Troubleshooting DAST browser-based analyzer

The following troubleshooting scenarios have been collected from customer support cases. If you experience a problem not addressed here, or the information here does not fix your problem, create a support ticket. For more details, see the GitLab Support page.

When something goes wrong

When something goes wrong with a DAST scan, if you have a particular error message then check known problems.

Otherwise, try to discover the problem by answering the following questions:

What is the expected outcome?

Many users who encounter issues with a DAST scan have a good high-level idea of what they think the scanner should be doing. For example, it’s not scanning particular pages, or it’s not selecting a button on the page.

As much as possible, try to isolate the problem to help narrow the search for a solution. For example, take the situation where DAST isn’t scanning a particular page. From where should DAST have found the page? What path did it take to navigate there? Were there elements on the referring page that DAST should have selected, but did not?

Is the outcome achievable by a human?

DAST cannot scan an application if a human cannot manually traverse the application.

Knowing the outcome you expect, try to replicate it manually using a browser on your machine. For example:

  • Open a new incognito/private browser window.
  • Open Developer Tools. Keep an eye on the console for error messages.
    • In Chrome: View -> Developer -> Developer Tools.
    • In Firefox: Tools -> Browser Tools -> Web Developer Tools.
  • If authenticating:
    • Navigate to the DAST_AUTH_URL.
    • Type in the DAST_USERNAME in the DAST_USERNAME_FIELD.
    • Type in the DAST_PASSWORD in the DAST_PASSWORD_FIELD.
    • Select the DAST_SUBMIT_FIELD.
  • Select links and fill in forms. Navigate to the pages that aren’t scanning correctly.
  • Observe how your application behaves. Notice if there is anything that might cause problems for an automated scanner.

Any reason why DAST would not work?

DAST cannot scan correctly when:

  • There is a CAPTCHA. Turn these off in the testing environment for the application being scanned.
  • It does not have access to the target application. Ensure the GitLab Runner can access the application using the URLs used in the DAST configuration.

How does your application work?

Understanding how your application works is vital to figuring out why a DAST scan isn’t working. For example, the following situations may require additional configuration settings.

  • Is there a popup modal that hides elements?
  • Does a loaded page change dramatically after a certain period of time?
  • Is the application especially slow or fast to load?
  • Is the target application jerky while loading?
  • Does the application work differently based on the client’s location?
  • Is the application a single-page application?
  • Does the application submit HTML forms, or does it use JavaScript and AJAX?
  • Does the application use websockets?
  • Does the application use a specific web framework?
  • Does selecting buttons run JavaScript before continuing the form submit? Is it fast, slow?
  • Is it possible DAST could be selecting or searching for elements before either the element or page is ready?

What is DAST doing?

Logging remains the best way to understand what DAST is doing:

Browser-based analyzer logging

The analyzer log is one of the most useful tools to help diagnose problems with a scan. Different parts of the analyzer can be logged at different levels.

Log message format

Log messages have the format [time] [log level] [log module] [message] [additional properties].

For example, the following log entry has level INFO, is part of the CRAWL log module, has the message Crawled path and the additional properties nav_id and path.

2021-04-21T00:34:04.000 INF CRAWL Crawled path nav_id=0cc7fd path="LoadURL [https://my.site.com:8090]"

Log destination

Logs are sent either to file or to console (the CI/CD job log). You can configure each destination to accept different logs using the environment variables DAST_BROWSER_LOG for console logs and DAST_BROWSER_FILE_LOG for file logs.

For example:

include:
  - template: DAST.gitlab-ci.yml

dast:
  variables:
    DAST_BROWSER_SCAN: "true"
    DAST_BROWSER_LOG: "auth:debug"                               # console log defaults to INFO level, logs AUTH module at DEBUG
    DAST_BROWSER_FILE_LOG: "loglevel:debug,cache:warn"           # file log defaults to DEBUG level, logs CACHE module at WARN
    DAST_BROWSER_FILE_LOG_PATH: "$CI_PROJECT_DIR/dast-scan.log"  # Save the file log in the project directory so it can be recognized as an artifact
  artifacts:
    paths:
      - dast-scan.log
    when: always

Log levels

The log levels that can be configured are as follows:

Log moduleComponent overviewMore
TRACEUsed for specific, often noisy inner workings of a feature. 
DEBUGDescribes the inner-workings of a feature. Used for diagnostic purposes. 
INFODescribes the high level flow of the scan and the results.Default level if none specified.
WARNDescribes an error situation where DAST recovers and continues the scan. 
FATAL/ERROR/PANIC Describes unrecoverable errors prior to exit. 

Log modules

LOGLEVEL configures the default log level for the log destination. If any of the following modules are configured, DAST uses the log level for that module in preference to the default log level.

The modules that can be configured for logging are as follows:

Log moduleComponent overview
ACTIVUsed for active attacks.
AUTHUsed for creating an authenticated scan.
BPOOLThe set of browsers that are leased out for crawling.
BROWSUsed for querying the state or page of the browser.
CACHEUsed for reporting on cache hit and miss for cached HTTP resources.
CHROMUsed to log Chrome DevTools messages.
CONTAUsed for the container that collects parts of HTTP requests and responses from DevTools messages.
CRAWLUsed for the core crawler algorithm.
DATABUsed for persisting data to the internal database.
LEASEUsed to create browsers to add them to the browser pool.
MAINUsed for the flow of the main event loop of the crawler.
NAVDBUsed for persistence mechanisms to store navigation entries.
REGEXUsed for recording performance statistics when running regular expressions.
REPTUsed for generating reports.
STATUsed for general statistics while running the scan.
VLDFNUsed for loading and parsing vulnerability definitions.
WEBGWUsed to log messages sent to the target application when running active checks.

Example - log crawled paths

Set the log module CRAWL to DEBUG to log navigation paths found during the crawl phase of the scan. This is useful for understanding if DAST is crawling your target application correctly.

include:
  - template: DAST.gitlab-ci.yml

dast:
  variables:
    DAST_BROWSER_LOG: "crawl:debug"

For example, the following output shows that four anchor links we discovered during the crawl of the page at https://example.com.

2022-11-17T11:18:05.578 DBG CRAWL executing step nav_id=6ec647d8255c729160dd31cb124e6f89 path="LoadURL [https://example.com]" step=1
...
2022-11-17T11:18:11.900 DBG CRAWL found new navigations browser_id=2243909820020928961 nav_count=4 nav_id=6ec647d8255c729160dd31cb124e6f89 of=1 step=1
2022-11-17T11:18:11.901 DBG CRAWL adding navigation action="LeftClick [a href=/page1.html]" nav=bd458cc1fc2d7c6fb984464b6d968866 parent_nav=6ec647d8255c729160dd31cb124e6f89
2022-11-17T11:18:11.901 DBG CRAWL adding navigation action="LeftClick [a href=/page2.html]" nav=6dcb25f9f9ece3ee0071ac2e3166d8e6 parent_nav=6ec647d8255c729160dd31cb124e6f89
2022-11-17T11:18:11.901 DBG CRAWL adding navigation action="LeftClick [a href=/page3.html]" nav=89efbb0c6154d6c6d85a63b61a7cdc6f parent_nav=6ec647d8255c729160dd31cb124e6f89
2022-11-17T11:18:11.901 DBG CRAWL adding navigation action="LeftClick [a href=/page4.html]" nav=f29b4f4e0bdee70f5255de7fc080f04d parent_nav=6ec647d8255c729160dd31cb124e6f89

Chromium DevTools logging

caution
Logging DevTools messages is a security risk. The output contains secrets such as usernames, passwords and authentication tokens. The output is uploaded to the GitLab server and may be visible in job logs.

The DAST Browser-based scanner orchestrates a Chromium browser using the Chrome DevTools Protocol. Logging DevTools messages helps provide transparency into what the browser is doing. For example, if selecting a button does not work, a DevTools message might show that the cause is a CORS error in a browser console log. Logs that contain DevTools messages can be very large in size. For this reason, it should only be enabled on jobs with a short duration.

To log all DevTools messages, turn the CHROM log module to trace and configure logging levels. The following are examples of DevTools logs:

2022-12-05T06:27:24.280 TRC CHROM event received    {"method":"Fetch.requestPaused","params":{"requestId":"interception-job-3.0","request":{"url":"http://auth-auto:8090/font-awesome.min.css","method":"GET","headers":{"Accept":"text/css,*/*;q=0.1","Referer":"http://auth-auto:8090/login.html","User-Agent":"Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) HeadlessChrome/105.0.5195.102 Safari/537.36"},"initialPriority":"VeryHigh","referrerPolicy":"strict-origin-when-cross-origin"},"frameId":"A706468B01C2FFAA2EB6ED365FF95889","resourceType":"Stylesheet","networkId":"39.3"}} method=Fetch.requestPaused
2022-12-05T06:27:24.280 TRC CHROM request sent      {"id":47,"method":"Fetch.continueRequest","params":{"requestId":"interception-job-3.0","headers":[{"name":"Accept","value":"text/css,*/*;q=0.1"},{"name":"Referer","value":"http://auth-auto:8090/login.html"},{"name":"User-Agent","value":"Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) HeadlessChrome/105.0.5195.102 Safari/537.36"}]}} id=47 method=Fetch.continueRequest
2022-12-05T06:27:24.281 TRC CHROM response received {"id":47,"result":{}} id=47 method=Fetch.continueRequest

Customizing DevTools log levels

Chrome DevTools requests, responses and events are namespaced by domain. DAST allows each domain and each domain with message to have different logging configuration. The environment variable DAST_BROWSER_DEVTOOLS_LOG accepts a semi-colon separated list of logging configurations. Logging configurations are declared using the structure [domain/message]:[what-to-log][,truncate:[max-message-size]].

  • domain/message references what is being logged.
    • Default can be used as a value to represent all domains and messages.
    • Can be a domain, for example, Browser, CSS, Page, Network.
    • Can be a domain with a message, for example, Network.responseReceived.
    • If multiple configurations apply, the most specific configuration is used.
  • what-to-log references whether and what to log.
    • message logs that a message was received and does not log the message content.
    • messageAndBody logs the message with the message content. Recommended to be used with truncate.
    • suppress does not log the message. Used to silence noisy domains and messages.
  • truncate is an optional configuration to limit the size of the message printed.

Example - log all DevTools messages

Used to log everything when you’re not sure where to start.

include:
  - template: DAST.gitlab-ci.yml

dast:
  variables:
    DAST_BROWSER_FILE_LOG: "chrom:trace"
    DAST_BROWSER_FILE_LOG_PATH: "/zap/wrk/dast-scan.log"
    DAST_BROWSER_DEVTOOLS_LOG: "Default:messageAndBody,truncate:2000"
  artifacts:
    paths:
      - dast-scan.log
    when: always

Example - log HTTP messages

Useful for when a resource isn’t loading correctly. HTTP message events are logged, as is the decision to continue or fail the request. Any errors in the browser console are also logged.

include:
  - template: DAST.gitlab-ci.yml

dast:
  variables:
    DAST_BROWSER_FILE_LOG: "chrom:trace"
    DAST_BROWSER_FILE_LOG_PATH: "/zap/wrk/dast-scan.log"
    DAST_BROWSER_DEVTOOLS_LOG: "Default:suppress;Fetch:messageAndBody,truncate:2000;Network:messageAndBody,truncate:2000;Log:messageAndBody,truncate:2000;Console:messageAndBody,truncate:2000"
  artifacts:
    paths:
      - dast-scan.log
    when: always

Chromium logs

In the rare event that Chromium crashes, it can be helpful to write the Chromium process STDOUT and STDERR to log. Setting the environment variable DAST_BROWSER_LOG_CHROMIUM_OUTPUT to true achieves this purpose.

DAST starts and stops many Chromium processes. DAST sends each process output to all log destinations with the log module LEASE and log level INFO.

For example:

include:
  - template: DAST.gitlab-ci.yml

dast:
  variables:
    DAST_BROWSER_LOG_CHROMIUM_OUTPUT: "true"

Known problems

Logs contain response body exceeds allowed size

By default DAST processes HTTP requests where the HTTP response body is 10 MB or less. Otherwise, DAST blocks the response which can cause scans to fail. This constraint is intended to reduce memory consumption during a scan.

An example log is as follows, where DAST blocked the JavaScript file found at https://example.com/large.js as it’s size is greater than the limit:

2022-12-05T06:28:43.093 WRN BROWS response body exceeds allowed size allowed_size_bytes=1000000 browser_id=752944257619431212 nav_id=ae23afe2acbce2c537657a9112926f1a of=1 request_id=interception-job-2.0 response_size_bytes=9333408 step=1 url=https://example.com/large.js
2022-12-05T06:28:58.104 WRN CONTA request failed, attempting to continue scan error=net::ERR_BLOCKED_BY_RESPONSE index=0 requestID=38.2 url=https://example.com/large.js

This can be changed using the configuration DAST_MAX_RESPONSE_SIZE_MB. For example,

include:
  - template: DAST.gitlab-ci.yml

dast:
  variables:
    DAST_MAX_RESPONSE_SIZE_MB: "25"