Azure Data Factory CVE-2026-66800

No special conditions are required for exploitation beyond network access to an Azure Data Factory instance - the CVSS vector (AV:N/AC:L/PR:N/UI:N) confirms this is a remote, unauthenticated, low-complexity attack requiring no user interaction. The primary limiting factor is that impact is strictly confined to confidentiality (I:N, A:N): an attacker cannot modify pipeline configurations or disrupt availability. The practical severity of information disclosure depends on what internal resources the affected Data Factory service identity can reach - instances configured with Private Endpoints, restricted managed identity scopes, or network egress controls will have substantially reduced exposure. There is no indication from available data that exploitation requires a non-default configuration or a specific feature to be enabled.

The CVSS 7.5 score is driven by a network-accessible, unauthenticated, low-complexity attack path with high confidentiality impact - a combination that makes this appear urgent on dashboards. However, cross-referencing with SSVC (Exploitation: none, Technical Impact: partial) and EPSS (0.99%, 60th percentile) significantly moderates that urgency: real-world exploitation has not been observed and automated exploitation, while theoretically feasible per SSVC 'Automatable: yes', has not materialized in tracked threat activity. The confidentiality-only impact (I:N, A:N) means attackers cannot modify data or disrupt the service - the realistic worst case is exfiltration of internal metadata or credentials accessible to the Data Factory service identity. Patch availability from Microsoft further lowers residual risk for patched environments. The key tension in this assessment is that CVSS signals high risk while SSVC and EPSS both signal low current exploitation pressure - organizations should prioritize patching but need not treat this as an emergency response incident unless the affected instance has particularly sensitive internal network access.

An unauthenticated attacker sends a crafted HTTP request to an exposed Azure Data Factory endpoint, supplying a malicious URL targeting an internal resource such as the Azure Instance Metadata Service or an internal storage endpoint. The Data Factory service fetches the attacker-controlled URL on the attacker's behalf and the response - potentially containing managed identity tokens, storage account keys, or internal service responses - is reflected back or accessible to the attacker. No public POC has been identified at time of analysis, but the low attack complexity (AC:L) and automatable SSVC flag indicate the technique is straightforward for a skilled attacker.