The CVSS 8.6 score for this Azure Data Factory SSRF materially understates the actual risk because it treats this as a standard server-side request forgery when it's actually operating inside a credential-broker service. ADF's core function is storing and using credentials to connect to Azure Storage, SQL Database, REST APIs, and other cloud resources — meaning an attacker who triggers SSRF isn't just reading network responses, they're operating from a service context that already holds authentication tokens and network positions those credentials provide.
The critical pivot point is reachability to the Azure Instance Metadata Service (IMDS) at 169.254.169.254. If the integration runtime can reach IMDS, this vulnerability transitions from information disclosure to full credential harvesting — service principal tokens, managed identity credentials, and storage keys become accessible. The CVSS calculation likely assumed scoped, isolated impact, but in a tenant with multiple ADF resources, one compromised pipeline can potentially reach metadata across the subscription.
Beyond the immediate exploit window, there's a temporal persistence problem the score doesn't capture. Azure-managed identity tokens typically valid 24-72 hours won't expire when you patch the SSRF. If an attacker harvested credentials during the active vulnerability window before patch deployment, those tokens remain valid post-patch. The standard remediation — apply the ADF patch — closes the vector but doesn't rotate the credentials that may have already been exfiltrated. Organizations should assume compromise and rotate any managed identities, storage account keys, and connection strings the affected ADF instance could access.
Multi-tenancy adds another layer: if the integration runtime's network position is shared across tenants, one tenant's exploited pipeline could bleed into adjacent tenants' credential contexts. This is rarely addressed in CVSS scoring but is essential for platform services.
Prioritize: confirm whether IMDS is reachable from your ADF integration runtime, assume credential exposure and initiate rotation for any secrets the factory accessed, and audit for lateral movement signs across the subscription rather than treating this as a single-resource vulnerability.