Skip to content

GitHub Data Loss Prevention (DLP) Action: Scan Pull Requests for sensitive data, like credentials & secrets, PII, credit card numbers, and more.

License

Notifications You must be signed in to change notification settings

nightfallai/nightfall_dlp_action

Use this GitHub action with your project
Add this Action to an existing workflow or create a new one
View on Marketplace

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

37 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Nightfall DLP GitHub Action

nightfalldlp

Nightfall DLP Action: A code review tool that protects you from committing sensitive information to your repositories.

The Nightfall DLP Action scans your code commits upon Pull Request for sensitive information - like credentials & secrets, PII, credit card numbers & more - and posts review comments to your code hosting service automatically. The Nightfall DLP Action is intended to be used as a part of your CI to simplify the development process, improve your security, and ensure you never accidentally leak secrets or other sensitive information via an accidental commit.

Example

Here's an example of the Nightfall DLP GitHub Action providing feedback on a Pull Request:

nightfall-dlp-action-example

The action runs when configured as a job in your GitHub Workflow:

Note: you must use the actions/checkout step as shown below before the running the nightfalldlp action in order for it to function properly

name: nightfalldlp
on:
  push:
    branches:
      - master
  pull_request:
jobs:
  run-nightfalldlp:
    name: nightfalldlp
    runs-on: ubuntu-latest
    steps:
      - name: Checkout Repo Action
        uses: actions/checkout@v2

      - name: nightfallDLP action step
        uses: nightfallai/[email protected]
        env:
          NIGHTFALL_API_KEY: ${{ secrets.NIGHTFALL_API_KEY }}
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
          EVENT_BEFORE: ${{ github.event.before }}
    permissions:
      checks: write
      contents: read

Usage

1. Get a Nightfall API key.

The Nightfall DLP Action is powered by the Nightfall Developer Platform. With the Nightfall Developer Platform, you can inspect & classify your data, wherever it lives. Nightfall's APIs allow callers to programmatically get structured results from deep learning-based detectors for data types like credit card numbers, API keys, and more. Scan data easily in your own third-party apps, internal apps, and data silos. Leverage these classifications in your own workflows - for example, saving them to a data warehouse or pushing them to a SIEM. Sign up for a free account here!

2. Set up config file to specify your condition set.

  • Place a .nightfalldlp/ directory within the root of your target repository, and inside it a config.json file in which you can configure your condition sets (see Nightfalldlp Config File section below for more information on condition sets)

  • See Additional Configuration section for more advanced configuration options

  • If a config is not included, a default config consisting of a condition set with the API_KEY and CRYPTOGRAPHIC_KEY detectors will be used.

3. Set up a few environment variables. These variables should be made available to the nightfall_dlp_action by adding them to the env: key in your workflow:

  • NIGHTFALL_API_KEY

    • Get a free Nightfall DLP API Key by registering for an account with the Nightfall Developer Platform
    • Add this variable to your target repository's "GitHub Secrets" and passed in to your GitHub Workflow's env.
  • GITHUB_TOKEN

    • This is automatically injected by GitHub inside each Workflow (via the secrets context), you just need to set it to the env key. This variable should always point to secrets.GITHUB_TOKEN
    • This token is used to authenticate to GitHub to write Comments/Annotations to your Pull Requests and Pushes
  • EVENT_BEFORE (*only required for GitHub Workflows running on a push event)

    • the value for this var lives on the github context object in a Workflow - EVENT_BEFORE should always point to ${{ github.event.before }} (as seen in the example above)
  • BASE_URL (*for Enterprise only)

    • if you are using Github Enterprise, you must set this variable to your enterprise domain name to connect to the Github API.

Supported GitHub Events

The Nightfall DLP Action can run in a GitHub Workflow triggered by the following events:

  • PULL_REQUEST
  • PUSH

The Nightfall DLP Action is currently unable to be used in forked GitHub repositories due to GitHub's disabling of secrets sharing when Workflows originate from forks.

NightfallDLP Config File

The .nightfalldlp/config.json file is used as a centralized config file to control which detectors to scan with and what content you want to scan for in your pull requests. It includes the following adjustable fields to fit your needs based on the Nightfall Developer Platform. For additional detail, review the code scanner documentation here.

DetectionRuleUUIDs

A Detection Rule UUID is a unique identifier of a Detection Rule, which can be created via the Nightfall Console. Once defined, you can input a list of up to 10 pre-built detection rules in your config file, e.g.

{ "detectionRuleUUIDs": ["A0BA0D76-78B4-4E10-B653-32996060316B"] }

Detection Rules

Detection Rules contain a list of detectors specified inline. A detectionRule contains a list of detector objects, as well as a logicalOp and an optional name. The logicalOp dictates when the detection rule should surface an alert, depending on whether all detectors in the provided list trigger a match, or if at least one triggers a match. Valid values for the logicalOp are ANY (logical OR), and ALL (logical AND).

{
  "detectionRules": [
    {
      "name": "my rule",
      "logicalOp": "ANY",
      "detectors": [
        {
          "minNumFindings": 1,
          "minConfidence": "POSSIBLE",
          "detectorType": "NIGHTFALL_DETECTOR",
          "nightfallDetector": "CREDIT_CARD_NUMBER"
        }
      ]
    }
  ]
}

minNumFindings specifies the minimum number of findings required to return for one request, e.g. if you set minNumFindings to be 2, and Nightfall identifies only 1 finding within the request payload related to that detector, that finding then will be filtered out from the response.

minConfidence specifies the minimum threshold to trigger a potential finding to be captured. We have five levels of confidence from least sensitive to most sensitive:

  • VERY_LIKELY
  • LIKELY
  • POSSIBLE
  • UNLIKELY
  • VERY_UNLIKELY

Detector

A detector is either a pre-built Nightfall detector or custom regex or wordlist detector that you can create. This is specified by the detectorType field.

Nightfall Pre-Built Detector

{
  "detectors": [{
    "detectorType": "NIGHTFALL_DETECTOR",
    "nightfallDetector": "API_KEY",
    "displayName": "apiKeyDetector",
    "minNumFindings": 1,
    "minConfidence": "POSSIBLE"
  }]
}
  • Within a detector struct

    • First specify detectorType as NIGHTFALL_DETECTOR

    • Choose the Nightfall detector you would like to use from our Detector Glossary. The Glossary includes a broad set of PII, PHI, PCI, credentials & secrets, and more to choose from. We recommend the following three as a simple starting point:

      • API_KEY
      • CRYPTOGRAPHIC_KEY
      • PASSWORD_IN_CODE
    • Set a display name for your detector, as this will be attached on your findings

Custom Regex

We also support custom regular expressions as a detector, which are defined as follows:

{
  "detectors": [{
    "detectorType": "REGEX",
    "regex": {
      "pattern": "coupon:\\d{4,}",
      "isCaseSensitive": false
    },
    "displayName": "simpleRegexCouponDetector",
    "minNumFindings": 1,
    "minConfidence": "POSSIBLE"
  }]
}

Custom Word List

Word List detectors trigger when a string payload contains any of the words you specify in the detector definition. For example:

{
  "detectors": [{
    "detectorType": "WORD_LIST",
    "wordList": {
      "values": ["key", "credential"],
      "isCaseSensitive": false
    },
    "displayName": "simpleWordListKeyDetector",
    "minNumFindings": 1,
    "minConfidence": "POSSIBLE"
  }]
}

Additional Detector Configuration Options

Aside from specifying which detectors to use in your scan, you can also specify some additional rules to attach. They are contextRules and exclusionRules.

Context Rules

A context rule evaluates the surrounding context (i.e. preceding and following characters) of a finding and adjusts the finding's confidence if the input context rule pattern exists.

Example:

{
  "detectors": [{
    // ...... other detector fields
    "contextRules": [
      {
        "regex": {
          "pattern": "my cc",
          "isCaseSensitive": true
        },
        "proximity": {
          "windowBefore": 30,
          "windowAfter": 30
        },
        "confidenceAdjustment": {
          "fixedConfidence": "VERY_LIKELY"
        }
      }
    ]
  }]
}
  • regex defines a regular expression to trigger a finding
  • proximity is defined as the number preceding and trailing characters surrounding the finding in which to conduct the search
  • confidenceAdjustment is the confidence level to adjust the finding to if a match is detected

As an example, say we have the line of text my cc number: 4242-4242-4242-4242 in a file, and 4242-4242-4242-4242 is detected as a credit card number with a confidence of POSSIBLE. If we had the context rule above, the confidence level of this finding will be bumped up to VERY_LIKELY because the characters preceding the finding, my cc, match the regex.

Exclusion Rules

Exclusion rules on detectors are used to mute findings according to the defined conditions:

Example:

{
  "detectors": [{
    // ...... other detector fields
    "exclusionRules": [
      {
        "matchType": "PARTIAL",
        "exclusionType": "REGEX",
        // specify one of these values based on the type specified above
        "regex": {
          "pattern": "4242-4242-\\d{4}-\\d{4}",
          "isCaseSensitive": true
        },
        "wordList": {
          "values": ["4242"],
          "isCaseSensitive": false
        }
      }
    ]
  }]
}
  • exclusionType specifies either a REGEX or WORD_LIST
  • regex specifies a regular expression that, if matched would trigger exclusion
  • wordList specifies a list of key words that, if matched would trigger exclusion
  • matchType can be either PARTIAL or FULL - to be a full match, the entire finding must match the regex pattern or word exactly, whereas findings containing more than just the regex pattern or word are considered partial matches. Example: Suppose we have a finding of "4242-4242" with exclusion regex of 4242. If you use PARTIAL, this finding will be excluded from results, while using FULL will not exclude this finding, since the regex only partially matches the finding.

Additional Configuration

You can add additional fields to your config file to ignore tokens and files from being flagged, as well as specify which files to exclusively scan.

Token Exclusion

To ignore specific tokens from being flagged globally, you can add the tokenExclusionList field to your nightfalldlp config. The tokenExclusionList is a list of strings, and it mutes findings that match any of the given regex patterns.

Here's an example use case:

  "tokenExclusionList": ["NF-gGpblN9cXW2ENcDBapUNaw3bPZMgcABs", "^127\\."]

In the example above, findings with the API token NF-gGpblN9cXW2ENcDBapUNaw3bPZMgcABs as well as local IP addresses starting with 127. would not be reported. For more information on how we match tokens, take a look at regexp.

File Inclusion & Exclusion

To omit files from being scanned, you can add the fileExclusionList field to your nightfalldlp config. In addition, to only scan specific files, add the fileInclusionList to the config.

Here's an example use case:

  "fileExclusionList": ["*/tests/*"],
  "fileInclusionList": ["*.go", "*.json"]

In the example, we are ignoring all file paths with a tests subdirectory, and only scanning on go and json files. Note: we are using gobwas/glob to match file path patterns. Unlike the token regex matching, file paths must be completely matched by the given pattern. e.g. If tests is a subdirectory, it will not be matched by tests/*, which is only a partial match.

Annotation Levels:

Annotations can be configured to be notice, warning, or failure, by setting the annotationLevel key in the configuration object. The check will only fail if failure annotations are written.

For example:

  "annotationLevel": "warning"

Versioning

The Nightfall DLP Action issues releases using semantic versioning.

Support

For help, please email us at [email protected].