Key facts
- The US Department of Justice has assessed that Cold Relic works inside Russia's Federal Security Service, the FSB.
- Google Threat Intelligence Group has followed and disrupted Cold Relic phishing operations starting in 2015.
- In late 2025, Cold Relic broke into the cloud productivity environment belonging to a US-based organization.
- The group maintained access for months, syncing email inboxes via OAuth to external clients and re-authenticating with attacker-enrolled passkeys that endured password and session resets.
Google Threat Intelligence Group and Mandiant reported that Cold Relic went after government officials, defense contractors and nongovernmental organizations in Europe and North America in research published Nov. 17, 2026.
Cold Relic is also tracked as COLDRIVER, UNC4057, Star Blizzard, Callisto and Gossamer Bear. Since 2015, Google Threat Intelligence Group has followed the group’s phishing operations and disrupted them, and it observed new operations as recently as November 2026. GTIG researchers said the campaigns combine human-centric social engineering with cloud identity abuse, turning legitimate business communications into attack vectors aimed at trusted partners. “COLD RELIC is a Russian state-sponsored threat group that the US Department of Justice (DOJ) has assessed operates within Russia’s Federal Security Service (FSB),” Josh Murchie et al. wrote in the Google Threat Intelligence Group blog post.
Cold Relic breached a US-based organization’s cloud productivity environment in late 2025. The intruders gained initial access using stolen Super Admin credentials and a multifactor authentication push notification fatigue attack (attackers spam push approvals until a user accepts one), then routed activity through a US residential proxy network located close to the victim to avoid impossible-travel alerts.
Cold Relic kept access for months by syncing email inboxes through OAuth to external clients and re-authenticating with attacker-enrolled passkeys that survived password and session resets. Operators enrolled a device never seen before and relied on passkeys together with multifactor authentication backup codes plus extra rogue multifactor devices, getting around device registration controls.
With Super Admin privileges, the group pulled down the whole cloud storage drive. It mass-shared files to an attacker-controlled e-mail account and downloaded files directly. The intruders duplicated complete staff rosters and launched a tenant-wide export using Google Workspace Customer Takeout, an administrative export logged as CUSTOMER_TAKEOUT_CREATED that administrators can turn off for users across the organization.
The intruders reviewed Admin security investigation and e-mail audit logs, then viewed, changed and deleted Alert Center notifications to remove forensic traces. Once access ended, Cold Relic used e-mail accounts that pretended to belong to victim employees to target internal staff and associates in the same professional circles.
During 2026, the group pretended to be an employee of the first organization and messaged an employee of a second US organization using an encrypted messaging app. The second victim relied on Microsoft Entra ID as its federated single sign-on identity provider for Google Workspace. “Cold Relic is highly adept at leveraging social networks and impersonation of key individuals,” said Wes Shields, a security engineer at GTIG.
Phishing links captured Entra ID credentials and authentication tokens, then sent victims to document lures that matched the conversation. After authentication, Cold Relic added a new multifactor authentication device to the single sign-on solution, kept access, and downloaded cloud files in bulk.
Cold Relic sends targets through compromised websites that conceal phishing infrastructure and check visitors before showing malicious content. Cold Relic customizes the links, which often mention a prior victim or lure topic and often contain an attacker-set parameter showing a relationship to a previous victim or the lure topic. In the second intrusion, phishing links arrived under the pretext of a document about a technical issue. GTIG found a Cold Relic redirector URL submitted to VirusTotal in June 2026, and its query parameter pointed to the first victim organization.
The group searches compromised environments for credentials, personnel data, .pst e-mail archives, event schedules, rosters, contact lists and resumes to support later spear-phishing. Storing credentials and secrets in cloud storage made the intrusions worse by allowing pivots to more accounts and services, including personal accounts that employers often do not monitor. Following the second compromise, the messaging app profile took on a new persona that GTIG believes was probably used to target other organizations. The researchers did not name either compromised U.S. organization or disclose what data was taken.
Google recommends enforcing phishing-resistant FIDO2/WebAuthn multifactor authentication (open standards for phishing-resistant authentication such as security keys), such as hardware security keys, and restricting and auditing passkey registrations. Titan Security Keys bind to the origin website and stop interception by adversary-in-the-middle reverse proxies and fatigue attack approval. New multifactor devices or secondary authenticators should require approval from existing administrators or out-of-band identity verification.
Google says remediation should require session and OAuth revocation as well as password resets when a compromise is suspected. Password changes should go together with immediate revocation of active sessions, refresh tokens and unauthorized third-party OAuth app grants. Short-lived sessions and least privilege limit the scope of access when credentials are compromised, according to Joseph Spero, a principal incident response consultant at Google, who also said session revocation should accompany password and multifactor authentication resets.
Google says organizations should add residential proxy and VPN reputation data to authentication logs and alert on sudden residential ASN changes for privileged accounts. Logons from known proxy networks should be blocked, and conditional access policies can block or detect those logons. Administrative access should be limited to networks other than consumer ISPs, VPNs and known residential proxy ASN ranges such as AS5650 and AS20115.
Google recommends cross-cloud Data Loss Prevention policies for cloud storage that alert on sensitive terms or block mass external file-sharing. The rules should cover Google Drive, OneDrive and SharePoint, watching for terms such as passwords, roster and .pst files. Defenders should review third-party OAuth grants and flag unexpected native desktop email clients or external mail services such as Thunderbird and Proton Mail.
The Advanced Protection Program is an option organizations and individuals may want to join. When Cold Relic targets Google accounts, GTIG sends Government-backed attack alerts. “As part of our efforts to thwart this threat actor, we issue Government-backed attack alerts when we detect COLD RELIC targeting Google accounts,” the Google Threat Intelligence Group said. GTIG plans to keep following the group, breaking down its systems and revealing its infrastructure to protect cloud users around the world.

Join the discussion