Imagine a cloud platform as a high-rise hotel: a guest occupies their own room – just like a tenant, or a logically isolated cloud account or virtual machine – while relying on the provider to safely manage the building. Kaspersky Security Services has published new research warning that traditional cloud audits often only inspect the lobby, failing to test whether a guest’s room key could secretly unlock staff corridors or neighboring suites.
Kaspersky’s new cloud penetration testing methodology shifts the focus away from surface checks to rigorously test the internal "trust boundaries" that prevent cross-tenant compromise. Detailed info is available in a new report.
Traditional penetration testing frequently stops at external web consoles and public APIs. However, Kaspersky’s research demonstrates that achieving remote code execution inside a client’s isolated room (container) is rarely the ultimate goal of an attacker. Without strict isolation controls, an attacker can leverage an initial breach to pivot into staff-only zones – stealing internal provider service identities (IAM tokens), accessing central backup pipelines or reaching other customers' private data.
To systematically evaluate these hidden pathways, Kaspersky categorizes cloud platforms into three operational layers: External (the public portal and APIs), Overlay (internal service runtimes, management tools and provider workflows) and Underlay (physical hardware, hypervisors and storage networks). By tracing how network connections, identities and commands move across these layers, the framework allows security teams to test critical attack vectors – including service token hijacking, lateral movement across provider networks and virtual machine breakouts onto underlying host hardware.
“Obtaining code execution in an isolated client container is not the end of a cloud penetration test-it is only the beginning. Just as a hotel guest shouldn't be able to enter staff corridors from their room, true cloud security requires proving that logical boundaries between tenant workloads, management planes and hypervisors hold up against real-world attacks,” said Matvey Bukhtoyarov, Security Services Expert at Kaspersky.
Based on the research findings, Kaspersky recommends that cloud service providers and enterprise architects first validate multi-tenant isolation by conducting targeted penetration tests designed to explicitly attempt cross-tenant movement between client workloads and provider management layers. They should also enforce strict IAM boundaries, limiting the scope, lifetime and permissions of service tokens used by background workers and automated management tasks to prevent identity theft. Finally, shared infrastructure services should be isolated, ensuring that central logging, telemetry and backup pipelines maintain strict logical and cryptographic separation between individual client tenants.
Tracy