nodejs-common

This library provides various utilities for parsing Adscore signatures, generating custom request payloads, and virtually anything that might be useful for customers doing server-side integration with the service.

Compatibility

Supported Signature v5 algorithms

1. v5_0200H - OpenSSL CBC, HTTP query
2. v5_0200S - OpenSSL CBC, PHP serialize
3. v5_0201H - OpenSSL GCM, HTTP query
4. v5_0201S - OpenSSL GCM, PHP serialize
5. v5_0101H - sodium secretbox, HTTP query
6. v5_0101S - sodium secretbox, PHP serialize
7. v5_0200J - OpenSSL CBC, JSON
8. v5_0201J - OpenSSL GCM, JSON
9. v5_0101J - sodium secretbox, JSON
10. v5_0101M - sodium secretbox, msgpack
11. v5_0200M - OpenSSL CBC, msgpack
12. v5_0201M - OpenSSL GCM, msgpack

Not supported Signature v5 algorithms

1. v5_0101I - sodium secretbox, igbinary
2. v5_0200I - OpenSSL CBC, igbinary
3. v5_0201I - OpenSSL GCM, igbinary

Install

Nodejs version >= 14.18.0 is required

Via NPM

npm i @adscore/nodejs-common

Optional dependencies:

Use npm i php-serialize package for following algorithms:

1. v5_0200S - OpenSSL CBC, PHP serialize
2. v5_0201S - OpenSSL GCM, PHP serialize
3. v5_0101S - sodium secretbox, PHP serialize

Use npm i sodium package for following algorithms:

1. v5_0101H - sodium secretbox, HTTP query
2. v5_0101S - sodium secretbox, PHP serialize
3. v5_0101M - sodium secretbox, msgpack
4. v5_0101J - sodium secretbox, JSON

Usage

V4 signature decryption

When zone's "Response signature algorithm" is set to "Hashing" or "Signing", it means that V4 signatures are in use. They provide basic means to check incoming traffic for being organic and valuable, but do not carry any additional information.

import { Judge } from "@adscore/nodejs-common/definition";
import { ParseError } from "@adscore/nodejs-common/errors";
import {
  Signature4,
  VerifyError,
  VersionError,
} from "@adscore/nodejs-common/signature";

/*  Replace <key> with "Zone Response Key" which you might find in "Zone Encryption" page for given zone.
    Those keys are base64-encoded and the library expects raw binary, so we need to decode it now. */
const cryptKey = Buffer.from("<key>", "base64");

/*  Three things are necessary to verify the signature - at least one IP address, User Agent string
        and the signature itself. */
const signature = request.body.signature; /* for example */
const userAgent = request.get("User-Agent") ?? "";
/*  You might want to use X-Forwarded-For or other IP-forwarding headers coming from for example load
        balancing services, but make sure you trust them and they are not vulnerable to user modification! */
const ipAddresses = [request.socket.remoteAddress];

try {
  const parser = Signature4.createFromRequest(
    signature,
    ipAddresses,
    userAgent,
    cryptKey
  );

  /*  Result contains numerical result value */
  const result = parser.getResult();
  /*  Judge is the module evaluating final result in the form of single score. RESULTS constant
            contains array with human-readable descriptions of every numerical result, if needed. */
  const humanReadable = Judge.RESULTS[result];
  console.log(humanReadable["verdict"] + " (" + humanReadable["name"] + ")");
} catch (e: unknown) {
  if (e instanceof VersionError) {
    /*  It means that the signature is not the V4 one, check your zone settings and ensure the signatures
            are coming from the chosen zone. */
  }

  if (e instanceof ParseError) {
    /*  It means that the signature metadata is malformed and cannot be parsed, or contains invalid data,
            check for corruption underway. */
  }

  if (e instanceof VerifyError) {
    /*  Signature could not be verified - usually this is a matter of IP / user agent mismatch (or spoofing).
            They must be bit-exact, so even excessive whitespace or casing change can trigger the problem. */
  }
}

V5 signature decryption

V5 is in fact an encrypted payload containing various metadata about the traffic. Its decryption does not rely on IP address nor User Agent string, so it is immune for environment changes usually preventing V4 to be even decoded. Judge result is also included in the payload, but client doing the integration can make its own decision basing on the metadata accompanying.

The format supports a wide variety of encryption and serialization methods, some of them are included in this repository, but it can be extended to fulfill specific needs.

It can be integrated in V4-compatible mode, not making use of any V5 features (see V4 verification):

import {
  Signature5,
  SignatureParseError,
  VerifyError,
  VersionError,
} from "@adscore/nodejs-common/signature";
import { Judge } from "@adscore/nodejs-common/definition";
import { CryptParseError } from "@adscore/nodejs-common";
// only when using sodium based algorithms
import * as sodium from "libsodium-wrappers";

const cryptKey: Buffer = Buffer.from("<key>", "base64");
const signature: string = request.body.signature;
const userAgent: string = request.get("User-Agent") ?? "";
const ipAddresses: string[] = [request.socket.remoteAddress];

await sodium.ready; // only when using sodium based algorithms

try {
  const parser = Signature5.createFromRequest(
    signature,
    ipAddresses,
    userAgent,
    cryptKey
  );
  const result = parser.getResult();
  const humanReadable = Judge.RESULTS[result];
  const humanReadableResult =
    humanReadable["verdict"] + " (" + humanReadable["name"] + ")";

  console.log(humanReadableResult);
} catch (e) {
  if (e instanceof VersionError) {
    /*  It means that the signature is not the V5 one, check your zone settings and ensure the signatures
        are coming from the chosen zone. */
  }

  if (e instanceof CryptParseError || e instanceof SignatureParseError) {
    /*  It means that the signature metadata is malformed and cannot be parsed, or contains invalid data,
        check for corruption underway. */
  }

  if (e instanceof VerifyError) {
    /*  Signature could not be verified - see error message for details. */
  }
}

The first difference is that now $cryptKey may be also a Closure instance (lambda function), accepting single int argument - zone ID and returning raw key as binary string. This is useful in scenarios, where signatures coming from different zones are handled at a single point. This is not possible for V4 signatures, as they do not carry over any zone information.

As we can see, createFromRequest also requires a list of IP addresses and User Agent string. This is used for built-in verification routine, but this time the verification is completely unrelated to decryption. Client integrating might want to replace the verification with its own implementation, so here is the extended example (without any exception handling for readability):

import { Signature5 } from "@adscore/nodejs-common/signature";
import * as sodium from "libsodium-wrappers";

await sodium.ready; // only when using sodium based algorithms

const signature: string = request.body.signature;
/*  An example structure holding keys for every zone supported */
const cryptKeys: Record<number, Buffer> = {
  123: Buffer.from("123456789abcdefghijklmn", "base64"),
};
const parser = new Signature5();
/*  Parsing/decryption stage */
parser.parse(signature, (zoneId: number | bigint) => {
  // Cast to number as bigint cannot be index type in TS yet, see: https://github.com/microsoft/TypeScript/issues/46395
  if (!cryptKeys[zoneId as number]) {
    throw new Error("Unsupported zone " + zoneId);
  }

  return cryptKeys[zoneId as number];
});
/*  The payload now contains a decrypted signature data which might be used to verify the signature */
const payload = parser.getPayload();
/*  We can still make use of built-in signature validator and only then getResult() is being populated */
const userAgent = request.get("User-Agent") ?? "";
const ipAddresses = [request.socket.remoteAddress];
parser.verify(ipAddresses, userAgent);
const result = parser.getResult();

The result field and its associated getResult() getter method return result score only after a successful verify() call. This is expected behavior, to preserve compliance with V4 behavior - the result is only valid when it's proven belonging to a visitor. For custom integrations not relying on built-in verification routines (usually more tolerant), the result is present also in payload retrieved via getPayload() call, but it's then the integrator's reponsibility to ensure whether it's trusted or not. When desired validation is more strict than the built-in one, the verify() can be called first, populating getResult() value, and after that any additional verification may take place.

Note: V4 signature parser also holds the payload, but it does not contain any useful informations, only timestamps and signed strings; especially - it does not contain any Judge result value, it is derived from the signature via several hashing/verification approaches.

Integration

Any questions you have with custom integration, please contact our [email protected]. Please remember that we do require adequate technical knowledge in order to be able to help with the integration; there are other integration methods which do not require any, or require very little programming.