Advertisement

Russian hackers exploit cloud accounts for cascading breaches

Russian cyber espionage group Star Blizzard has used stolen cloud identities and corporate records to breach multiple US organisations, according to a new Google investigation showing how one intrusion can enable another.

Google Threat Intelligence Group and its Mandiant incident response division disclosed on 17 November that attackers exploited information stolen from a US organisation in late 2025 to target a second organisation in 2026. The operation involved impersonating an employee of the first victim and using convincing communications to obtain access to the second organisation's cloud services.

The group, which Google now tracks as Cold Relic and previously identified as Coldriver, is linked by US authorities to Russia's Federal Security Service. Its targets have included government officials, defence contractors and non-governmental organisations across North America and Europe. Google said it had detected further phishing activity during November, indicating that the operation remains active.

The first investigated intrusion began when attackers authenticated to a Google Workspace environment using credentials stolen from an account with Super Admin privileges. They routed their connection through a residential proxy geographically close to the victim, reducing the likelihood that an unusual login location would trigger suspicion.

The attackers then repeatedly prompted the account holder to approve multifactor authentication. After a request was accepted, they enrolled an unfamiliar device and established additional ways of signing in, including passkeys, backup codes and unauthorised authentication devices.

With administrative access, the operators shared cloud storage files with an attacker-controlled email account and downloaded documents directly. They exported employee directories and initiated a domain-wide Google Workspace Customer Takeout export, an administrative function capable of collecting volumes of organisational data.

Mandiant also found that the attackers examined security investigation records and email audit logs, while modifying or deleting notifications in the Google Workspace Alert Center. These measures were intended to reduce the visibility of the intrusion to defenders.

Access persisted for several months. During that period, the group synchronised email with external clients through OAuth, searched stored files for credentials and personnel information, and reused attacker-enrolled passkeys. Google observed that some of these authentication methods continued working after password and session resets.

The second case exposed how stolen information could be converted into a credible approach to another target. Investigators identified communications through an encrypted messaging application in which an attacker posed as an employee of the first compromised organisation.

The impersonator sent phishing links that sought Microsoft Entra ID credentials and authentication tokens, then redirected the recipient towards documents matching the conversation. The second organisation used Entra ID to provide federated single sign-on access to Google Workspace, allowing a stolen identity to cross the boundary between cloud platforms.

After compromising a user account, the attackers registered a multifactor authentication device and downloaded files accessible through the account's cloud drive. Google also linked a phishing redirect submitted to VirusTotal in June to a parameter referencing the earlier victim organisation.

The findings distinguish the campaign from attacks concerned solely with stealing passwords. Employee lists, correspondence, archived email and event schedules became intelligence resources for selecting and persuading subsequent victims. Google said the group also searched for documents containing credentials that could permit access to additional services, including personal accounts outside corporate monitoring.

Microsoft separately documented an expansion of Star Blizzard's phishing operations during 2026. Its September assessment identified at least 13 larger-scale campaigns since January and activity affecting more than 100 organisations, mainly in the United States and Britain. Those operations included malicious attachments using a technique Microsoft calls RedFlick to deploy the CosmicPulse backdoor.

Google advised organisations to review privileged authentication events, identify suspicious residential proxy connections and revoke unauthorised devices, sessions and passkeys following suspected compromise. It also recommended controls restricting sensitive files from being shared externally and closer scrutiny of administrative exports.

Google said defenders should inspect newly registered authentication factors rather than relying on password changes alone. The company highlighted a case in which an organisation reset affected passwords but failed to invalidate existing sessions, enabling continued attacker access. Its guidance also calls for monitoring changes to privileged accounts.

Investigators observed another detail after the second breach: the profile name on the encrypted messaging account used to contact its victim changed to an entirely different identity. Google assessed that the replacement persona was probably intended for further approaches to contacts in the same professional networks.
Previous Post Next Post

Advertisement

Advertisement

نموذج الاتصال