Grafana Mimir vs ThingsBoard

ThingsBoard ships every release twice — once on the 4.3 line and once as a 4.2 backport with an identical security section — and those security sections now dominate the notes, running to twenty or thirty CVEs per release. The recurring classes are telling: SSRF through AI model provider URLs, SSRF and file access escapes from the TBEL script sandbox, DNS rebinding bypasses, and access control on alarm comments. Feature work continues underneath, mostly IoT Hub integration and an Angular 20 UI migration.

The platform is paying down the security cost of being extensible. TBEL scripting and user-configurable AI model endpoints are exactly the features that make ThingsBoard useful for industrial rule engines, and both are repeatedly the source of sandbox and SSRF findings — so the work has shifted to fencing them with allow-lists, opt-in SSRF protection and configurable security headers. Meanwhile the AI surface keeps growing, with structured output support spreading across more model providers.

The dual-branch pattern will hold, with 4.2 continuing to receive the same security sets as 4.3 until it reaches end of life; expect further hardening of the TBEL sandbox rather than new scripting capability.