Redaction

Utilities for implementing common redaction logic for the Meticulous Recorder Middleware API.

Utilities for implementing redaction middleware

Example Usage

import { dropRequestHeader, transformJsonResponse, redactRecursively, asterixOut } from "@alwaysmeticulous/redaction";

const middleware = [
  dropRequestHeader("Authorization"),
  transformJsonResponse({
    urlRegExp: /https:\/\/api\.example\.com\/.*/,
    transform: (data) => {
      return { ...data, ...(data.sensitive ? { sensitive: asterixOut(data.sensitive) } : {}) };
    },
  }),
  transformJsonResponse({
    urlRegExp: /https:\/\/api\.example\.com\/sensitive.*/,
    transform: (data) => redactRecursively(data, {
      redactString: (str, path) => path[path.length - 1] === "uuid" ? str : asterixOut(str),
    }),
  }),
];

Utilties for redacting javascript objects

To make it easier to redact large complex javascript types we provide a number of helpers.

asterixOut

This ensures the redacted text is the same length and has the same whitespace as the original text, thereby allowing you to test the same layout cases.

import { asterixOut } from "@alwaysmeticulous/redaction";

const redacted = asterixOut("some sensitive text"); // returns "**** ********* ****"

redactString

Redact string intelligently redacts the string by checking for common data formats. This reduces the risk of your app error'ing during the replay of Meticulous tests due to it failing to operate on the redacted data (e.g. "*******" is not a valid URL):

import { redactString } from "@alwaysmeticulous/redaction";

const redacted = redactString("some sensitive text"); // returns "**** ********* ****"
const redacted2 = redactString("[email protected]"); // returns "[email protected]"
// etc.

See redactString.spec.ts for more examples.

NestedFieldsRedactor

NestedFieldsRedactor allows you to specify a redaction policy for each distinct field name (for example 'ssn' or 'email'). It'll then recursively apply this redaction policy across all nested fields inside an object. Type safety ensures that compile time errors will be produced if you provide an object that has a field name (nested at any level inside the object) that you have not specified a redaction policy for.

Basic Usage

import { transformJsonResponse, NestedFieldsRedactor, redactString } from "@alwaysmeticulous/redaction";

interface MyComplexApiType {
  details: {
    ssn: string;
    phoneNumbers: Array<{ mobile?: string; home?: string }>;
  };

  // By default we do not require redacting boolean fields.
  isUSBased: boolean;
}

// Important: include your API type explictly in the call to createRedactor (`createRedactor<MyComplexApiType>`)
const complexApiTypeRedactor = NestedFieldsRedactor.builder().createRedactor<MyComplexApiType>({
  strings: {
    ssn: redactString,
    mobile: redactString,
    home: redactString,
  },
});

const middleware = [
  transformJsonResponse({
    urlRegExp: /https:\/\/api\.example\.com\/.*/,
    transform: (data: MyComplexApiType) => {
      return complexApiTypeRedactor(data);
    },
  }),
];

If you update MyComplexApiType to add a new name string field:

interface MyComplexApiType {
  name: string; // <<< ADDED

  details: {
    ssn: string;
    phoneNumbers: Array<{ mobile?: string; home?: string }>;
  };

  isUSBased: boolean;
}

But don't add a corresponding redaction policy for the new name field to the 'createRedactor' call, then your code will fail to compile. This ensures that the redaction is exhaustive.

createRedactor will force you to redact all non-enum string fields, but it won't force exhaustive redaction of other types of fields (dates, booleans, numbers, etc.). See createRedactorStrict if you need to enforce exhaustive redaction at compile time of additional data types.

With Defaults

We recommend however using NestedFieldsRedactor.builderWithDefaults(), which will provide default redactors for most common field names. If there are any string fields not covered by those defaults then the compiler will force you to specify a redaction policy for them:

const complexApiTypeRedactor = NestedFieldsRedactor.builderWithDefaults().createRedactor<MyComplexApiType>({
  strings: {
    // Don't need to specify a redaction policy for `ssn` as it's covered by the defaults,
    // but we do need to specify a redaction policy for `mobile` and `home` as they're not covered by the defaults.
    mobile: redactString,
    home: redactString,
  },
});

Pattern Based Redactors

You can also specify redactors that match field names that end with a given postfix, while preserving compile-time type safety. See common-redactors.ts and redact-nested-fields.ts for some examples.

redactRecursively

Recursively iterates through a JSON object applying the provided redaction function. See redact-recursively.spec.ts for more details.

This can be combined with NestedFieldsRedactor to provide extra safety. For example:

const complexApiTypeRedactor = NestedFieldsRedactor.builder().createRedactor<MyComplexApiType>({
  strings: {
    ssn: redactString,
    mobile: redactString,
    home: redactString,
  },
});

const redactAnythingThatLooksLikeAnSSN = <T>(data: T) => redactRecursively(
    data,
    {
      redactString: (str) => looksLikeAnSSN(str) ? asterixOut(str) : str,
    }
  );

const middleware = [
  transformJsonResponse({
    urlRegExp: /https:\/\/api\.example\.com\/.*/,
    transform: (data: MyComplexApiType) => {
      return redactAnythingThatLooksLikeAnSSN(complexApiTypeRedactor(data));
    },
  }),
];