"Once Cold Relic compromises a target, it uses stolen emails and data to establish authentic pretexts and impersonate contacts when reaching out to subsequent targets," Google said.
Cold Relic: aliases, targets, and operational pattern
The actor Microsoft and security vendors track as Cold Relic—formerly COLDRIVER and UNC4057 and also tracked as Callisto, Gossamer Bear, and Star Blizzard—has a documented history of targeting high-value Western organizations. According to Google's Mandiant and Threat Intelligence Group (GTIG), Cold Relic has successfully compromised government officials, defense contractors, and non-governmental organizations across Europe and North America by combining phishing and malware with targeted identity theft to acquire credentials and long-term access.
From a single Google Workspace compromise to chained intrusions
Google investigators described a playbook that begins with one cloud compromise and escalates into a campaign of linked intrusions. In late 2025 the adversary obtained corporate staff directories, identities, and other data from a U.S.-based organization. Those harvested assets were then used in 2026 to mount a subsequent breach at a different U.S.-based organization. The second intrusion used the legitimacy of stolen emails and contact lists to craft pretexts and impersonate trusted colleagues, enabling the actor to move laterally into the victim's professional networks.

This site is the portfolio.
OSINTSights runs on Cloudflare Workers, D1, R2, and Vectorize, with an AI pipeline on Hetzner ARM. Nubivance designed, built, and operates it. We do the same for clients.
See what we buildTechniques: MFA fatigue, passkeys, and administrative abuse
Google's teams reported a sequence of technical steps that turned tenant-level control into mass data theft. Initial access was secured via a residential proxy network positioned near the victim. The attackers used stolen Super Admin credentials combined with a push-notification MFA fatigue attack until a user accepted a request. After access, the threat actor registered their own devices in the tenant and employed passkeys and MFA backup codes as alternate access mechanisms.
Once inside, Cold Relic exercised Super Admin privileges to download the organization's entire cloud storage by mass-sharing files to attacker-controlled email accounts and by exporting data through Google Takeout. The intruders inspected the Admin Console, queried Admin security investigation and email audit logs to assess defensive visibility, and viewed, modified, and deleted Alert Center notifications to remove forensic evidence.
Persistence techniques observed included syncing inboxes via OAuth to external mail clients and re-authenticating using unrevoked attacker-enrolled passkeys that persisted across password and session resets and bypassed device registration controls. The actors also targeted cloud drive documents tied to credentials, email archives, and personnel records.
Second breach via Microsoft Entra ID SSO and re-weaponized trust
The campaign shifted from Google Workspace to Microsoft Entra ID when the second U.S. organization used Entra ID as its federated single sign-on (SSO) provider for Google Workspace. There, Cold Relic impersonated an individual from the first compromised organization, initiated contact over a secure messaging app, and sent phishing links under the pretext of sharing a document about a technical issue. Those links harvested Entra ID credentials and authentication tokens.
After successful authentication, the actor registered a new MFA device in the SSO system to retain access, then performed bulk downloads of files accessible through the compromised user's cloud drive. Google noted that the messaging account used to initiate the operation had its profile name changed to a different persona, likely to facilitate targeting other organizations in the same professional networks. Even after being expelled from the initial tenant, the group continued to impersonate employees of the first victim to target internal staff and associated contacts in renewed social-engineering campaigns.
What this means for security teams, policymakers, and affected enterprises
- Security teams and technologists: Expect adversaries to combine human-centric social engineering with identity and cloud-account abuse. The techniques observed—MFA push fatigue, attacker-enrolled passkeys that survive password resets, OAuth mailbox syncs, Google Takeout mass exports, Admin Console log queries, and Alert Center manipulation—are all part of the recorded chain of actions used to maintain persistence and exfiltrate data.
- Policymakers and regulators: The investigation directly ties the campaign's utility to intelligence objectives, with Google stating that "Russian intelligence services turn legitimate business communications into attack vectors against trusted partners." That framing highlights cross-border intelligence exploitation of commercial cloud platforms and third-party trust relationships as a regulatory consideration.
- Affected enterprises and procurement leaders: Federated identity and SSO configurations can transform a compromise at one partner into a breach at another. The case shows how employee rosters, operational documents, and genuine correspondence can be harvested and re-used to increase success rates of phishing against downstream organizations.
Google's Mandiant and GTIG investigators depict Cold Relic's operations as a methodical conversion of a single cloud compromise into a platform for chained intrusions aimed at networks of trusted partners. The record here is concrete: residential-proxy initial access, Super Admin credential theft, MFA fatigue and passkey persistence, administrative log manipulation, Google Takeout exports, and Entra ID federated-account exploitation. The sequence leaves one clear operational question unanswered by the report itself—how many downstream targets remain unknowingly exposed to resurrected personas and roving messaging accounts that began life inside a single compromised tenant?




