November 19, 2026

The Message Came From a Colleague at a Partner Firm. It Was Russian Intelligence: How One Cloud Breach Fueled the Next

Google’s threat researchers describe how the Russian espionage group Cold Relic, also known as Star Blizzard, used data stolen from one U.S. organization to impersonate its employee, phish Microsoft Entra ID credentials at a second organization and plant its own MFA device. Here is the story, and the layers that break the chain.

A man in a navy sweater at an office desk with a laptop and a coffee cup, looking thoughtful

Here Is How Layered Defenses Help When Attackers Borrow the Identity of Someone You Trust

Most phishing advice assumes the dangerous message comes from a stranger. This week, Google’s threat intelligence team and its Mandiant responders described a case where it came from someone the target thought they knew. The account below is based on Google’s published case study, with additional reporting from SecurityWeek and The Record.

The first break-in

The group behind it is known to Google as Cold Relic and to others as Star Blizzard, Callisto, Gossamer Bear and formerly ColdRiver. The U.S. Justice Department has assessed that it operates within Russia’s FSB intelligence service, and Google has tracked it since 2015 as it went after government officials, defense contractors and nonprofits in Europe and North America.

In late 2025 the group got into the cloud productivity environment of a U.S. organization. It had the stolen password of a top-level “super admin” account, and it got past multi-factor authentication by flooding the user with push prompts until one was accepted. To avoid location-based alarms, it connected through a residential proxy near the victim’s own area. It then registered a device the tenant had never seen and, over time, added passkeys, MFA backup codes and other rogue MFA methods so it could keep getting back in.

With admin rights, it took nearly everything: the organization’s entire cloud drive, through mass sharing to an attacker-controlled account and direct downloads, full employee lists and a tenant-wide data export. It checked the security and email audit logs to see what defenders could see, and edited and deleted security alerts to erase traces of the intrusion. It stayed for months, and passkeys it had enrolled kept working even after password and session resets. In at least one case Google describes, an organization reset passwords but did not revoke existing sessions, so the attacker simply kept its access. Even after it was finally removed, it kept emailing the victim’s staff and contacts from accounts impersonating employees.

The second break-in

In 2026, a second U.S. organization was approached on an encrypted messaging app by someone posing as an employee of the first. Google says the group likely drew on the contact and relationship details stolen earlier; one of its phishing links, seen in June 2026, even carried a reference to the first victim. This second organization used Microsoft Entra ID as the single sign-on provider for its Google Workspace. After a period of conversation, the impostor shared links that redirected to documents fitting the discussion while capturing the target’s Entra ID credentials and authentication tokens. Once in, the group registered a new MFA device in the sign-on system to hold on to the account and bulk-downloaded the files in the user’s cloud drive. Afterwards it renamed its messaging profile to a new persona, which Google believes was meant for the next target.

Google says the group looks specifically for credentials, personnel records, email archives, rosters, contact lists, resumes and event schedules, the raw material for the next convincing approach. In some cases, passwords stored in cloud documents let it reach additional accounts, including employees’ personal ones. Google says it has seen new phishing activity from the group as recently as this month. Neither organization was named.

The layered defenses

  • Layer 1: phishing-resistant MFA. Google’s top recommendation is to replace push, SMS and voice codes with FIDO2/WebAuthn hardware security keys, which are bound to the real site and cannot be relayed by a fake page or approved out of fatigue. Start with administrators.
  • Layer 2: lock down who can add an MFA method. Require an existing administrator or an out-of-band identity check before a new authenticator, passkey or device is registered, and alert on enrollments right after a burst of MFA prompts. In Microsoft Entra ID, for example, Conditional Access can govern registration of security information; other identity platforms have comparable controls.
  • Layer 3: shrink and watch the admin accounts. Keep very few super admins, use them only for admin work, and restrict admin sign-ins from consumer ISPs, VPNs and residential proxy networks. Alert on directory exports, tenant-wide data exports, audit-log searches and deleted security alerts.
  • Layer 4: verify trusted contacts out of band. A familiar name in a messaging app is not proof of identity. Before opening a shared document link or signing in after a chat, confirm through a channel you already have, such as a known phone number.
  • Layer 5: watch the data, not just the door. Data loss prevention rules on cloud storage such as OneDrive, SharePoint and Google Drive can flag files containing passwords or staff rosters and limit mass external sharing. Alert on sudden spikes in downloads or external shares.
  • Layer 6: remediate completely. Password resets alone did not evict this attacker. Google advises revoking all sessions and refresh tokens, removing unknown passkeys, backup codes and MFA devices, and reviewing third-party OAuth app grants in both Google Workspace and Microsoft 365.

What this means for your company

Your business does not have to be an espionage target to face this pattern; criminals use the same trick in invoice fraud, writing from a real supplier’s hijacked mailbox. That means a breach at a partner can become your problem. Move your admins and finance staff to phishing-resistant sign-in first, and make sure only a small, named group can approve new MFA devices. Teach staff that a known name in a chat or email is not enough when a link, login prompt or payment change is involved, and agree a call-back habit with key partners. Finally, write down what “fully cleaned up” means for an account compromise, including session revocation and MFA method review, so a password change is never mistaken for the finish line.

If you want to know how your sign-in settings, admin accounts and partner communications would hold up against this kind of impersonation, we can review them with you. Reach out to Layer 7 Data Solutions to discuss how we’ve helped companies similar to yours.

About The Author