CVE-2026-73616 is a cross-realm authorization bypass in OpenRemote's notification deletion endpoints. Any authenticated realm administrator can delete notifications belonging to any other realm — the endpoints never verify that the notification's realm matches the requestor's realm context. The CVSS 6.5 (AV:N/AC:L/PR:H/UI:N/S:U/C:N/I:H/A:N) reflects this: authenticated, moderate-complexity exploitation with high impact on integrity but no confidentiality or availability component.
What makes this worth your attention is the pattern, not the single vulnerability. Multiple deletion paths lack the same check, suggesting the gap emerged from how OpenRemote's framework handles multi-tenant authorization rather than scattered developer oversight. The framework provides authentication primitives but places realm boundary enforcement entirely on individual developers at the application logic layer — there's no middleware or decorator that automatically validates realm context. When building notification functionality, the primary mental model is "delete this notification," and developers reasonably assume the authentication layer handles cross-realm concerns. Under deadline pressure, that assumption ships.
Here's what to check and do:
First, identify your exposure. Any OpenRemote instance running with multiple realms (master realm plus tenant realms) is affected if those tenants have administrative users. The vulnerability exists in the notification deletion API paths — you should audit which endpoints in your deployment expose this surface and whether custom extensions add more.
Second, prioritize the fix. The patch adds explicit realm validation to the deletion endpoints, verifying that the target notification's realm matches the authenticated user's realm context before executing the delete. If you're on an unpatched version, treat this as a scheduled-maintenance priority rather than emergency — there's no evidence of active exploitation, but the attack surface is straightforward to reach over the network.
Third, look for similar patterns. The architectural gap that produced this vulnerability likely exists elsewhere. Endpoints handling write operations on tenant-scoped resources without explicit realm validation are the search target. Deletion endpoints are particularly dangerous because they combine write access with no natural resource ownership check — a pattern that has recurred across multi-tenant platforms. Audit any legacy or deprecated endpoints that predate the current multi-tenant architecture; these are the most likely survivors of the same gap.
Finally, consider the upstream question. Whether OpenRemote's framework should enforce realm isolation at the routing or gateway layer rather than the application layer is a design decision worth revisiting. The fix here is correct, but a framework that makes the secure path the easy path would prevent the next CVE in this family.