What happened
Replaying the Azure account hijack, early 2024
- 1
Stolen consultant tokens
Early 2024Access tokens stolen from an external consultant's computer are accepted by the customer's Azure environment from an unusual location.
- 2
Changes in the subscription
Early 2024, within about ten minutesSigned in as the consultant, the attacker opens a network security rule to inbound SSH, raises a service limit, registers a resource provider and creates or modifies a virtual machine disk.
- 3
Persistence in the directory
Several weeks later, on two consecutive daysThe account invites an external free webmail address, and the next day new MFA information is registered on it.
The attacker held the consultant's access to the customer's Azure environment, opened a network security rule to inbound SSH, prepared to create resources and added persistence in the directory. No data loss is reported and the outcome is not stated.
Where Azure and Microsoft Entra controls fit
Collaboration restrictions that deny invitations to personal email domains, or allow only your partner domains, would have refused the invitation to the webmail address, whoever sent it. Conditional Access policies that limit Azure management to your compliant network and the registration of security information to your trusted locations, Azure Policy deny assignments on inbound SSH rules and on who may create or update disks and virtual machines, and a security admin rule that drops SSH from the internet would have narrowed the other steps, though the sources do not say where the attacker connected from, which kind of token was stolen or which roles the consultant's account held. No Azure setting refuses the registration of a resource provider: the role assignments you give a consultant's account decide it. The tokens themselves were taken on the consultant's own computer, outside your cloud.
Could this have been prevented?
In hindsight, not as a whole. 1 control would have refused a step outright, and 6 controls would have limited one. 2 steps have no cloud-provider control. The counts below are control assessments, not attack stages: several can address the same step. The same replay with the controls in place before the incident shows where each phase would have been refused or bounded.
- 1
- Prevented
- 6
- Contained
- 2
- Not covered
How it could have been prevented
The same replay with the controls in place. Green marks where a step is refused, blue where it is bounded, and red where no control reaches it. Open a row to read the control, why it acts, and its caveat.
Replaying the Azure account hijack, early 2024
1 · Stolen consultant tokens
1 Contained1 Not covered3 on other clouds2 · Changes in the subscription
3 Contained1 Not covered9 on other clouds3 · Persistence in the directory
1 Prevented2 Contained6 on other cloudsKey facts
- Darktrace published the case on July 8, 2025 and dated it only to early 2024; the customer, in the EMEA region, was not named
- Access tokens of a third-party consultant's account were stolen after the consultant downloaded cracked software; how the software took them, and which kind of token it was, was not stated
- With the tokens the attacker authenticated to the customer's Azure environment from a location Darktrace called unusual, and modified a network security rule to allow inbound SSH from a specific address range
- The account raised a resource service limit and registered the Microsoft Capacity resource provider; about ten minutes later a disk tied to a virtual machine was created or modified, which Darktrace read as an attempt to create a rogue VM that no source confirmed
- Several weeks later the account sent a collaboration invitation to an external free webmail address, and the next day new MFA information was registered on it, which the customer confirmed in its logs
- The outcome and the end of the access were not stated; a later Wiz catalog entry, citing only the Darktrace post, stated that a rogue VM was created and labeled the impact data exfiltration, neither of which the Darktrace post reported
Sources
Every fact in this breakdown comes from the public reporting below. The replay is a model of that reporting, not a reconstruction of the environment it describes.
- 1.Defending the Cloud: Stopping Cyber Threats in Azure and AWS with DarktraceDarktrace · Jul 8, 2025 · ResearchToken theft, sign-in to Azure, network security rule change, resource provider registration, disk activity, external invitation, MFA registration, dating
- 2.Azure Account Hijack via Stolen TokensWiz Cloud Threat Landscape · Jul 16, 2025 · ResearchSummary of the chain and technique labels, citing only the Darktrace post; its wording on the VM and the impact
The estate in the replay is illustrative and drawn at the level of attack stages; the victim's real environment is not public. Verdicts describe what a control does when it is in place before the incident. They are not an assessment of the victim's security program.