Threat Model: Attacking the Plasma-Control Loop
An attacker who reaches the fast control loop would try to drive the plasma off its operating point; layered identity, attestation, and an independent safety path deny that.
The most consequential target
The L1 fast loop keeps the breeder plasma on its operating point - shape and vertical position at 9.66 MA with negative triangularity delta -0.30 - and manages the burner's mirror/plug regime. An attacker who could issue or alter commands here could attempt to induce a disruption, a vertical displacement event, or an off-normal excursion. This is the scenario the whole architecture is built to defeat.
Attack paths and denials
- Forge a command: denied by mTLS identity - unsigned or wrong-identity commands are rejected.
- Substitute firmware on an FPGA: denied by secure boot and caught by attestation.
- Reach the loop laterally: denied by microsegmentation default-deny.
- Feed false sensor data: constrained by diagnostic integrity and independent SIS sensing.
- Succeed anyway: the independent SIS can still force a safe state.
Timing as a defense and a constraint
The loop is deterministic and sub-millisecond; injected commands that violate timing or rate envelopes are anomalous and rejected. But the same timing budget forbids heavy per-cycle cryptography on the innermost loop, so integrity is enforced at configuration and identity time (attested firmware, authenticated bitstream) rather than by signing every cycle. This is an honest trade: the fast path trusts an attested, isolated FPGA rather than re-verifying each command.
# Command acceptance combines identity, rate envelope, and mode
def accept_cmd(cmd):
if not verify_issuer(cmd): return REJECT # attested identity
if not within_rate_envelope(cmd): return REJECT # physics-plausible
if cmd.op not in mode_allowed(state): return REJECT # legal in this mode
return ACCEPT
Design status: these denials are validated against the digital twin and hardware-in-the-loop rigs, including simulated injection attacks. No live reactor plasma exists yet; the loop runs against simulated plasma models pending FOAK from Q2 2027.