Skip to content
Security & Trust

Google: data from one cloud breach was used to phish a second organization

Google's threat team says Cold Relic used one US organization's stolen records to pose as a trusted colleague.

W
WebPulse Newsroom
AI-assisted · 4 min read
Share on X LinkedIn
Google: data from one cloud breach was used to phish a second organization
In brief
  • Google Threat Intelligence Group says Cold Relic used data stolen from one US organization's cloud to impersonate an employee and phish a second organization.
  • A breach can reach partners, because stolen staff lists and email make a fake message look real.
  • Google advises phishing-resistant MFA, audited passkey registrations, short-lived sessions and full session revocation after a suspected compromise.

A stolen employee list is more than a privacy problem. In patient hands, it is a set of believable voices. Google Threat Intelligence Group (GTIG) describes a Russian espionage group that used one organization's cloud data to speak, in that organization's name, to another.

What GTIG says happened

GTIG tracked a group called Cold Relic across two compromises of US organizations over the past year. Google has followed the group since 2015. It also goes by the names Star Blizzard, Callisto and Gossamer Bear. Dark Reading reports that the US government has placed it inside Russia's Federal Security Service.

The first break-in came in late 2025. The attackers held stolen credentials for a super admin account. That is the account with the widest rights in a cloud environment.

They routed their traffic through a residential proxy near the victim. The logins therefore looked local. Then they kept sending the real account holder multifactor authentication (MFA) push prompts. Eventually the user approved one.

Once inside, the group registered its own device. It also added new ways to sign in: passkeys, backup MFA codes and extra rogue MFA devices. GTIG says it copied cloud storage holding full employee lists, email threads, directories and contact lists. It then took steps to hide its tracks and kept access for several months.

Earlier this year, the group put that material to use. It posed as an employee of the first organization and contacted a person at a second organization over an encrypted messaging app. It sent links disguised as documents linked to their conversation.

The links captured the target's Microsoft Entra ID credentials and authentication tokens. With those, the attackers entered the organization's Google Workspace through federated single sign-on. That setup lets one identity open several systems. They took files from the user's cloud drive and searched for email archives, chat records, credentials and personnel data.

The idea: your records vouch for the attacker

We usually picture a breach ending at the victim's door. This case has a longer reach. The stolen material was not only data to read or sell. It showed who knows whom and what they are working on.

GTIG's researchers wrote that the campaigns "demonstrate the convergence of human-centric social engineering and cloud identity abuse." In plain terms, a cloud account breach can supply the detail that makes a scam message believable.

The lesson for leaders is that a breach can lend your reputation to the attacker. Partners, suppliers and clients trust messages that carry your people's names and project details. Those partners carry the risk, and they cannot see your logs.

How it works, in plain terms

Two weak points stand out. The first was the push prompt. If one tap approves a login, a persistent attacker needs only one tired approval.

The second was registration. After entering, the group added its own devices, passkeys and backup codes. A password reset alone would not remove devices and passkeys the attacker added. That is our inference, not a finding from GTIG.

In the second case, the entry point was a phishing link from a fake colleague. Federation widened the damage. One stolen Entra ID sign-in opened a separate Google Workspace. Convenience has a cost: a single theft can open several doors. That reading is ours, not GTIG's.

The sources do not say how the first credentials were stolen. They do not say how many people were affected. They do not say whether other organizations were targeted. Those questions remain open.

What to ask your team

Joseph Spero, a principal incident response consultant at Google, lists several controls. They are MFA that resists phishing, restricting and auditing passkey registrations, hardening MFA registration, and audits of OAuth grants. An OAuth grant is a permission that lets one app act for a user.

He also points to short-lived sessions and least privilege. Both limit what a stolen credential can reach.

If compromise is suspected, Spero says to mandate session revocation in addition to resetting passwords and MFA mechanisms. A reset alone does not end live sessions.

Wes Shields, a GTIG security engineer, adds advice for staff. Be wary of unexpected messages, especially on channels a colleague does not normally use. Confirm the sender through a separate, known channel first.

Leaders can turn this into five questions for their teams.

Can any administrator approve a login with one tap? Who is alerted when a new passkey, backup code or MFA device is added to a privileged account? When did we last audit OAuth grants? After an incident, do we revoke sessions as well as reset passwords? Can our partners check a message that claims to come from us?

Your staff directory and inbox history map your relationships. Protect them as you would protect the relationships themselves.

Produced by the WebPulse Newsroom with AI assistance from the original reporting credited below, and checked against that source by our editorial review. How we use AI.
Original reporting: Dark Reading.