JadePuffer, tracked by Microsoft as Storm-3168, used agent-driven attacks to survey Azure environments, collect credentials and destroy cloud resources, with one destructive operation affecting more than 100 storage accounts in seven minutes. The incidents show how compromised service identities can give automated attack workflows considerable reach across cloud infrastructure.
Service identities enabled the attacks
Microsoft observed two attacks in June involving two compromised service principals in the same Azure tenant. One identity was used for reconnaissance and resource discovery, while the other carried out discovery, destructive actions and credential collection. Service principals allow applications, hosted services and automated tools to authenticate to Azure using assigned permissions.
The initial access route remains unclear. Microsoft noted that credentials for one of the service principals appeared in a public GitHub issue before the attacks, pointing to exposed secrets as one possible avenue for compromise.
Destruction spread across Azure services
The attackers deleted Azure Storage accounts and targeted Key Vaults, Function Apps, Virtual Machines and App Services. They also removed Azure Site Recovery locks and other recovery protections, making restoration more difficult. About 30 minutes after the wipe attempts, Storm-3168 returned with more than 30 requests for storage account keys, most of which succeeded.
Some storage accounts survived because Azure resource locks and account-level protections blocked the deletions. Attempts to remove Azure SQL databases also failed because the attacker used an unsupported API version, while efforts to remove recovery-protection locks were unsuccessful.
Broader ransomware implications
Sysdig previously reported that JadePuffer automated much of the attack chain, including reconnaissance, credential theft, lateral movement, persistence and data encryption. The operation later expanded toward AI assets, training datasets and vector databases through a tool called EncForge.
The activity could support ransomware extortion, but Microsoft did not report financial demands and did not confirm data theft in the observed cases. Its description of the parallel targeting of storage and SQL resources indicates an effort to increase destructive impact across several data services, rather than evidence that every targeted system was successfully erased.
Defensive priorities
The researchers recommend enabling cloud workload protections, searching public repositories for exposed secrets and reviewing Azure RBAC permissions against least-privilege principles. The observed failures also show that resource locks, storage protections and recovery controls can limit damage when they remain in place and are configured to resist the compromised identities.
AI-generated editorial illustration; not a documentary photograph.
Source: JadePuffer agentic AI attacks target Azure, destroy cloud resources
SIGNAL TITANS
Leave a Reply