A Russian-speaking hacking group has been harvesting login credentials for U.S. banks, IRS, and Social Security accounts directly from infected Windows computers—surprisingly bypassing both password managers and multi-factor authentication (MFA). The group intercepts credentials saved on users’ own devices rather than breaching the banks or government agencies themselves.
Security researchers discovered that the threat actors collected data from over 600 victims including at least 57 with bank or government account logins—among them 46 who had credentials tied to critical U.S. platforms like ID.me, Login.gov, Social Security, and the IRS.
Malware planted via phishing dropped a small Python script on victims’ machines, which then communicated with the attackers’ servers. After profiling the victim’s stored data, the attackers deployed tools specifically to target browser-based password stores. They extracted encrypted password vaults, saved browser cookies, and decrypted them using embedded keys stored within the malware. In one case, 3,345 credentials were decrypted from just 16 victims.
Erosion of Cookie & Vault Trust
The attackers used stolen session cookies to bypass MFA—once the cookie is accepted by a site, the account appears logged in even without reentry of credentials or additional verification. They also decrypted whole vaults: one LastPass user lost access to over 2,800 saved logins including for major financial accounts like Chase, Wells Fargo, IRS.gov, Medicare and PayPal. Some users of 1Password had their master passwords successfully guessed, and a developer lost files stored in a KeePass database.
Targets included major U.S. banks: Chase showed up in 31 cases, Capital One in 20, Citi in 17, Wells Fargo in 15. Accounts tied through ID.me and Login.gov were especially alarming given how these platforms can grant access to government services.
Tools, Infrastructure & Wider Impact
Besides banking and government credentials, the group also stole login data from 67 victims covering 44 cryptocurrency platforms like Binance, Coinbase, MetaMask, and Kraken. Tokens for AI tools—such as Claude, Codex, and Gemini—were also lifted, exposing potentially sensitive project data.
To retain access, attackers installed legitimate remote-management utilities such as Level.io and ScreenConnect—tools which can evade detection because they’re genuine software. Their operations were masked behind Cloudflare Workers, which acted as fronts for servers named HavoX, Phantom53, and Kraken Bridge. Over a period of about three weeks, the malware infrastructure logged more than 1.2 million check-ins from infected machines.
Investigators traced clues in the attackers’ own files—Russian-language strings found across almost 250 scripts and logs—suggesting a Russian-speaking operation. The data trove was discovered via an open directory in the Netherlands, where logs, scripts, and data exfiltrated from over 600 compromised Windows PCs had been stored.
Victims are advised to act immediately: use a clean device to reset passwords—starting with email and financial accounts—and sign out of all active sessions. Resetting passwords alone is not enough, since valid session cookies may still grant access. Security teams should flag suspicious Python scripts running from AppData\Roaming, restrict or monitor remote-management tools, bind sessions to trusted devices, and keep session durations short.
Indicators of compromise include exposed IPs and domains tied to the operation, including ports used by HavoX, Phantom53, various Cloudflare Worker frontends, and file paths to malware implants.