Summary
- Mandiant investigated linked intrusions involving the Russian espionage group COLD RELIC.
- Stolen relationship information helped attackers impersonate trusted contacts and pursue a second organisation.
- Cloud identity compromise can expose sensitive documents while supporting subsequent phishing against third parties.
A Russian cyber espionage group used information obtained from one organisation’s cloud environment to support an intrusion against another, according to an investigation demonstrating how stolen relationships and legitimate account access can extend an attack beyond its initial victim.
Google Threat Intelligence Group and Mandiant published the findings on 17 November 2026, documenting linked incidents involving COLD RELIC, previously tracked as COLDRIVER and UNC4057 and also known as Star Blizzard and Callisto.
The researchers examined how information about professional contacts and organisational relationships was used to construct convincing impersonation attempts. In a subsequent intrusion during 2026, attackers obtained access to cloud resources belonging to a second US organisation.
Google says the group remains active, with phishing operations observed as recently as November. Although the detailed linked cases concern US organisations, COLD RELIC has an established history of targeting people and institutions of intelligence interest across Europe and other regions.
Compromised information enables further impersonation
The Google Cloud investigation describes how the group incorporated information obtained through an earlier compromise into its subsequent social engineering activity.
Rather than relying exclusively on generic phishing messages, attackers used existing relationships and contextual details to make communications appear credible. Information about colleagues, advisers, external contacts and ongoing work can provide a foundation for impersonation without requiring attackers to invent an entirely false scenario.
Mandiant identified evidence indicating that material from a first victim had informed approaches to a second organisation. Investigators examined phishing links and impersonation pretexts associated with the later intrusion.
The researchers assessed that the attackers obtained credentials and tokens associated with the targeted cloud identity and gained access to the organisation’s cloud-stored information. Their findings included activity involving additional authentication registration and bulk downloads from cloud storage.
These observations distinguish the compromise from a phishing attempt that ends with credential collection. Once a valid account is obtained, attackers can use legitimate service functionality to review files, access shared information and identify further opportunities.
Such activity may not immediately resemble malware execution. Authentication events, newly registered devices, unusual access patterns and unexpected data transfers become important evidence of compromise.
Access can survive a simple password reset
Cloud identity incidents often involve several forms of access beyond the original password. Sessions, authentication tokens, registered devices and authorised applications may continue to provide access unless investigators identify and revoke the affected mechanisms.
Mandiant’s findings describe techniques involving multifactor authentication abuse and cloud account access. Their relevance depends on the configuration and controls protecting each affected identity.
A password change can address one route into an account while leaving other access mechanisms requiring separate investigation. Attackers who have registered an additional authentication method or obtained usable session credentials may not depend on the password that was originally stolen.
The operational consequences can extend to external organisations. A compromised mailbox or document repository may contain information about partners, contractors, customers and advisers that can support new phishing approaches.
Those third parties may not have suffered direct unauthorised access during the first incident. Nevertheless, information obtained from the victim can expose them to more credible impersonation attempts.
For organisations handling sensitive policy, research, legal or government-related work, this relationship creates a continuing security problem after the initial environment has been restored.
Incident response must consider downstream targeting
Establishing what was accessed becomes particularly important when stolen material includes correspondence or information about external relationships.
Cloud audit records may help identify downloaded files, new authentication registrations and unusual access. Their usefulness depends on the available logging, retention period and whether the attacker operated through channels that the organisation monitored.
Organisations may also need to assess whether external contacts require notification or additional verification procedures. The appropriate response depends on what was exposed and whether investigators identify credible evidence of follow-on targeting.
Phishing-resistant authentication, restrictions on unnecessary cloud access and controls around newly registered authentication methods can reduce specific attack paths. None independently eliminates the risk that stolen information will later be used to deceive another organisation.
The Google and Mandiant investigation establishes a chain of related espionage activity, although it does not identify every subsequent target or confirm that all organisations mentioned in stolen documents were attacked.
COLD RELIC’s continued activity in November indicates that the campaign has not ended with the incidents examined in the report. The investigation shows how a compromised cloud account can become both a source of sensitive information and a means of preparing further intrusion attempts.





