Threat Modeling for NeuroTech & BCIs

NeuroTech threat modeling has to handle a system with three distinct trust domains - implant, clinician programmer, patient-side companion - and a closed-loop control path inside the implant itself. Apply STRIDE per data flow across all of those boundaries, not a single block diagram with a few arrows. Our threat models for this segment use STRIDE-per-element on a system data-flow diagram that includes the in-implant sense-stim loop as a first-class boundary, then trace each identified threat to a risk control in your ISO 14971 file and an SPDF mitigation.

We model the wireless link to the implant explicitly: pairing, session, command authorization, and parameter-write authorization as separate controls. We model the patient app as compromised by default and verify that no implant control depends on app-side enforcement. We model the clinician programmer as both a trusted clinician tool and a potentially stolen device, and demand differentiated controls for each. For closed-loop systems we add a control-integrity threat-modeling pass: which elements of the sense-to-stim pipeline can be tampered with, and what is the patient-harm scenario if they are? Output is an AAMI SW96-aligned threat model document that maps to your hazard analysis, plus the global system view that FDA guidance recommends for connected devices.

For brain-computer interfaces (BCIs) we add BCI-specific elements to the model. Neural recordings are treated as a sensitive asset that cannot be changed once exposed. Decoded output to an external device, such as a cursor, speller, or prosthetic, is modeled as its own data flow. The inductive charging and telemetry link, research data exports, and multi-vendor interfaces between sensors, processors, and stimulators each get their own STRIDE pass. Each threat traces to a patient-harm scenario in your ISO 14971 file.