---
url: "https://www.xcademia.com/news/google-details-how-cold-relic-turns-cloud-compromises-into-follow-on-espionage-attacks"
title: "Google Details How COLD RELIC Turns Cloud Compromises into Follow-On Espionage Attacks"
description: "Google and Mandiant detail how COLD RELIC uses stolen identities, cloud data, and phishing to compromise organisations and maintain access."
publishedAt: "2026-11-18T09:02:33.169+00:00"
updatedAt: "2026-11-22T10:31:50.70741+00:00"
type: news
category: "cloud-security"
source_name: Google Cloud Blog
source_url: "https://cloud.google.com/blog/topics/threat-intelligence/cold-relic-a-case-study-in-impersonation-and-compromise"
tags:
  - "#COLDRELIC"
  - "#Cybersecurity"
  - "#ThreatIntelligence"
  - "#CloudSecurity"
  - "#IdentitySecurity"
  - "#Phishing"
  - "#GoogleWorkspace"
  - "#Mandiant"
---

# Google Details How COLD RELIC Turns Cloud Compromises into Follow-On Espionage Attacks

> Google and Mandiant detail how Russian-linked COLD RELIC used stolen credentials, cloud data, and employee impersonation to compromise multiple organisations. The report highlights cross-cloud identity abuse, persistent MFA access, and targeted social engineering.

Source: **Google Cloud Blog** · 18 November 2026

Google Threat Intelligence Group (GTIG) and Mandiant have published an investigation into **COLD RELIC**, a Russian state-sponsored cyber espionage group that uses compromised cloud environments and stolen organisational information to conduct follow-on attacks against other targets.

The investigation connects two intrusions at separate US-based organisations. During a late-2025 incident, the group compromised a cloud productivity environment and collected large volumes of organisational data. In a subsequent 2026 intrusion, investigators found evidence that COLD RELIC used information and identities obtained during the earlier compromise to impersonate an employee and approach people at another organisation.

The group is also tracked under the names **COLDRIVER and UNC4057**, and is known as **Star Blizzard, Callisto, and Gossamer Bear**.

According to Google, the US Department of Justice has assessed that COLD RELIC operates within Russia's Federal Security Service (FSB). The group has targeted individuals and organisations with information of strategic or political interest, including government officials, defence contractors, and non-governmental organisations across Europe and North America.

GTIG has tracked and disrupted COLD RELIC phishing campaigns since 2015. The latest investigation provides additional detail about its social engineering, cloud identity abuse, data collection, and persistence techniques.

Google said it observed evidence of new COLD RELIC phishing operations as recently as November 2026. The group remains active.

## 
How One Cloud Breach Became a Route to Another Organisation

The investigation describes a pattern in which information stolen during one intrusion becomes the foundation for a later attack.

During the late-2025 compromise, COLD RELIC downloaded substantial volumes of data from a US-based organisation. The group searched for names, email addresses, and email archives, likely to meet broader intelligence requirements.

In 2026, Mandiant investigated another intrusion at a separate US-based organisation. Phishing URL parameters and the impersonation pretexts used by the attackers indicated that COLD RELIC likely drew on contact information and professional relationships obtained during the earlier incident.

Investigators also found that the group initiated contact through an encrypted messaging application while impersonating an employee of the first victim organisation.

Separately, GTIG connected a redirector URL from the later intrusion to a URL submitted to VirusTotal in June 2026. Its query parameter referenced the first victim organisation.

This connection illustrates how data stolen from one organisation can support a subsequent intrusion without requiring attackers to rely on generic phishing messages. Information about employees, their roles, contacts, and ongoing professional relationships can make an impersonation attempt appear credible.

### 
The role of compromised websites in phishing

COLD RELIC frequently uses compromised websites as the first link in a phishing chain.

Instead of directing a victim straight to the final phishing domain, the attacker sends a link hosted on a compromised website. That website redirects selected visitors to the actual phishing destination.

According to Google, this approach provides several advantages:

- It conceals the final phishing domain behind a legitimate but compromised website.
- It gives the attacker an opportunity to fingerprint visitors before redirecting them.
- It allows the attacker to decide whether a visitor should be sent to the phishing page.
- It can incorporate URL parameters associated with a previous victim or the subject of a phishing lure.

The report provides examples of URL parameters containing a previous victim's name or a lure topic:

```
https://compromised.example.com/some_path/some_file.php?file=

https://compromised.example.com/some_path/some_file.php?ref=
```

These parameters can provide investigators with clues about the relationship between a phishing campaign and an earlier victim. Google notes that this URL pattern is common in the group's campaigns, but not universal.

![info-1](https://0a515t3ure77wbvx.public.blob.vercel-storage.com/articles/1794991751022-info-1--277-.webp)

## Case Study 1: Stolen Super Admin Credentials Led to Extensive Cloud Data Theft

During the late-2025 intrusion, COLD RELIC gained access to the cloud productivity environment of a US-based organisation.

GTIG observed the group using stolen credentials belonging to a **Super Admin** account. The attackers connected through a US residential proxy network located geographically close to the victim.

The login activity also involved an MFA push notification fatigue attack. The victim accepted an MFA request, allowing the attackers to authenticate using the stolen credentials.

After gaining access, COLD RELIC registered a device model that had not previously appeared in the organisation's tenant. The group then used passkeys, MFA backup codes, and unauthorised MFA devices to maintain access.

### 
Mass collection through cloud storage and administrative tools

The compromised Super Admin account provided access to the organisation's cloud environment.

COLD RELIC collected files from the organisation's cloud storage drive using two methods:

- Sharing large volumes of files with an email account operated by the attackers.
- Downloading files directly from the cloud storage environment.

The group also accessed the Google Workspace Admin Console and examined security investigation and email audit logs to understand the organisation's defensive visibility.

Its activity went beyond data collection. GTIG observed the attackers viewing, modifying, and deleting Alert Center notifications to remove evidence of the initial intrusion.

The attackers also exported employee lists and initiated a domain-wide data export through **Google Workspace Customer Takeout**.

Customer Takeout is an administrative tool for exporting data across an organisation's tenant. It differs from the individual Google Takeout service available to end users.

Google states that Customer Takeout requests are recorded in Workspace Admin audit logs under the event name:

`CUSTOMER_TAKEOUT_CREATED`

Administrators can also disable Google Takeout for users at the organisational level.

### 
Persistence continued after password and session resets

The initial data theft did not end the intrusion.

GTIG observed COLD RELIC maintaining access to the organisation for several months after the initial compromise. During this period, the group:

- Synchronized email inboxes through OAuth with external email clients.
- Reauthenticated using attacker-enrolled passkeys that had not been revoked.
- Continued accessing cloud documents related to credentials, email archives, personnel information, and other intelligence requirements.

The passkeys were particularly significant because they persisted across password and session resets and bypassed device registration controls, according to the report.

After the threat actor was removed from the organisation's infrastructure, GTIG observed renewed social engineering activity. COLD RELIC used email accounts impersonating employees of the compromised organisation to contact internal staff and people in the same professional circles.

This meant the stolen organisational identities remained useful even after the original cloud access had been removed.

## 
Case Study 2: COLD RELIC Used an Employee's Identity to Approach Another Organisation

The 2026 intrusion demonstrated how COLD RELIC combined social engineering with identity infrastructure spanning multiple cloud platforms.

GTIG observed the group compromise a Microsoft Entra ID user account at a second US-based organisation. That organisation used Entra ID as its federated single sign-on (SSO) identity provider for Google Workspace.

The attackers first established contact through a secure messaging application. They impersonated an employee of the organisation compromised during the earlier incident.

The conversation provided a believable context for sharing phishing links. Rather than directing the victim to an unrelated login page, the links redirected the user to document lures relevant to the ongoing discussion.

The phishing pages were designed to steal Microsoft Entra ID credentials and authentication tokens.

After obtaining the credentials, COLD RELIC authenticated to the account and registered a new MFA device in the SSO solution. The device helped the attackers retain access throughout the intrusion.

GTIG said the group ultimately obtained access to the organisation and performed a bulk download of files available in the compromised user's cloud drive.

## 
Why the impersonation approach mattered

The attacker did not approach the second organisation as an unfamiliar sender. It used the identity of someone connected to the target's professional network.

That context made the conversation more credible and allowed the phishing links to appear relevant to an existing discussion.

Google included a redacted and abbreviated depiction of the social engineering exchange in its report. The image was generated with large language models and represents the observed engagement rather than an original screenshot of the full conversation.

<text color="secondary" size="xs">Source image: Google Cloud, Figure 2, redacted LLM-generated depiction of the social engineering engagement.</text>

The investigation shows why organisations should not assess suspicious messages solely by checking whether the sender's name is familiar. An attacker using a compromised identity or a convincing impersonation can exploit existing trust between employees and professional contacts.

## 
How COLD RELIC Maintains Access After Compromising an Account

Across the two cases, GTIG identified a recurring sequence of post-compromise activity.

The group first establishes persistent access through rogue MFA devices, passkeys, and MFA backup codes. It then collects information from email, chat, and cloud storage services.

Its primary objective is broad collection and downloading of files and email archives. However, the group also searches for specific information that could support its intelligence requirements or enable further intrusions.

The report identifies several categories of interest:

Information targeted

How it can support the operation

Credentials and other secrets

Can provide access to additional accounts or services

Personnel information

Helps identify employees and relevant individuals

Email archives, including `.pst` files

Provide historical communications and relationship context

Contact lists and organisational rosters

Support employee impersonation and follow-on targeting

Resumes and event schedules

Provide information that can make social engineering more convincing

Operational documents

Help identify information relevant to intelligence collection

GTIG also observed a possible change of identity after the second compromise. The encrypted messaging application used for the initial approach recorded that the attacker changed the profile name to a completely different person.

Google believes the new persona was likely intended to support further targeting of organisations in the same professional circles. The report presents this as an assessment, not a confirmed account of every subsequent operation.

![info-2](https://0a515t3ure77wbvx.public.blob.vercel-storage.com/articles/1794991829739-info-2--259-.webp)

## Cross-Cloud Identity Abuse Creates a Wider Security Challenge

The second case involved Microsoft Entra ID and Google Workspace working together through federated SSO.

This is important because a compromise affecting one identity provider can have consequences for services that rely on it for authentication.

In the investigation, COLD RELIC obtained credentials and authentication tokens for a Microsoft Entra ID account. The victim organisation used Entra ID as its federated identity provider for Google Workspace.

The group then registered a new MFA device in the SSO solution and accessed files available through the compromised account.

The incident demonstrates that organisations need visibility across their connected identity environments. Reviewing Google Workspace activity alone may not provide the complete picture when authentication is federated through another provider.

The same principle applies to incident response. Revoking access in one service does not necessarily invalidate every active session, token, or authentication method associated with the identity.

GTIG's recommendations therefore cover Google Workspace and Microsoft Entra ID, including active browser sessions, refresh tokens, third-party OAuth permissions, passkeys, backup codes, and MFA devices.

## 
Google Recommends Stronger Identity Controls and Session Revocation

GTIG's remediation guidance focuses on identity security, session management, network visibility, and data protection.

### 
1. Replace vulnerable MFA methods with phishing-resistant authentication

Google recommends enforcing phishing-resistant MFA based on **FIDO2 and WebAuthn**.

The guidance recommends replacing SMS, voice calls, and push notifications with hardware security keys, such as Titan Security Keys.

The report explains that hardware security keys are bound to the origin URL. This helps protect authentication from adversary-in-the-middle (AiTM) reverse proxies and MFA fatigue attacks.

Organisations should also:

- Monitor passkey registrations, particularly on administrative accounts.
- Investigate unfamiliar device enrolments.
- Require existing administrators or out-of-band identity verification to approve new MFA devices and secondary authenticators.
- Review the use of MFA backup codes.
- Consider Google's Advanced Protection Program.

The emphasis is not simply on requiring MFA. Organisations also need to control how new authentication methods are registered and who can approve them.

### 
2. Revoke sessions and OAuth access during incident response

One of the report's central findings concerns the limitations of password resets.

GTIG described a case in which an organisation attempted to remove COLD RELIC by resetting passwords for affected accounts. However, the attackers retained access because existing sessions had not been revoked.

Google recommends that organisations responding to suspected account compromise revoke:

- Active browser sessions.
- Refresh tokens.
- Authorised third-party OAuth application permissions.
- Unauthorised passkeys.
- MFA backup codes.
- Rogue MFA devices.

These actions should be included in Business Email Compromise (BEC) response procedures.

Organisations should also regularly audit third-party OAuth grants across Google Workspace and Microsoft 365. Unexpected email clients and external mail services, including Thunderbird and Proton Mail, should receive additional scrutiny when they are not approved for organisational use.

### 
3. Improve visibility into residential proxy activity

COLD RELIC used residential proxy networks with IP addresses geographically close to its victims.

This approach can make malicious logins appear consistent with the expected location of the legitimate user. As a result, traditional impossible-travel detections may not identify the activity.

GTIG recommends enriching authentication logs with residential proxy and VPN IP reputation feeds.

Organisations should also:

- Alert on sudden residential autonomous system number (ASN) changes involving privileged accounts.
- Apply conditional access policies to block or detect logins from known proxy networks.
- Review authentication activity from consumer internet service providers and unfamiliar networks.
- Investigate suspicious access even when the geographic location appears plausible.

The report identifies **AS5650** and **AS20115** as examples of residential proxy ASN ranges to consider in network monitoring.

### 
4. Prevent sensitive information from being stored and shared in cloud storage

GTIG found that credentials and other secrets stored in organisational cloud storage increased the potential impact of the intrusions.

The information gave COLD RELIC opportunities to access additional accounts and services, including personal accounts that employers might not monitor.

Google recommends using cloud storage Data Loss Prevention (DLP) policies to detect sensitive information.

For example, organisations can configure rules to warn or block users attempting to create, upload, or externally share files matching sensitive patterns.

The guidance includes Google Workspace Security and Data Protection as one example. It also recommends cross-cloud DLP policies covering platforms such as Google Drive, OneDrive, and SharePoint.

Rules can focus on sensitive terms associated with passwords, credentials, personnel rosters, and email archives.

## 
Detection Opportunities: What Security Teams Should Monitor

The report provides detection guidance mapped to MITRE ATT&CK techniques. These recommendations focus on identifying suspicious authentication, data collection, persistence, and attempts to interfere with security monitoring.

The following table summarises the activities and detection opportunities identified by GTIG.

Observed activity

MITRE ATT&CK

Detection opportunity

MFA fatigue attack

T1621: Multi-Factor Authentication Request Generation

Alert on repeated failed logins or MFA challenges for one user within a short period, followed by successful authentication from the same IP address

OAuth abuse and mail synchronization

T1114.002: Email Collection - Remote Email Collection; T1528: Steal Application Access Token

Monitor new OAuth authorisations granting broad mail scopes, such as `https://mail.google.com/`, to unapproved clients

Automated data export

T1119: Automated Collection

Alert on tenant-wide exports, including `CUSTOMER_TAKEOUT_CREATED` events in Google Workspace Admin logs

Bulk cloud file downloads

T1530: Data from Cloud Storage

Detect unusual increases in downloads by one user, especially from non-corporate IP addresses or residential proxy networks

External file sharing

T1567.002: Exfiltration to Cloud Storage

Monitor unusual increases in external sharing, including changes to `shared_externally` visibility or sharing with free email providers

New device and passkey enrolment

T1098.005: Account Manipulation - Device Registration

Alert on unexpected passkey enrolments and device registrations, particularly following MFA fatigue activity

MFA backup code use

T1078.004: Valid Accounts - Cloud Accounts

Investigate backup-code authentication from unfamiliar networks or residential proxy infrastructure

Alert deletion and modification

T1562.008: Impair Defenses - Disable or Modify Cloud Logs / Alerts

Monitor changes to or deletion of Alert Center notifications, particularly alerts associated with the same account

Employee directory export

T1087.004: Account Discovery - Cloud Account

Review user-list downloads from unfamiliar devices or unexpected administrative sessions

Audit log investigation

T1082: System Information Discovery

Investigate unexpected queries against security, access, and audit logs that could indicate counter-forensic activity

Searches for credentials and archives

T1083: File and Directory Discovery; T1552.001: Credentials In Files

Monitor searches for terms such as `.pst`, `.bak`, `password`, and `cred`, along with repeated access to credential-related documents

## 
Google SecOps includes dedicated detection rules

Google Security Operations customers can access relevant detection content through the **Mandiant Frontline Threats rule pack**.

GTIG identifies the following Google SecOps rule names for activity covered in the report:

- Google Workspace Successful MFA Fatigue Attack
- Google Workspace Customer Takeout Initiated
- Google Workspace Alert Center Alert Deletion
- Google Workspace Directory User List Download
- Google Workspace Drive Bulk External or Freemail File Sharing
- Google Workspace Device or Passkey Enrollment Following MFA Fatigue
- Google Workspace OAuth Broad Mail Scope Granted to Unapproved Client

These rules cover several stages of the activity, including suspicious authentication, data export, external sharing, device enrolment, and attempts to remove security alerts.

Organisations should review the relevant rules and ensure their logging and alerting configurations provide visibility into the identity and cloud services they use.

![info-3](https://0a515t3ure77wbvx.public.blob.vercel-storage.com/articles/1794991859487-info-3--235-.webp)

## Why COLD RELIC's Campaigns Matter to Organisations Using Cloud Services

The cases described by Google and Mandiant show how a cloud compromise can extend beyond the original victim.

COLD RELIC collected information that could serve two purposes: intelligence gathering and preparation for later social engineering. Employee directories, email archives, contact lists, and operational documents helped the group understand relationships and construct more convincing approaches to other people.

The investigation also highlights the difficulty of responding to identity compromises that involve multiple authentication methods. Password resets alone did not remove the attackers' access in the case described by GTIG. Unrevoked sessions and attacker-enrolled passkeys allowed the group to reauthenticate.

For organisations using federated identity across cloud platforms, the findings reinforce the need to coordinate monitoring and incident response across identity providers and connected services.

The report also highlights a limitation in location-based detection. Residential proxies selected near a victim's location can make suspicious logins appear less unusual. Authentication monitoring therefore needs to consider network reputation, device changes, and the sequence of events surrounding a login, rather than relying on geographic anomalies alone.

COLD RELIC remains active, with GTIG reporting new phishing activity as recently as November 2026. The group's use of stolen organisational context demonstrates why cloud security must extend beyond protecting credentials. Organisations also need to protect the data, identities, sessions, and professional relationships that attackers may exploit after an initial compromise.

## Original source

https://cloud.google.com/blog/topics/threat-intelligence/cold-relic-a-case-study-in-impersonation-and-compromise

## Tags

`#COLDRELIC` · `#Cybersecurity` · `#ThreatIntelligence` · `#CloudSecurity` · `#IdentitySecurity` · `#Phishing` · `#GoogleWorkspace` · `#Mandiant`

---

## About this content

This Markdown news article is the citation-grade twin of [Google Details How COLD RELIC Turns Cloud Compromises into Follow-On Espionage Attacks](https://www.xcademia.com/news/google-details-how-cold-relic-turns-cloud-compromises-into-follow-on-espionage-attacks). It is published by **Xcademia** (UK Companies House 12322710) and is available for AI search engines and large language models to index, summarise, and cite.

When citing or quoting, please attribute *Xcademia* and link back to the source URL above.

- Source: https://www.xcademia.com/news/google-details-how-cold-relic-turns-cloud-compromises-into-follow-on-espionage-attacks
- Publisher: Xcademia, https://www.xcademia.com
- Catalogue index: https://www.xcademia.com/llms-full.txt
