list the names of all the tests that can be executed by the
run
command with their description.$ syntribos --config-file keystone.conf list_tests
download
This command will download templates and payload files. By default, it will download a set of OpenStack template files (with the
--templates
flag), or a set of payloads (with the--payloads
flag) to your syntribos root directory. However, the behavior of this command can be configured in the[remote]
section of your config file.$ syntribos download --templates
Important
All these commands, except init
, will only work if a configuration file
is specified. If a configuration file is present in the default
path ( ~/.syntribos/syntribos.conf
), then you
do not need to explicitly specify a config file and
can run syntribos using the command syntribos run
.
By default, syntribos looks in the syntribos home directory (the directory
specified when running the syntribos init
command on install) for config
files, payloads, and templates. This can all be overridden through command
line options. For a full list of command line options available, run
syntribos --help
from the command line.
To run syntribos against all the available tests, specify the
command syntribos
, with the configuration file (if needed), without
specifying any test type.
$ syntribos --config-file keystone.conf run
It is possible to limit syntribos to run a specific test type using
the -t
flag.
$ syntribos --config-file keystone.conf -t SQL run
This will match all tests that contain SQL
in their name. For example:
SQL_INJECTION_HEADERS
, SQL_INJECTION_BODY
, etc.
If you set up the syntribos environment with a custom root (i.e. with
syntribos --custom_root init
), you can point to it with the
--custom_root
configuration option. Syntribos will look for a
syntribos.conf
file inside this directory, and will read further
configuration information from there.
There are two types of logs generated by syntribos:
- The results log is a collection of issues generated at the end of a syntribos run to represent results.
- The debug log contains debugging information captured during a particular run. Debug logs may include exception messages, warnings, raw but sanitized request/response data, and a few more details. A modified version of Python logger is used for collecting debug logs in syntribos.
The results log is displayed at the end of every syntribos run, it can be
written to a file by using the -o
flag on the command line.
The results log includes failures and errors. The "failures"
key represents
tests that have failed, indicating a possible security vulnerability. The
"errors"
key gives us information on any unhandled exceptions, such as
connection errors, encountered on that run.
Example failure object:
{ "defect_type": "xss_strings", "description": "The string(s): '[\"<STYLE>@import'http://xss.rocks/xss.css';</STYLE>\"]', known to be commonly returned after a successful XSS attack, have been found in the response. This could indicate a vulnerability to XSS attacks.", "failure_id": 33, "instances": [ { "confidence": "LOW", "param": { "location": "data", "method": "POST", "type": null, "variables": [ "type", "details/name", ] }, "severity": "LOW", "signals": { "diff_signals": [ "LENGTH_DIFF_OVER" ], "init_signals": [ "HTTP_CONTENT_TYPE_JSON", "HTTP_STATUS_CODE_2XX_201" ], "test_signals": [ "FAILURE_KEYS_PRESENT", "HTTP_CONTENT_TYPE_JSON", "HTTP_STATUS_CODE_2XX_201", ] }, "strings": [ "<STYLE>@import'http://xss.rocks/xss.css';</STYLE>" ] } ], "url": "127.0.0.1/test" }
Error form:
ERROR: { "error": "Traceback (most recent call last):\n File \"/Users/test/syntribos/tests/fuzz/base_fuzz.py\", line 58, in tearDownClass\n super(BaseFuzzTestCase, cls).tearDownClass()\n File \"/Users/test/syntribos/tests/base.py\", line 166, in tearDownClass\n raise sig.data[\"exception\"]\nReadTimeout: HTTPConnectionPool(host='127.0.0.1', port=8080): Read timed out. (read timeout=10)\n", "test": "tearDownClass (syntribos.tests.fuzz.sql.image_data_image_data_get.template_SQL_INJECTION_HEADERS_sql-injection.txt_str21_model1)" }
Debug logs include details about HTTP requests, HTTP responses, and other
debugging information such as errors and warnings across the project. The
path where debug logs are saved by default is .syntribos/logs/
.
Debug logs are arranged in directories based on the timestamp in these
directories and files are named according to the templates.
For example:
$ ls .syntribos/logs/ 2016-09-15_11:06:37.198412 2016-09-16_10:11:37.834892 2016-09-16_13:31:36.362584 2016-09-15_11:34:33.271606 2016-09-16_10:38:55.820827 2016-09-16_13:36:43.151048 2016-09-15_11:41:53.859970 2016-09-16_10:39:50.501820 2016-09-16_13:40:23.203920
$ ls .syntribos/logs/2016-09-16_13:31:36.362584 API_Versions::list_versions_template.log API_Versions::show_api_details_template.log availability_zones::get_availability_zone_detail_template.log availability_zones::get_availability_zone_template.log cells::delete_os_cells_template.log cells::get_os_cells_capacities_template.log cells::get_os_cells_data_template.log
Each log file includes some essential debugging information such as the string representation of the request object, signals, and checks used for tests, etc.
Example request:
------------ REQUEST SENT ------------ request method.......: PUT request url..........: http://127.0.0.1/api request params.......: request headers size.: 7 request headers......: {'Content-Length': '0', 'Accept-Encoding': 'gzip, deflate', 'Accept': 'application/json', 'X-Auth-Token': <uuid>, 'Connection': 'keep-alive', 'User-Agent': 'python-requests/2.11.1', 'content-type': 'application/xml'} request body size....: 0 request body.........: None
Example response:
----------------- RESPONSE RECEIVED ----------------- response status..: <Response [415]> response headers.: {'Content-Length': '70', 'X-Compute-Request-Id': <random id>, 'Vary': 'OpenStack-API-Version, X-OpenStack-Nova-API-Version', 'Openstack-Api-Version': 'compute 2.1', 'Connection': 'close', 'X-Openstack-Nova-Api-Version': '2.1', 'Date': 'Fri, 16 Sep 2016 14:15:27 GMT', 'Content-Type': 'application/json; charset=UTF-8'} response time....: 0.036277 response size....: 70 response body....: {"badMediaType": {"message": "Unsupported Content-Type", "code": 415}} ------------------------------------------------------------------------------- [2590] : XSS_BODY (<syntribos.clients.http.client.SynHTTPClient object at 0x102c65f10>, 'PUT', 'http://127.0.0.1/api') {'headers': {'Accept': 'application/json', 'X-Auth-Token': <uuid> }, 'params': {}, 'sanitize': False, 'data': '', 'requestslib_kwargs': {'timeout': 10}} Starting new HTTP connection (1): 127.0.0.1 "PUT http://127.0.0.1/api HTTP/1.1" 501 93
Example signals captured:
Signals: ['HTTP_STATUS_CODE_4XX_400', 'HTTP_CONTENT_TYPE_JSON'] Checks used: ['HTTP_STATUS_CODE', 'HTTP_CONTENT_TYPE']
Debug logs are sanitized to prevent storing secrets to log files. Passwords and other sensitive information are marked with asterisks using a slightly modified version of oslo_utils.strutils.mask_password.
Debug logs also include string compression, wherein long fuzz strings are
compressed before being written to the logs. The threshold to start data
compression is set to 512 characters. Although it is not recommended to turn
off compression, it is possible by setting the variable
"http_request_compression"
, under the logging section in the config file,
to False
.
This section describes how to write templates and how to run specific tests. Templates are input files which have raw HTTP requests and may be supplemented with variable data using extensions.
In general, a request template is a marked-up raw HTTP request. It's possible for you to test your application by using raw HTTP requests as your request templates, but syntribos allows you to mark-up your request templates for further functionality.
A request template looks something like this:
POST /users/{user1} HTTP/1.1 Content-Type: application/json X-Auth-Token: CALL_EXTERNAL|syntribos.extensions.vAPI.client:get_token:[]| {"newpassword": "qwerty123"}
For fuzz tests, syntribos will automatically detect URL parameters, headers,
and body content as fields to fuzz. It will not automatically detect URL path
elements as fuzz fields, but they can be specified with curly braces {}
.
Note: The name of a template file must end with the extension .template
Otherwise, syntribos will skip the file and will not attempt to parse any files
that do not adhere to this naming scheme.
Extensions can be used to supplement syntribos template files with variable data, or data retrieved from external sources.
Extensions are found in syntribos/extensions/
.
Calls to extensions are made in the form below:
CALL_EXTERNAL|{extension dot path}:{function name}:[arguments]
One example packaged with syntribos enables the tester to obtain an AUTH
token from keystone. The code is located in identity/client.py
.
To use this extension, you can add the following to your template file:
X-Auth-Token: CALL_EXTERNAL|syntribos.extensions.identity.client:get_token_v3:["user"]|
The "user"
string indicates the data from the configuration file we
added in examples/configs/keystone.conf
.
Another example is found in random_data/client.py
. This returns a
UUID when random, but unique data is needed. The UUID can be used in place of
usernames when fuzzing a create user call.
"username": "CALL_EXTERNAL|syntribos.extensions.random_data.client:get_uuid:[]|"
The extension function can return one value, or be used as a generator if you want it to change for each test.
Syntribos comes with a slew of utility functions/extensions, these functions can be used to dynamically inject data into templates.
Method | Parameters | Description |
---|---|---|
hash_it | [data, hash_type (optional hash type, default being SHA256)] | Returns hashed value of data |
hmac_it | [data, key, hash_type (optional hash type, default being SHA256)] | Returns HMAC based on the has algorithm, data and the key provided |
epoch_time | [offset (optional integer offset value, default is zero)] | Returns the current time minus offset since epoch |
utc_datetime | [] | Returns current UTC date time |
base64_encode | [data] | Returns base 64 encoded value of data supplied |
url_encode | [url] | Returns encoded URL |
All these utility functions can be called using the following syntax:
CALL_EXTERNAL|common_utils.client.{method_name}:{comma separated parameters in square brackets}
For example:
"encoded_url": "CALL_EXTERNAL|common_utils.client:url_encode:['http://localhost:5000']|
Other functions that return random values can be seen below:
Method | Parameters | Description |
---|---|---|
get_uuid | [] | Returns a random UUID |
random_port | [] | Returns random port number between 0 and 65535 |
random_ip | [] | Returns random ipv4 address |
random_mac | [] | Returns random mac address |
random_integer | [beg (optional beginning value, default is 0), end (optional end value)] | Returns an integer value between 0 and 1468029570 by default |
random_utc_datetime | [] | Returns random UTC datetime |
These can be called using:
CALL_EXTERNAL|random_data.client.{method_name}:{comma separated parameters in square brackets}
For example:
"address": "CALL_EXTERNAL|random_data.client:random_ip:[]|"
While syntribos is designed to test all fields in a request, it can also ignore specific fields through the use of Action Fields. If you want to fuzz against a static object ID, use the Action Field indicator as follows:
"ACTION_FIELD:id": "1a16f348-c8d5-42ec-a474-b1cdf78cf40f"
The ID provided will remain static for every test.
Syntribos allows for templates to read in variables from a user-specified meta variable file. These files contain JSON objects that define variables to be used in one or more request templates.
The file must be named meta.json, and they take the form:
{ "user_password": { "val": 1234 }, "user_name": { "type": config, "val": "user.username" "fuzz_types": ["ascii"] }, "user_token": { "type": "function", "val": "syntribos.extensions.identity:get_scoped_token_v3", "args": ["user"], "fuzz": false } }
To reference a meta variable from a request template, reference the variable name surrounded by | (pipe). An example request template with meta variables is as follows:
POST /user HTTP/1.1 X-Auth-Token: |user_token| { "user": { "username": "|user_name|", "password": "|user_password|" } }
Note: Meta-variable usage in templates should take the form |user_name|, not user_|name| or |user|_|name|. This is to avoid ambiguous behavior when the value is fuzzed.
- val - All meta variable objects must define a value, which can be of any json DataType. Unlike the other attributes, this attribute is not optional.
- type - Defining a type instructs syntribos to interpret the variable in a
certain way. Any variables without a type defined will be read in directly
from the value. The following types are allowed:
- config - syntribos reads the config value specified by the "val" attribute and returns that value.
- function - syntribos calls the function named in the "val" attribute with any arguments given in the optional "args" attribute, and returns the value from calling the function. This value is cached, and will be returned on subsequent calls.
- generator - Works the same way as the function type, but its results are not cached and the function will be called every time.
- args - A list of function arguments (if any) which can be defined here if the variable is a generator or a function
- fuzz - A boolean value that, if set to false, instructs syntribos to ignore this variable for any fuzz tests
- fuzz_types - A list of strings which instructs syntribos to only use certain
fuzz strings when fuzzing this variable. More than one fuzz type can be
defined. The following fuzz types are allowed:
- ascii - strings that can be encoded as ascii
- url - strings that contain only url safe characters
- min_length/max_length - An integer that instructs syntribos to only use fuzz strings that meet certain length requirements
Meta variable files inherit based on the directory it's in. That is, if you have foo/meta.json and foo/bar/meta.json, templates in foo/bar/ will take their meta variable values from foo/bar/meta.json, but they can also reference meta variables that are defined only in foo/meta.json. This also means that templates in foo/baz/ cannot reference variables defined only in foo/bar/meta.json.
Each directory can have no more than one file named meta.json.
As mentioned above, some tests included with syntribos by default are: LDAP injection, SQL injection, integer overflow, command injection, XML external entity, reflected cross-site scripting, Cross Origin Resource Sharing (CORS), SSL, Regex Denial of Service, JSON Parser Depth Limit, and User defined.
In order to run a specific test, use the -t, --test-types
option and provide syntribos
with a keyword, or keywords, to match from
the test files located in syntribos/tests/
.
For SQL injection tests, see below:
$ syntribos --config-file keystone.conf -t SQL run
To run SQL injection tests against the template body only, see below:
$ syntribos --config-file keystone.conf -t SQL_INJECTION_BODY run
For all tests against HTTP headers only, see below:
$ syntribos --config-file keystone.conf -t HEADERS run
To execute unit tests automatically, navigate to the syntribos
root
directory and install the test requirements.
$ pip install -r test-requirements.txt
Now, run the unittest
as below:
$ python -m unittest discover tests/unit -p "test_*.py"
If you have configured tox you could also run the following:
$ tox -e py27 $ tox -e py35
This will run all the unit tests and give you a result output containing the status and coverage details of each test.
Syntribos is an open source project and contributions are always welcome. If you have any questions, we can be found in the #openstack-security channel on Freenode IRC.
- Follow all the OpenStack Style Guidelines (e.g. PEP8, Py3 compatibility)
- Follow secure coding guidelines
- Ensure all classes/functions have appropriate docstrings in RST format
- Include appropriate unit tests for all new code(place them in the
tests/unit
folder) - Test any change you make using tox:
pip install tox tox -e pep8 tox -e py27 tox -e py35 tox -e cover
Anyone wanting to contribute to OpenStack must follow the OpenStack development workflow
Submit all changes through the code review process in Gerrit described above. All pull requests on Github will be closed/ignored.
File bugs on the syntribos launchpad site, and not on Github. All Github issues will be closed/ignored.
Submit blueprints here for all breaking changes, feature requests, and other unprioritized work.
Note
README.rst is a file that can be generated by running
python readme.py
from the syntribos/scripts
directory. When the
README file needs to be updated; modify the corresponding rst file in
syntribos/doc/source
and have it generate by running the script.