> ## Content Index
> Fetch the complete content index at: https://f4n6.co.uk/llms.txt
> Use this file to discover other available public pages before exploring further.

# Cold Relic leverages stolen cloud data for targeted espionage
- URL: https://f4n6.co.uk/security-feed/cold-relic-leverages-stolen-cloud-data-for-targeted-espionage/
- Published: 2026-11-18T23:08:34.000Z
- Updated: 2026-11-18T23:08:34.000Z
- Author: Jeff Davies
- Tags: #security-feed, Cold Relic, ColdRiver, FSB

## 1\. Executive summary

Mandiant, Google's threat intelligence unit, has published an investigation into two linked intrusions by the Russia-linked espionage group Cold Relic (also tracked as COLDRIVER, UNC4057, Callisto, Gossamer Bear and Star Blizzard). In a late-2025 compromise at a US organization the group took an administrator account, downloaded the entire cloud drive and employee directories, and stayed resident for several months; in a 2026 intrusion at a second US organization it used that stolen contact and relationship context to impersonate an employee over an encrypted messaging app and phish Microsoft credentials and authentication tokens. The US Department of Justice has assessed that Cold Relic operates within Russia's Federal Security Service (FSB). No CVE, CVSS score or CISA KEV entry is associated with this campaign, and no EMEA financial services victim is named. The risk to clients is identity and cloud-data exposure through social engineering that borrows a trusted colleague's identity, not exploitation of a software flaw.

## 2\. Regulatory framing

No specific DORA/NIS2 article is directly engaged by this item.

## 3\. Technical analysis & attack chain

The campaign is a chain of two intrusions in which material stolen in the first is used to make the second credible. Mandiant investigated both. Neither victim is named, and neither is described as a financial services firm.

Confirmed steps, first intrusion (late 2025, US organization):

1. Cold Relic took control of an administrator account at the organization.
2. From that account it downloaded the organization's entire cloud drive and its employee directories.
3. It searched for emails, passwords and personnel records, which Mandiant assesses was likely to fulfil broader intelligence requirements (names, email addresses and email archives).
4. It remained inside the organization's systems for several months, continuing to read email through third-party applications.
5. After the organization changed passwords and reset user sessions, the group found ways to regain access.
6. After losing network access, it kept targeting the organization and its contacts by sending emails that appeared to come from employees.

Confirmed steps, second intrusion (2026, second US organization):

1. Using information stolen in the first breach, Cold Relic posed as an employee and approached a person at the second organization through an encrypted messaging app.
2. It opened a conversation before sending malicious links built to steal Microsoft account login details and authentication tokens.
3. The links led to documents related to the conversation, so the phishing matched the pretext.
4. The group gained access to an account at the second organization and downloaded large amounts of data from that user's cloud storage.

The delivery technique is the part a defender can act on. Cold Relic uses a compromised website as the first link, which redirects the target to the real phishing domain. That hides the phishing domain and lets the group fingerprint each visitor before deciding whether to redirect or serve malicious content. Phishing links usually carry a parameter at the end whose value the attacker sets, often indicating a relationship to a previous victim or the lure. Mandiant's example format is `https://compromised.example.com/some_path/some_file.php?file=<previous_victim_name>`. The parameter value is the tell: a link whose query string names a person, a prior contact or a document topic is consistent with this tradecraft.

Mandiant also describes the group's use of localized residential proxies and cross-cloud identity abuse as part of its post-compromise tradecraft. The public summary does not give proxy infrastructure, IP ranges, domains, file hashes or the specific third-party applications used to keep reading email. Those details are not stated in public reporting.

Attribution: Mandiant and Google Threat Intelligence Group attribute the activity to Cold Relic and link the group to Russia's FSB via a US Department of Justice assessment. The Record separately reports that US authorities have linked the group to the FSB. MITRE ATT&CK records Star Blizzard as G1033; Cold Relic, COLDRIVER, UNC4057, Callisto and Gossamer Bear are names the sources give for the same group. The FSB itself has no MITRE ATT&CK profile, so the state link rests on the DOJ assessment as reported, not on an independent technical attribution in this material.

Google says it has tracked the group since 2015 and observed new phishing operations as recently as November 2026\. The reporting is single-sourced to Google's Mandiant and GTIG; The Record and The Hacker News both draw on the same Google report rather than adding independent telemetry.

## 4\. Mitigation & containment

No software vulnerability is involved, so there is no patch. The controls that matter are identity, session and link-handling controls around cloud productivity suites.

P1 (within 24 hours)

- Hunt for cloud administrator accounts that downloaded an entire drive or the full employee directory in one session, and for mailbox access through third-party OAuth applications. Revoke OAuth grants to third-party apps that staff do not recognise, since the group kept reading email through third-party applications after the first compromise.
- After any credential reset, verify that sessions and tokens are actually invalidated. The group regained access after the first victim changed passwords and reset user sessions, so a password reset alone is not containment. Revoke refresh tokens and re-register affected devices.
- Block or alert on inbound links whose query string carries a person's name, a prior contact or a document topic, matching the `?file=<previous_victim_name>` pattern Mandiant describes. Treat unsolicited links arriving over encrypted messaging apps as phishing until the sender is verified out of band.

P2 (within 72 hours)

- Require phishing-resistant authentication (FIDO2 security keys or passkeys) for cloud administrator accounts and for staff who handle sensitive correspondence, so a stolen password plus token cannot be replayed.
- Restrict cloud drive sharing and bulk-download permissions. Alert on a single account downloading the full drive or the full directory.
- Add monitoring for lookalike or compromised third-party sites used as redirectors: flag links that land on a domain unrelated to the one in the message, and log the redirect chain.

P3 (within 7 days)

- Brief staff who deal with external contacts that an approach from a known colleague over a messaging app can be an impersonation built from stolen correspondence. Verification should use a channel the sender did not initiate.
- Review joiner, leaver and directory-sync processes so that a stolen staff directory has a shorter useful life, and limit what a compromised administrator account can export in one action.

No fixed release, patch identifier or vendor mitigation specific to this campaign is named in public reporting.

## 5\. Indicators of compromise

No atomic indicators of compromise (hashes, domains, IP addresses, file paths) are given in the public reporting. The sources describe behaviours and a URL pattern rather than machine-pivotable values. The example URL in Mandiant's report uses placeholder hostnames and is not a live indicator.

Behavioural indicators:

| behaviour                                                                 | where to observe                                         | confidence                                                     |
| ------------------------------------------------------------------------- | -------------------------------------------------------- | -------------------------------------------------------------- |
| Cloud administrator account downloads entire drive and employee directory | Cloud audit logs (drive export, directory export events) | High, stated by Mandiant                                       |
| Mailbox read through third-party OAuth applications                       | OAuth grant logs, mailbox access logs                    | High, stated by Mandiant                                       |
| Access regained after password change and session reset                   | Authentication logs, token refresh events                | High, stated by Mandiant                                       |
| Phishing link carries a query parameter naming a prior victim or lure     | Web proxy, mail gateway, messaging-app link logs         | High, pattern described by Mandiant                            |
| Link to a compromised site redirects to an unrelated phishing domain      | Web proxy redirect chains, DNS logs                      | High, stated by Mandiant                                       |
| Approach over an encrypted messaging app impersonating a known employee   | User reports, endpoint messaging-app telemetry           | Medium, described in The Record's summary of the Google report |

## 6\. Detection

The sources give a URL parameter pattern but no file artefacts, command lines, registry keys or process names, so a YARA rule would have nothing genuine to match. Detection here is network and identity behaviour.

```yaml
title: Cold Relic style phishing redirect with victim-named query parameter
id: 8f2c1a4e-cold-relic-redirect
status: experimental
description: Detects HTTP requests to a redirector whose query string carries a file parameter naming a person or prior contact, consistent with COLD RELIC phishing tradecraft described by Mandiant.
author: Adverse Trace
date: 2026/11/18
references:

  - https://cloud.google.com/blog/topics/threat-intelligence/cold-relic-a-case-study-in-impersonation-and-compromise/
logsource:
  category: proxy
detection:
  selection_path:
    cs-uri-query|contains: 'file='
  selection_redirect:
    cs-uri-stem|endswith:

      - '.php'
  filter_benign:
    cs-host|endswith:

      - '.example.com'
  condition: selection_path and selection_redirect and not filter_benign
fields:

  - c-ip
  - cs-host
  - cs-uri-stem
  - cs-uri-query
falsepositives:

  - Legitimate file-download endpoints that pass a filename in a file parameter
level: medium

```

```yaml
title: Bulk cloud drive and directory export by administrator account
id: 3b7d9e21-cold-relic-bulk-export
status: experimental
description: Detects a cloud administrator account exporting the full drive or employee directory, the collection step Mandiant attributes to COLD RELIC in the late-2025 intrusion.
author: Adverse Trace
date: 2026/11/18
references:

  - https://cloud.google.com/blog/topics/threat-intelligence/cold-relic-a-case-study-in-impersonation-and-compromise/
logsource:
  product: google_workspace
  service: admin
detection:
  selection:
    event.name:

      - 'drive.export'
      - 'directory.users.download'
      - 'admin.drive.download'
  condition: selection
fields:

  - actor.email
  - event.name
  - ipAddress
falsepositives:

  - Scheduled compliance or backup exports by IT
level: high

```

The proxy rule keys on the `file=` parameter and `.php` stem that Mandiant's example shows; tune the benign-host filter to your own estate. The cloud rule uses event names that vary by tenant, so map them to your provider's audit schema before enabling.

## Threat actor context

**Star Blizzard** · [G1033](https://attack.mitre.org/groups/G1033?ref=f4n6.co.uk) · aka SEABORGIUM, Callisto Group, TA446, COLDRIVER

[Star Blizzard](https://attack.mitre.org/groups/G1033?ref=f4n6.co.uk) is a cyber espionage and influence group originating in Russia that has been active since at least 2019\. [Star Blizzard](https://attack.mitre.org/groups/G1033?ref=f4n6.co.uk) campaigns align closely with Russian state interests and have included persistent phishing and credential theft against academic, defense, government, NGO, and think tank organizations in NATO countries, particularly the US and the UK.

*No MITRE ATT&CK profile for: Cold Relic, FSB.*

## 7\. Sources

- Google Threat Intelligence / Mandiant, "COLD RELIC: A Case Study in Impersonation and Compromise", https://cloud.google.com/blog/topics/threat-intelligence/cold-relic-a-case-study-in-impersonation-and-compromise/
- The Record (Recorded Future News), Daryna Antoniuk, "Russian cyber-espionage group used stolen cloud data to hack into organizations, Google says", https://therecord.media/russia-cyber-espionage-google, 2026-11-18
- The Hacker News, "Russian COLD RELIC Hackers Use Stolen Google Workspace Data to Breach Second U.S. Organization", https://thehackernews.com/2026/11/russian-cold-relic-hackers-use-stolen.html

## 8\. Adverse Trace position

Adverse Trace assesses this as a targeted espionage campaign with no CVE, no CVSS score and no CISA KEV entry, so it does not carry a vulnerability severity rating; the exposure is identity and cloud-data theft through impersonation. No EMEA financial services victim is named, and the two victims are described only as US organizations, so direct client impact is not established. The relevance to financial services is that the same cloud productivity suites, administrator accounts and third-party OAuth applications are in wide use across the sector, and a stolen staff directory plus correspondence is enough to make a follow-on approach convincing. The reporting is single-sourced to Google's Mandiant and GTIG; the FSB link rests on a US DOJ assessment as reported, not on an independent technical attribution in this material, and should be treated as the sources' attribution rather than a confirmed fact. We will monitor for Mandiant or Google publishing infrastructure, domains or hashes from these intrusions, and will issue a follow-up if atomic indicators become available. Clients who want to test exposure now should run the P1 hunts in §4 against cloud audit and OAuth logs.

---

[Read the original source →](https://therecord.media/russia-cyber-espionage-google?ref=f4n6.co.uk)

*Published via PulseTrace — Adverse Trace threat intelligence.*