Azure's Weakest Link – Five Full Cross-Tenant Compromises

Accessing Azure Key Vault secrets and executing arbitrary SQL queries across tenants is now possible.

Security researcher Haakon Wik Gulbrandsrud has detailed five previously undisclosed cross-tenant vulnerabilities in Azure's API Connections, a component of Azure Logic Apps. These flaws allow an attacker to gain full access to any connected backend service, including Azure Key Vaults, Azure SQL databases, and third-party services like Jira and Salesforce, by tricking the Azure API Management instance into using a victim's connection token. The vulnerabilities were exploited using undocumented ARM API endpoints, such as DynamicInvoke and DynamicList, enabling attackers to perform actions like reading secrets from Key Vaults or executing arbitrary SQL queries on databases. Gulbrandsrud reported these issues, which could have potentially been exploited to compromise various services across different Azure tenants. The initial fix involved blocking these specific endpoints, but the underlying architectural weaknesses may persist if the endpoints are re-enabled without more robust security measures.

The discovery of these vulnerabilities in Azure's API Connections highlights a significant architectural weakness in how Azure manages connections and API calls between different services and tenants. The ability to exploit undocumented ARM API endpoints and bypass standard security controls suggests a deeper issue with privilege escalation and inter-service communication security within the Azure ecosystem. This breach of trust between services, where a connection intended for one purpose can be hijacked for another, poses a substantial risk to organizations relying on Azure for sensitive data and operations.

The widespread impact of these vulnerabilities means that any Azure customer utilizing Logic Apps, particularly those integrating with external services or storing sensitive data in Key Vaults or SQL databases, could be at risk. While Microsoft has reportedly blocked the specific exploited endpoints, the underlying mechanism for cross-tenant access remains a concern. Future vigilance will be required to ensure that subsequent fixes address the root cause rather than just symptomatic exploits, and that organizations themselves review their access control configurations to minimize potential exposure.