Picture an employee who installs a cracked copy of a design program on a Friday evening, or follows the instructions on a page that claims to be checking whether a visitor is a robot. Minutes later a compressed bundle of text files leaves the machine: every password the browser had saved, the autofill entries, the cookies that keep that person signed in to the company's email and cloud consoles, and a list of the software installed. Criminals call the bundle a log, and it is the raw material for a large share of the intrusions that follow.
The trade is industrial in scale. Flashpoint counted 7.4 million infected hosts and devices across the first six months of 2026, 27% more than in the previous half year, and put the number of credentials harvested from them at 1.7 billion. Vidar, StealC and Lumma were the most prolific families by infected devices. Microsoft alone identified more than 394,000 Windows computers infected with Lumma between March 16 and May 16, 2025.
What a log holds, and how it is taken
Microsoft's dissection of Lumma gives a good inventory. The malware pulls stored logins, autofill data and session cookies out of Chromium and Mozilla based browsers, Edge included. It hunts for the local keys, wallet files and browser extensions of cryptocurrency wallets such as Exodus, Electrum and MetaMask. It also lifts VPN and email client configurations, documents from the user profile, and enough system detail for a buyer to know what kind of machine the log came from.
No flaw on the victim's own machine has to be exploited for any of this. Microsoft lists the routes it watched Lumma take: poisoned search ads for software downloads and browser updates, pirated applications bundled with the stealer, phishing emails, hacked websites with injected JavaScript, and fake CAPTCHA pages. That last trick, known as ClickFix, tells the visitor to prove they are human by pasting a command into the Windows Run box; the command fetches the stealer and runs it in memory. One Claude user whose account was hijacked in August 2026 traced the infection to a pirated game picked up on a Russian underground forum.
From a log to a login
Lumma is rented out as a service through Telegram and Russian language forums, and its affiliates manage the stolen data from a panel. Logs are then traded in bulk. Flashpoint describes automated pipelines that ingest fresh logs, pick out the valuable entries and immediately run credential stuffing and live session tests against thousands of services at once. The buyer does not break in. With a valid password or, better still, a valid cookie, they sign in to VPNs, cloud instances, webmail and business applications as if they were the employee.
The cookie is the part that defeats two factor authentication. A second factor protects the login itself; once the site has issued a session cookie, the browser presents it on every request and is not challenged again, so an attacker who replays a stolen cookie is simply an already signed in user. That is what happened to Claude accounts in late August 2026. Anthropic told affected users that the malware seen in the campaign included RedLine, Acreed, StealC, Lumma and Vidar on Windows machines and Atomic Stealer on a handful of Macs, and that it had signed the accounts out to kill the stolen sessions, removed saved payment methods and refunded unauthorized charges. Microsoft has seen ransomware crews including Octo Tempest use Lumma, and told a US court in June 2026 that StealC credentials are sold on to initial access brokers and then used to break into networks and deploy ransomware.
Takedowns buy time, not safety
Two of the largest law enforcement actions against the supply side came thirteen months apart. In May 2025 Microsoft used a US court order to seize or block about 2,300 domains behind Lumma, while the US Department of Justice took down its central command structure and Europol and Japan's JC3 cybercrime center suspended local infrastructure. In June 2026 the Operation Endgame coalition turned to Amadey, StealC and the SocGholish loader: Europol reported 326 servers and 142 domains disrupted, roughly 27 million stolen credentials recovered from more than 385,000 compromised systems, and more than 41 million euros in criminal crypto assets frozen. Neither action ended the trade; Lumma and its clones were back in Flashpoint's top three a year after its domains were seized. The defensive work therefore falls on employers, and three measures matter most.
- Move the accounts that matter to passkeys or hardware keys, and pair them with short session lifetimes and tokens bound to the device, because a stealer copies the cookie whatever the strength of the login.
- Give work devices a managed browser that keeps corporate passwords out of the local profile and controls extensions, and block unapproved software. Microsoft points to rules that stop executables failing prevalence, age or trusted list checks and that stop scripts from launching downloaded programs, the exact path ClickFix relies on.
- Watch for the company's own domains in leaked logs, and treat a hit as a live session problem rather than a password problem: revoke every active session for that user first, then reset the credentials.
Anthropic's instructions to its locked out users are a usable template for the order of operations. Signing out kills the stolen sessions but leaves the malware in place, so the company told people to remove the infection first, then set a new password and turn on two factor authentication for the email account tied to Claude, and only after that add a payment method back. For an organization the lesson is the same in a different order: the buyer of a log paid for the cookie, so revoking sessions has to travel with the password reset, because a new password on its own does not always end the sessions that cookie opened.