In this post we present a small program where we use CrossBrowserTesting to test Fp-Collect, the fingerprinting library used to collect fingerprints that can be analyzed by Fp-Scanner, a bot detection library.

CrossBrowserTesting is a service that enables to test your website from a large number of device and browser configurations. You can either run the tests manually by interacting with the remote devices with your mouse through CrossBrowserTesting website or you can use their API with Selenium or any other testing framework to automate your tests.

Since browser fingerprinting heavily relies on APIs available only on certain browsers and devices, it is frequent to encounter bugs. While we currently test the code of Fp-Collect using Chrome headless and Puppeteer, it is unfortunately not always enough to catch all the bugs that may occur on exotic devices and browsers. Thus, in this post, we present a simple example where we use CrossBrowserTesting to test Fp-Collect using different device configurations.

Testing strategy

We use a simple testing strategy; the idea is to visit the following page to verify if the fingerprint of the automated browser is properly collected.

If you want to run the following tests you will need a CrossBrowserTesting account.

First, we start by importing the Selenium web driver.

const webdriver = require("selenium-webdriver");

Then, we create a function that reads a JSON file provided by CrossBrowserTesting in order to generate different OS and browser configurations that we use for our tests. While we generate only around 30 different configurations, it is possible to generate more configurations by combining different browsers with different OS.

function getDevicesCaps() {
    const dataDevices = Array.from(require('./browsers.json'));
    // For the moment we extract only 1 configuration per element of the array
    return => {
        const result = {};
        Object.keys(info.caps).forEach((property) => {
            result[property] = info.caps[property];

        Object.keys(info.browsers[0].caps).forEach((property) => {
            result[property] = info.browsers[0].caps[property];

        Object.keys(info.resolutions[0].caps).forEach((property) => {
            result[property] = info.resolutions[0].caps[property];

        return result;

Then, we define a function whose goal is to verify the content of the div with the “fp” id. In the case where its content is an empty string, we add the device configuration to the list of configurations that bug.

const failedCaps = [];

async function getResult(driver, cap) {
    const fp = await driver.findElement('fp')).getText();
    if (typeof fp === "undefined" || fp === "") {
    return true;

Then, we can launch our tests for each of the configuration we have generated. While here we launch the tests sequentially, CrossBrowserTesting provides a way to run multiple tests in parallel.

(async () => {
    const cbtHub = "";
    const devicesCaps = getDevicesCaps(); // generate the list of configurations

    for (let deviceCap of devicesCaps) {
        deviceCap.username = username; // replace username by your username
        deviceCap.password = authkey; // replace authkey by your authentication key

        const driver = new webdriver.Builder()

        try {
            // Loads the page with the fingerprinting script
            await driver.get('');

            // Verify if the value of the fingerprint is properly collected
            await driver.wait(getResult(driver, deviceCap), 3000);
        } catch (e) {
        } finally {
            await driver.quit();

    console.log('List of configurations that failed:');

While the testing strategy is quite naive, it still helped to catch an error that occurred in the promise returned by the navigator.mediaDevices.enumerateDevices function.

One of the next steps would be to improve the testing strategy in order to not only verify that the fingerprint collect does not crash but also to verify more in details the values returned depending on the configuration.

Antoine Vastel

Head of research at Datadome.