Answer per control. Each answer feeds the evaluation and the traceability chain Actor → Vector → Component → Effect → Audit.
Is a complete SBOM available for every shipped software & firmware artefact?
A Software Bill of Materials lists every open-source and third-party component (and version) inside a build. Without it you cannot answer 'are we affected by CVE-X?' in hours. SBOMs must cover firmware and embedded artefacts, not just server software.
Are all firmware and software updates signed, with on-device signature verification?
Every firmware/software update must be cryptographically signed by a controlled key, and the device must verify the signature before accepting it. This blocks the classic 'malicious update' supply chain attack — the highest-impact path into deployed equipment.
Is bitstream origin verified and locked against unauthorised reconfiguration?
FPGAs are reconfigured from a bitstream that defines their hardware behavior. Unverified bitstreams can implement hidden backdoors invisible to software analysis. Origin verification, encrypted bitstreams and lockbit settings make unauthorized reconfiguration practically impossible.
Is the build toolchain reproducible and integrity-verified across releases?
If the build toolchain (compiler, linker, signing tools) is tampered with, every binary it produces is suspect — the classic Ken Thompson attack. Reproducible builds let independent parties verify that source code maps deterministically to the released binary.
Are ATR / autonomy models tracked with training data lineage and version control?
AI/ML models (automatic target recognition, autonomy) inherit risks from their training data and pipeline. Tracking model versions, dataset hashes and training environment lets you reproduce a model, investigate misbehavior and respond to data-poisoning disclosures.
Are Tier-2 and Tier-3 sub-suppliers disclosed for safety- and mission-critical components?
Risk is rarely at the integrator level — it is in the chip foundry, the sensor crystal supplier, the open-source library two layers deep. Disclosure of Tier-2/3 sub-suppliers for safety- and mission-critical parts lets the buyer assess country-of-origin and concentration risk.
Is there a defined SLA for CVE triage, patching and incident-triggered re-validation of firmware?
New vulnerabilities are disclosed daily. A defined SLA for triage (is this CVE applicable?), patching, and re-validation (does the patched firmware still meet safety requirements?) prevents both 'unpatched for years' and 'rushed patch breaks the platform' failure modes.
Are build servers, signing infrastructure and engineering workstations isolated from general IT and internet?
Build infrastructure and signing keys are the crown jewels. If a build server is on the corporate intranet with email access, a phishing email can compromise every product the company ships. Isolation, MFA, and audit logging are non-negotiable.
Are ATR / autonomy models tested against adversarial inputs, sensor evasion and data poisoning?
ATR and autonomy models can be fooled by inputs crafted to evade detection (adversarial patches, acoustic shapes) or by long-term data-poisoning of public training sets. Red-team testing against these attack classes must happen before deployment, not after an incident.
Are third-party payload interfaces (data + power) gated by an authenticated, schema-validated bridge?
Third-party payloads (additional sensors, communication modules) plug into the platform. Without an authenticated, schema-validated bridge, a malicious or buggy payload can inject commands, exfiltrate data, or cause faults in the host platform.
