Nuclei supports fuzzing of HTTP requests based on rules defined in the fuzzing section of the HTTP request. This allows creating templates for generic Web Application vulnerabilities like SQLi, SSRF, CMDi, etc without any information of the target like a classic web fuzzer. We call this concept as Fuzzing for Unknown Vulnerabilities.

pre-condition

More often than not, we want to only attempt fuzzing on those requests where it makes sense. For example,

  • Fuzz Body When Body is Present
  • Ignore PreFlight and CONNECT requests

and so on. With Nuclei v3.2.4 we have introduced a new pre-condition section which contains conditions when the fuzzing template should be executed.

pre-condition can be considered a twin of matchers in nuclei. They support all matcher types, including DSL, and the only difference is that this serves a different purpose.

For example, to only execute template on POST request with some body, you can use the following filter.

- pre-condition:
    - type: dsl
      dsl:
        - method == POST
        - len(body) > 0
      condition: and
Currently, Only request data like header, host, input, method, path, etc is available, but soon, response data will be available once the support for loading the response along with the request is added.

When writing/executing a template, you can use the -v -svd flags to see all variables available in filters before applying the filter.

Part

Part specifies what part of the request should be fuzzed based on the specified rules. Available options for this parameter are -

query (default) - fuzz query parameters for URL

fuzzing:
  - part: query # fuzz parameters in URL query

path - fuzz path parameters for requests

fuzzing:
  - part: path # fuzz path parameters

header - fuzz header parameters for requests

fuzzing:
  - part: header # fuzz headers

cookie - fuzz cookie parameters for requests

fuzzing:
  - part: cookie # fuzz cookies

body - fuzz body parameters for requests

fuzzing:
  - part: body # fuzz parameters in body

Special Part

request - fuzz the entire request (all parts mentioned above)

fuzzing:
  - part: request # fuzz entire request

Multiple selective parts

Multiple parts can be selected for fuzzing by defining a parts field which is the plural of above allowing selected multiple parts to be fuzzed.

fuzzing:
  - parts:
      - query
      - body
      - header

Type

Type specifies the type of replacement to perform for the fuzzing rule value. Available options for this parameter are -

  1. replace (default) - replace the value with payload
  2. prefix - prefix the value with payload
  3. postfix - postfix the value with payload
  4. infix - infix the value with payload (place in between)
  5. replace-regex - replace the value with payload using regex
fuzzing:
  - part: query
    type: postfix # Fuzz query and postfix payload to params

Key-Value Abstraction

In a HTTP request, there are various parts like query, path, headers, cookies, and body and each part has different in various formats. For example, the query part is a key-value pair, the path part is a list of values, the body part can be a JSON, XML, or form-data.

To effectively abstract these parts and allow them to be fuzzed, Nuclei exposes these values as key and value pairs. This allows users to fuzz based on the key or value of the request part.

For example, Below sample HTTP request can be abstracted as key-value pairs as shown below.

POST /reset-password?token=x0x0x0&source=app HTTP/1.1
Host: 127.0.0.1:8082
User-Agent: Go-http-client/1.1
Cookie: PHPSESSID=1234567890
Content-Length: 23
Content-Type: application/json
Accept-Encoding: gzip
Connection: close

{"password":"12345678"}
  • part: Query
keyvalue
tokenx0x0x0
sourceapp
  • part: Path
keyvalue
value/reset-password
  • part: Header
keyvalue
Host127.0.0.1:8082
User-AgentGo-http-client/1.1
Content-Length23
Content-Typeapplication/json
Accept-Encodinggzip
Connectionclose
  • part: Cookie
keyvalue
PHPSESSID1234567890
  • part: Body
keyvalue
password12345678

Note: XML, JSON, Form, Multipart-FormData will be in kv format, but if the Body is binary or in any other format, the entire Body will be represented as a single key-value pair with key as value and value as the entire Body.

keyvalue
value”\x08\x96\x01\x12\x07\x74”

This abstraction really levels up the game since you only need to write a single rule for the Body, and it will be applied to all formats. For example, if you check for SQLi in body values, a single rule will work on all formats, i.e., JSON, XML, Form, Multipart-FormData, etc.

Mode

Mode specifies the mode in which to perform the replacements. Available modes are -

  1. multiple (default) - replace all values at once
  2. single - replace one value at a time
fuzzing:
  - part: query
    type: postfix
    mode: multiple # Fuzz query postfixing payloads to all parameters at once

Note: default values are set/used when other options are not defined.

Component Data Filtering

Multiple filters are supported to restrict the scope of fuzzing to only interesting parameter keys and values. Nuclei HTTP Fuzzing engine converts request parts into Keys and Values which then can be filtered by their related options.

The following filter fields are supported -

  1. keys - list of parameter names to fuzz (exact match)
  2. keys-regex - list of parameter regex to fuzz
  3. values - list of value regex to fuzz

These filters can be used in combination to run highly targeted fuzzing based on the parameter input. A few examples of such filtering are provided below.

# fuzzing command injection based on parameter name value
fuzzing:
  - part: query
    type: replace
    mode: single
    keys:
      - "daemon"
      - "upload"
      - "dir"
      - "execute"
      - "download"
      - "log"
      - "ip"
      - "cli"
      - "cmd"
# fuzzing openredirects based on parameter name regex
fuzzing:
  - part: query
    type: replace
    mode: single
    keys-regex:
      - "redirect.*"
# fuzzing ssrf based on parameter value regex
fuzzing:
  - part: query
    type: replace
    mode: single
    values:
      - "https?://.*"

Fuzz

Fuzz specifies the values to replace with a type for a parameter. It supports payloads, DSL functions, etc and allows users to fully utilize the existing nuclei feature-set for fuzzing purposes.

# fuzz section for xss fuzzing with stop-at-first-match
payloads:
  reflection:
    - "6842'\"><9967"
stop-at-first-match: true
fuzzing:
  - part: query
    type: postfix
    mode: single
    fuzz:
      - "{{reflection}}"
# using interactsh-url placeholder for oob testing
payloads:
  redirect:
    - "{{interactsh-url}}"
fuzzing:
  - part: query
    type: replace
    mode: single
    keys:
      - "dest"
      - "redirect"
      - "uri"
    fuzz:
      - "https://{{redirect}}"
# using template-level variables for SSTI testing
variables:
  first: "{{rand_int(10000, 99999)}}"
  second: "{{rand_int(10000, 99999)}}"
  result: "{{to_number(first)*to_number(second)}}"

http:
    ...
    payloads:
      reflection:
        - '{{concat("{{", "§first§*§second§", "}}")}}'
    fuzzing:
      - part: query
        type: postfix
        mode: multiple
        fuzz:
          - "{{reflection}}"

Analyzer

Analyzers is a new concept introduced in nuclei fuzzing which allow the engine to make additional verification requests based on a certain logic to verify the vulnerability.

time_delay

The time_delay analyzer verifies that the response time of the request is controllable by the fuzzed payload. It uses a Linear Regression algorithm ported from ZAP with alternating requests to determine the server time is actually controllable rather than just noise. You can configure it like so

# Create a new time delay analyzer
analyzer:
  name: time_delay
  # Optionally, you can define parameters for the
  # analyzer like below.
  # 
  # the defaults are good enough for most use cases. 
  parameters:
    sleep_duration: 10 # sleep for 10 seconds (default: 5)
    requests_limit: 6 # make 6 verification requests (default: 4)
    time_correlation_error_range: 0.30 # error range for time correlation (default: 0.15)
    time_slope_error_range: 0.40 # error range for time slope (default: 0.30)

The following dynamic placeholders are available in payloads with time_delay analyzer.

  • [SLEEPTIME] - The sleep time in seconds for the time delay analyzer.
  • [INFERENCE] - The inference condition (%d=%d) for the time delay analyzer.

These values are substituted at runtime with the actual values for the analyzer. The following is how a usual verification process looks.

  1. Send the request with the payload to the target with 5 second delay.
  2. If the response time is less than 5, do nothing.
  3. Send the request to the analyzer which queues it with 5 seconds delay.
  4. Next a 1 second delay
  5. Next a 5 second delay
  6. Finally, the last 1 second delay.

If the response time is controllable, the analyzer will report the vulnerability.

Matching for the analyzer matches is pretty straightforward as well. Simiar to interactsh, you can use part: analyzer to match the analyzer response.

matchers:
  - type: word
    part: analyzer
    words:
      - "true"

Optionally, you can also extract the analyzer_details from the analyzer for matches.

Example Fuzzing template

An example sample template for fuzzing XSS vulnerabilities is provided below.

id: fuzz-reflection-xss

info:
  name: Basic Reflection Potential XSS Detection
  author: pdteam
  severity: low

http:
  - pre-condition:
      - type: dsl
        dsl:
          - 'method == "GET"'       # only run if method is GET
    payloads:
      reflection:
        - "6842'\"><9967"

    stop-at-first-match: true
    fuzzing:
      - part: query
        type: postfix
        mode: single
        fuzz:
          - "{{reflection}}"

    matchers-condition: and
    matchers:
      - type: word
        part: body
        words:
          - "{{reflection}}"

      - type: word
        part: header
        words:
          - "text/html"

More complete examples are provided here