osec.one case analysis cover: Roku 2024, two credential stuffing waves and 591,000 accounts

Roku wasn't hacked in 2024. Roku's statement says there was "no indication that Roku was the source of the account credentials" and no sign its systems were breached. Attackers took email and password pairs leaked by other services and tried them on Roku's sign-in page, millions of times, until hundreds of thousands worked. The first wave took over 15,363 accounts between 28 December 2023 and 21 February 2024. The second, disclosed on 12 April 2024, took over about 576,000. That's roughly 591,000 accounts in all, and fewer than 400 of them were used to buy things with the card the owner had saved.

Roku's answer was to turn on two-factor verification for every one of its more than 80 million active accounts, including the ones nobody had touched. This write-up walks through both waves step by step: the tools used, why per-IP limits and a CAPTCHA didn't stop them, the arithmetic on how many login attempts it takes to get 576,000 hits, and the control that breaks each step on a sign-in page of your own.

Six steps: leaked combo lists; OpenBullet configs with rotating proxies; valid logins; email, password and shipping address changed; stored card used for subscriptions and hardware; accounts resold from 50 cents
Six steps from a leaked password list to a soundbar shipped to someone else's address. Roku's own systems weren't breached.

The two waves in numbers

Wave 1Wave 2
Accounts taken over15,363about 576,000
When the attackers had access28 Dec 2023 to 21 Feb 2024 (56 days, Maine AG filing)Not published
Disclosed8 Mar 2024 (customer notices)12 Apr 2024 (Roku blog)
How Roku found itNot stated"Ongoing security monitoring" after wave 1
Fraudulent purchasesHardware and streaming subscriptions, shown off by buyers on Telegram"Less than 400" cases, refunded
Full card numbers takenNoNo
Two-factor on the web accountNot offeredTurned on for all 80M+ accounts

Combined: 15,363 + 576,000 = 591,363 accounts, about 0.74% of Roku's 80 million active accounts (591,363 / 80,000,000). Small as a share, large as a number.

How the attack worked, step by step

1. A list of passwords from somewhere else

Credential stuffing starts with "combo lists": email and password pairs from older breaches and infostealer logs, traded and resold in bulk. Nothing on the list came from Roku. It works because people reuse passwords: if your Roku password is the one that leaked from a forum in 2019, the attacker already has it.

2. Off-the-shelf tools with a Roku config

BleepingComputer found the first wave was run with OpenBullet 2 and SilverBullet, free automation tools that replay a site's sign-in request for every line of a combo list. Users share "configs" per target site. A researcher told BleepingComputer the Roku config had been in use for months, and that it got past Roku's brute-force limits and CAPTCHA by sending requests to specific URLs and rotating proxies, so each IP address made only a few attempts.

3. A valid login, then the owner is locked out

On a hit, the attackers changed the account's password, email address and shipping address. Changing the email does two things at once: the owner can no longer reset the password, and order confirmation emails stop reaching them. Many owners found out from a charge on their card.

4. The saved card pays

Roku accounts often keep a card on file for subscriptions. Attackers used it for streaming subscriptions and Roku hardware: streaming sticks, soundbars, remotes, light strips and cameras. Buyers posted redacted order confirmations on Telegram as proof. Roku counted fewer than 400 fraudulent purchases in wave 2, so most taken-over accounts were never used to buy anything. The resale market is the bigger worry.

5. Accounts sold for as little as 50 cents

BleepingComputer saw hijacked Roku accounts sold for as little as $0.50 each, one listing offering 439 accounts. The buyer gets a working login with a stored card and a delivery address they control.

The arithmetic: how many attempts make 576,000 hits

Roku didn't publish attempt counts, and there is no solid public number for stuffing success rates; industry estimates usually run from about 0.1% to 2% of attempts. Here is the second wave at three rates:

Assumed success rateAttempts for 576,000 hitsAttempts for both waves (591,363)
2%28.8 million29.6 million
1%57.6 million59.1 million
0.1%576 million591 million

Now spread wave 1 over its 56 known days. At a 1% success rate, 15,363 hits means 1.54 million attempts: about 27,400 a day, or one every 3 seconds. Run through 5,000 rotating proxy IPs, each IP makes about 5.5 attempts a day. A per-IP limit of 5 attempts a minute, a common default, allows 7,200 a day (5 × 1,440 minutes) and never fires: the busiest IP stays about 1,300 times below it.

That's why per-IP rate limits and a CAPTCHA on the login form weren't enough. The signal is in other counts: the share of failed logins across the whole site, how many distinct accounts one network or one browser fingerprint tries, and sign-ins to accounts that have never been used from that region. For scale: Akamai counted 193 billion credential stuffing attempts across its customers in 2020, and Okta found that 34% of Auth0 sign-in traffic in 2022 was credential stuffing.

Roku's fix: two-factor for everyone, by email link

Roku's 2FA isn't an authenticator app or an SMS code. The next time anyone signs in to a Roku account on the web, Roku emails a verification link to the account's address, and the sign-in goes through only after the link is clicked. People who can't reach that mailbox can fall back to entering the last five digits of a device ID linked to the account.

What it fixes and what it doesn't:

• Fixes: a correct password alone no longer opens the account. A stuffing tool would also need the owner's mailbox.
• Doesn't fix: accounts where the attacker had already changed the email. On those, the link goes to the attacker, and recovery depends on proving who owned the original address.
• Weak spot: the device ID fallback is a short number on hardware in someone's living room. Reviewers called it too basic.
• Moves the risk: the account is now only as safe as the email account. If someone reused their Roku password on their email too, the link protects nothing.

Controls that break each step

StepControl on your own sign-inEffect here
Reused password triedCheck new and existing passwords against known breached ones (the Pwned Passwords range API takes only the first 5 characters of a SHA-1 hash); force a reset on a matchRemoves most of the combo list's hits before they happen
Millions of attempts through proxiesCount failures per account, per device fingerprint and site-wide, not only per IP; slow down or challenge when the site's failure rate jumpsCatches the low-and-slow pattern a per-IP limit misses
Login from a new deviceStep-up on new devices: email code or link, passkey, or app approval (what Roku did, months later)A correct password isn't enough
Email changedConfirm at the old address with a link that undoes the change; hold the change 24-72 hours; no password reset to the new address in that windowThe owner can take the account back
Shipping address changed, then a purchaseRe-authenticate before payment or address changes; a new address plus a hardware order within minutes goes to reviewStops the stored-card purchase
Account resoldSign out every session after a password or email change; alert when one device uses many accountsResold logins stop working

The cheapest control is not to have a password at all. A sign-in with only a one-time email code, or Google or Apple, gives a combo list nothing to try. osec.one's passwordless sign-in does email codes, Google and Apple with no OAuth setup on your side, and it's free. If you keep passwords, sign-in rules that count emails and wrong codes, not just IPs cover the second row of the table.

How it differs from 23andMe

Both started with reused passwords. In 23andMe's 2023 breach, about 14,000 logins led to data on 6.9 million people, because each account could see its DNA relatives: the damage came from a sharing feature. At Roku, each login was one account, and the damage was money: a stored card and a shipping address field. Different payoffs, same first step, and the same fix for that step: a correct password should never be the only thing between an attacker and an account.

Sign-ins with email codes, Google and Apple: no password to stuff, no OAuth app to set up on your side. Free.

Add passwordless sign-in

Sources

• BleepingComputer, Over 15,000 hacked Roku accounts sold for 50¢ each to buy hardware (11 Mar 2024): https://www.bleepingcomputer.com/news/security/over-15-000-hacked-roku-accounts-sold-for-50-each-to-buy-hardware/
• Maine Attorney General, Roku breach notification (15,363 accounts, 28 Dec 2023 to 21 Feb 2024): https://www.maine.gov/agviewer/content/ag/985235c7-cb95-4be2-8792-a1252b4f8318/e9cc298b-379b-47ba-a10d-e2263963b574.shtml
• Roku, Protecting your Roku account (12 Apr 2024): https://www.roku.com/blog/protecting-your-roku-account
• CBS News, Roku says 576,000 accounts compromised (12 Apr 2024): https://www.cbsnews.com/news/roku-576000-accounts-compromised-recent-security-breach/
• Security Affairs, Roku second data breach (12 Apr 2024): https://securityaffairs.com/161765/data-breach/roku-second-data-breach.html
• Digital Trends, Roku makes 2FA mandatory after breach: https://www.digitaltrends.com/home-theater/roku-account-breach-2fa/
• Axios, Roku 576,000 accounts (12 Apr 2024): https://www.axios.com/2024/04/12/roku-576000-accounts-security-breach
• Akamai, State of the Internet 2020 (193 billion credential stuffing attempts), reported by Dark Reading
• Okta, State of Secure Identity 2022 (34% of Auth0 traffic was credential stuffing)
• Have I Been Pwned, Pwned Passwords range API: https://haveibeenpwned.com/API/v3#PwnedPasswords