Skip to content

ModSecurity Regression Test set - Waf testing master thesis

License

Notifications You must be signed in to change notification settings

Sebitosh/MRTS

Β 
Β 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

11 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

MRTS

MRTS is a utility that helps you create rule sets and their tests for ModSecurity or ModSecurity compliant engines (eg. Coraza) for regression testing. The format of the test cases is compatible with go-ftw.

Please note that this project is in very beta state.

Goals

The goals of this project:

  • create as many rules as possible for ModSecurity to test its behavior
  • create as many tests as possible for each rule

ModSecurity uses its rules targets, operators, transformations (special actions) and so many other components. It is necessary to test their behavior.

Note, that libmodsecurity3 has a regression test framework with several test cases, but it tests only the library, not the embedded state. For example we don't know anything about behavior of Nginx connector.

With the generated rules and tests we can check the operation of mod-security2 and Nginx-connector.

The generated rules can help in the quality assurance of these engines, e.g. after sending pull requests, we can verify that the change did not change the expected behavior.

Idea

The idea is to generate rules to see what happens to a particular component. It's not as trivial as it seems at first glance. Consider there are 5 phases - can we be sure of behaviors are same in each phases? Variables can be collections, every combinations of collections works as we need? Do you want to check the variable against multiple operator? With multiple operator arguments?

It's easy to see that the number of possible combinations can be infinite. It could be too much efford to write a rule for every possible format - and a test case too.

Instead of doing this, we can make a description about the object, and expand the possible combinations and their test cases.

Rules are generated based on templates. You can define as many templates as you want, and you can apply them for each rule description.

The operation is very simple: create one or more configuration files, and run the generator script with those files. the format of the files is some structured data (YAML, JSON) which can be human readable (and writable). Generator will produce rules with combination of given:

  • target + colkeys (collection keys) (eg. ARGS:arg1, ARGS:arg2, ARGS:arg1|ARGS:arg2)
  • operators (you can pass multiple operators)
  • operator arguments - also can pass several arguments
  • phases - it depends on your choose, in which phases you want to check the target

API

The framework has an API that describes which keywords can be used for the description. To avoid unwanted typing, there are several global settings that are derived in each case.

The syntax of API can be YAML or JSON.

Global keywords

Every global settings should be put under the global keyword, eg:

global:
  version: MRTS/0.1
  baseid: 100000

You can place global keywords in every file, each subsequent occurrence will overwrite the previous one. The files are processed in ABC order, later overwriting does not change the previous settings.

global

This keyword shows that the next block contains global settings.

version

version shows the current version of framework and can appear as constant in templates (see later).

baseid

baseid defines the first id what a rule can use. Inside the generator increments that for every rule, and that variable is avaluable as $CURRID (see later).

default_operator

This global variable defines the default operator for rules. You can overwrite it at every case, moreover you can add more operators for every case. But if you don't want to type, the operator member can be omitted.

Syntax:

global:
  default_operator: "@rx"

templates

templates defines a list of templates. Each item in the list is a template block - see template section.

default_tests_phase_methods

This keyword describes an object. Each keys of the object is a phase value, and the value is the method what you prefer to send the request during the test (with go-ftw). In phase:1 we prefer to use GET method, in case of each other the POST. Example:

global:
  default_tests_phase_methods
  - 1: get
  - 2: post
  - 3: post
  - 4: post
  - 5: post

Template

You can create one or more template which can be used for generated rules. A template object has two other named objects: name and template.

name must be a unique name, and template is a text with the rule definition. This definition can contain macros - see [macros][#macros] section.

An example for templates:

  - name: "SecRule for TARGETS"
    template: |
      SecRule $TARGET "$OPERATOR $OPARG" \
          "id:$CURRID,\
          phase:$PHASE,\
          deny,\
          t:none,\
          log,\
          msg:'%{MATCHED_VAR_NAME} was caught in phase:$PHASE',\
          ver:'$VERSION'"

As you can see the teplate macros begins with the dollar sign ($).

macros

Marcos are coming from the definition. That can be from the unique definition or if there no such variable, then from the globals.

Avaliable macros:

  • $TARGET the variable name when you want to check the SecRule's variable
  • $OPERATOR is the used operator; it must be placed with the leading @, eg. @rx.
  • $OPARG is the argument of the operator in the rule
  • $CURRID is the incremented id, which guaranties that every generated rule will have a unique id
  • $PHASE is the current phase in the list that you define in the definition file (see later its syntax)
  • $VERSION is the VERSION, see above

Please note that %{MATCHED_VAR_NAME} is not a tool macro, but the ModSecurity's macro. You can use them where you want.

Definition

In a definition file there also many keywords are avaliable. See an example then expand the meanings:

target: null
rulefile: MRTS_001_INIT.conf
testfile: null
objects:
- object: secaction
  actions:
    id: 10001
    phase: 1
    pass: null
    nolog: null
    msg: "'Initial settings'"
    ctl: ruleEngine=DetectionOnly
- object: secrule
  target: REQUEST_HEADERS:X-MRTS-Test
  operator: '@rx ^.*$'
  actions:
    id: 10002
    phase: 1
    pass: null
    t: none
    log: null
    msg: "'%{MATCHED_VAR}'"

or

target: ARGS_COMBINED_SIZE
rulefile: MRTS_003_ARGS_COMBINED_SIZE.conf
testfile: MRTS_003_ARGS_COMBINED_SIZE.yaml
templates:
- SecRule for TARGETS
colkey:
- - ''
operator:
- '@lt'
oparg:
- 2
testdata:
  phase_methods:
    1: get
    2: post
    3: post
    4: post
  targets:
    - target: 2
      test:
        data:
          foo: attack
    - target: arg1
      test:
        data:
          arg1: attack
    - target: arg2
      test:
        data:
          arg2: attack
  • target - defines the variable name what you want to test; it can be null, but then you must define the expected rules or actions under the object block
  • rulefile - the name of generated file; the path will be passed as cli argument, you should define here the relative path
  • testfile - the name of generated test file; can be null if you don't want to make tests against rules. The path here also will be passed as cli argument.
  • objects - a list type item, you can order the object which describes a SecRule or a SecAction. This is necessary because there are some special rules/actions, which can't described as regular rule. The first example generates the file MRTS_001_INIT.conf with a SecAction and a SecRule:
SecAction \
    "id:10001,\
    phase:1,\
    pass,\
    nolog,\
    msg:'Initial settings',\
    ctl:ruleEngine=DetectionOnly"

SecRule REQUEST_HEADERS:X-MRTS-Test "@rx ^.*$"\
    "id:10002,\
    phase:1,\
    pass,\
    t:none,\
    log,\
    msg:'%{MATCHED_VAR}'"

These are necessary for go-ftw.

  • templates - you can list the name of templates what you want to apply
  • colkey - list collection keys what you want to test; note that each item in the list is a list too! See this example:
colkey:
- - ''
- - arg1
- - arg1
  - arg2
- - /^arg_.*$/

will produce: [[''], ['arg1'], ['arg2'], ['arg1', 'arg2'], ['/^arg_.*$']]. This will generate rules with targets:

SecRule ARGS
SecRule ARGS:arg1
SecRule ARGS:arg1|ARGS:arg2
SecRule ARGS:/^arg_.*$/
  • operator - list of used operators
  • oparg - list of used operator arguments
  • testdata - list of expected test cases - see testdata section

testdata

testdata is a keyword in the definition file. Here you can list the necessary test case definitions. A testdata item can contain two member:

  • phase_methods - where you can owerwrite the default_tests_phase_methods - this keyword is optional
  • targets - here you can define the posible collection keys that can occurres in generated rules

test case defition

Let's see a test case definition example:

  targets:
    - target: ''
      test:
        data:
          foo: attack
    - target: arg1
      test:
        data:
          arg1: attack

As it described above, targets is list of tests. A test case contains two keywords:

  • target - describes the collection key which used at the rule (can be empty: ``)
  • test - is an object

The test object can contains these keywords:

  • data - which can be a single string or a key:value pair
  • input - a structure which overrides the test case in predefined structure

Note, that the go-ftw test structure is hard-coded in the script, the input overwrites that structure.

The given example above contains two test cases: one if the collection key is empty, and another one if the collection key is the arg1 - see the generated rules example above. You must give at least one test for each used collection keys at the rules definition!

Here are some examples for test cases:

  targets:
    - target: ''
      test:
        data:
          foo: attack
    - target: ''
      test:
        data:
          arg1: attack

This will generate two test cases for empty collection key with data: foo=attack and arg1=attack.

    - target: ''
      test:
        data:
          foo: attack
    - target: arg1
      test:
        data:
          arg1: attack

This will generate one test for empty collection key and one for the collection key arg1. The data for the first case will be foo=attack and arg1=attack for the second.

  targets:
    - target: '/*'
      test:
        data:
          <level1><level2>foo</level2><level2>bar</level2></level1>
        input:
          headers:
            - name: Content-Type
              value: application/xml

This will genreate a test case for collection key /* (usually used for XML), the data will be the given XML string, and the test add an extra header for go-ftw test.

Run the tool

To generate the rules and their tests, run the tool:

$ ./generate-rules.py 
usage: generate-rules.py [-h] -r [/path/to/mrts/*.yaml ...] -e /path/to/mrts/rules/ -t /path/to/mrts/tests/
generate-rules.py: error: the following arguments are required: -r/--rulesdef, -e/--expdir, -t/--testdir

As you can see there are few command line arguments.

  • -r - rules' definition files
  • -e - export directory where rules will be written
  • -t - export test directory where tests will be written
$ ./mrts/generate-rules.py -r config_tests/*.yaml -e generated/rules/ -t generated/tests/regression/tests/

Once generated, rules need to be added to your ModSecurity configuration file.

Change mΜ€rts.load with your absolute path to the generated rules:

Include /Absolute/Path/To/MRTS/generated/rules/*.conf

In modsecurity.conf include your absolute path to mrts.load:

...

Include /Absolute/Path/To/MRTS/mrts.load

Don't forget to restart your server each time you generate new rules.

If you finished the generation and configuration process, you can download go-ftw and run it.

For more info about go-ftw please see its README or CRS's excellent documentation.

Here is an example:

$ cat .ftw.apache-mrts.yaml 
---
logfile: '/var/log/apache2/error.log'
logmarkerheadername: 'X-MRTS-TEST'
logtype:
  name: 'apache'
  timeregex:  '\[([A-Z][a-z]{2} [A-z][a-z]{2} \d{1,2} \d{1,2}\:\d{1,2}\:\d{1,2}\.\d+? \d{4})\]'
  timeformat: 'ddd MMM DD HH:mm:ss.S YYYY'


$ ./go-ftw run --config .ftw.apache-mrts.yaml -d generated/tests/regression/tests/
πŸ› οΈ Starting tests!
πŸš€ Running go-ftw!
πŸ‘‰ executing tests in file MRTS_002_ARGS.yaml
	running 2-1: βœ” passed in 12.842548ms (RTT 54.970028ms)
	running 2-2: βœ” passed in 12.049459ms (RTT 54.891019ms)
	running 2-3: βœ” passed in 10.790834ms (RTT 53.365412ms)
	running 2-4: βœ” passed in 10.695786ms (RTT 53.515826ms)
πŸ‘‰ executing tests in file MRTS_002_ARGS.yaml
	running 2-1: βœ” passed in 8.615306ms (RTT 52.334647ms)
	running 2-2: βœ” passed in 7.64326ms (RTT 52.301444ms)
	running 2-3: βœ” passed in 8.353395ms (RTT 52.289161ms)
	running 2-4: βœ” passed in 8.704224ms (RTT 52.993254ms)
...

Check the state of covered variables

When you finished the build process, you can check which variables (and later the other entities) are covered by the generated rule set.

You should type:

$ cd mrts/collect_rules

$ ./collect-rules.py 
usage: collect-rules.py [-h] -r [/path/to/mrts/*.conf ...]
collect-rules.py: error: the following arguments are required: -r/--rules

As you can see here are also a mandatory argument, the path of generated rules.

$ ./collect-rules.py -r ../../generated/rules/*.conf
Config file: ../../generated/rules/MRTS_001_INIT.conf
 Parsing ok.
Config file: ../../generated/rules/MRTS_002_ARGS.conf
 Parsing ok.
Config file: ../../generated/rules/MRTS_003_ARGS_COMBINED_SIZE.conf
 Parsing ok.
Config file: ../../generated/rules/MRTS_004_ARGS_GET.conf
 Parsing ok.
Config file: ../../generated/rules/MRTS_005_ARGS_GET_NAMES.conf
 Parsing ok.
Config file: ../../generated/rules/MRTS_110_XML.conf
 Parsing ok.

=====
Covered TARGETs: REQUEST_HEADERS, ARGS, ARGS_COMBINED_SIZE, ARGS_GET, ARGS_GET_NAMES, XML

UNCOVERED TARGETs: ARGS_NAMES, ARGS_POST, ARGS_POST_NAMES, ...

Based on the output, we actually covered 6 targets, so there are lot of works to cover all variables.

About

ModSecurity Regression Test set - Waf testing master thesis

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%