Transport. Software supply chain security for rail, transit, road, and coach.
Rail operators, urban transit authorities, road transport networks running autonomous and driver-assist software, and inter-city coach fleets all now sit inside the TSA Rail Cyber Directive, EN 50701, UNECE R155, and NIS2 perimeters. Safeguard makes signalling, dispatch, ticketing, and AV / ADAS evidence a live query, not a quarterly binder.
Four forces converging on transport software.
Regulator, safety case, passenger-data, and operational pressures are collapsing into one continuous evidence requirement.
TSA Rail Cyber Directive
US rail operators now face binding cybersecurity performance requirements with named cyber-incident reporting timelines. Spreadsheet evidence on signalling and dispatch software is no longer enough — continuous attestation is.
EU NIS2 transport
Rail, urban transit, road, and coach operators classified as essential entities under NIS2 now answer to ENISA on supply chain risk and incident reporting. The clock is shorter than most internal audits can move.
Signalling-system cyber
Rail signalling and dispatch increasingly run on COTS stacks with long-tail OSS underneath. A single unpatched KEV in a signalling SDK is a safety-critical incident, not a vulnerability ticket.
AV / ADAS safety + cyber convergence
UN R155 and adjacent regimes are pulling autonomous-vehicle and driver-assist software into the same regulator review as safety. Adversarial-input attacks on perception models are now a recall-class event.
Capability mapped to transport regulator expectation.
Signalling-system signed SBOM
Every signalling, dispatch, and interlocking release emits a CycloneDX SBOM with signed provenance. Regulator evidence is a query against a live store, not a binder assembled the week before audit.
AV / ADAS reachability-aware patching
Reachability analysis decides which CVEs are actually exposed in a given AV / ADAS build, not which versions match. Combined with KEV and EPSS, recall risk gets a defendable, ranked worklist.
Ticketing-platform vendor concentration
See your single-point-of-failure components across ticketing, contactless, and revenue-share platforms before procurement signs the next operator contract. Concentration risk surfaces component-by-component.
Control-room MCP-server governance
Agentic copilots in dispatch, OCC, and traffic-management control rooms run through a governed MCP layer. Tool calls are scoped, logged, and prompt-injection tested before they touch a safety-critical interlock.
Frameworks the platform is mapped to.
Pre-mapped control narratives and evidence in the formats your auditor, safety case, and regulator already accept.
A typical deployment in a rail or transit operator.
Rail-DMZ control plane, signalling-CI signing pipeline, per-vendor trust packet for OEMs and integrators, and a regulator-facing evidence export portal for TSA, ENISA, and UNECE oversight.
Rail-DMZ control plane
Control plane sits in the operator's rail-DMZ, between corporate IT and signalling OT. No cross-tenant traffic, no shared key material, no shared dispatch logs.
Signalling-CI signing pipeline
Every signalling, interlocking, and dispatch CI release passes through a signing pipeline. SBOM, VEX, and model attestation are pinned to the build, not retrofitted at audit.
Vendor trust packet
Per-vendor signed bundle of SBOM, VEX, and provenance flows to the operator's TPRM stack. Concentration risk is visible at the component level, across signalling, ticketing, and AV / ADAS suppliers.
Regulator evidence export
Read-only portal exposes signed evidence to TSA, ENISA, UNECE and national regulators on demand — no email attachments, no last-minute spreadsheets.
Four risk surfaces your safety case already worries about.
Signalling-system compromise (rail)
Modern signalling, dispatch, and interlocking systems are increasingly COTS underneath. A single unpatched KEV in a signalling SDK is a safety-critical event, not a vulnerability ticket — and the regulator agrees.
AV / ADAS adversarial-input attacks (road)
Perception models for autonomous and driver-assist stacks are recall-class assets. Adversarial-input attacks on lane keeping, classification, or planning models now require model-attestation evidence, not just unit tests.
Ticketing-platform breach
Ticketing, contactless, and revenue-share platforms concentrate passenger-data and payment flows. A single shared transitive dependency cascades across operators, agencies, and authorities at once.
Control-room ransomware
OCC, dispatch, and traffic-management control rooms are increasingly Windows-stack with long-tail OSS underneath. One ransomware event can stall an entire urban transit network for a working day.
What is actually hitting transport this year.
- Signalling-system KEV CVEsDisclosure-to-exploit cycle frequently under 72 hours; reachability decides which signalling, dispatch, and interlocking builds are actually exposed.We address this through Eagle reachability + KEV prioritisation
- AV / ADAS adversarial-input attacksPerception and planning models for autonomous and driver-assist stacks are recall-class assets. Adversarial-input attacks now need model-attestation evidence.We address this through AI governance + model attestation
- Ticketing-vendor compromiseA single shared transitive dependency across ticketing, contactless, and revenue-share platforms cascades across operators, agencies, and authorities.We address this through Third-party risk concentration heatmap
- Control-room ransomwareOCC, dispatch, and traffic-management control rooms with long-tail OSS underneath can stall an entire urban transit network for a working day.We address this through SCA + reachability prioritisation
- Sanctioned-OEM signalling exposureAcross rail, transit, and AV / ADAS vendor networks, a single sanctioned OEM hidden in the dependency graph can trigger multi-jurisdiction fines.We address this through Sovereign deployment + jurisdictional controls
Quantified benefits for transport operators.
Numbers from production deployments. Same regulator, same vendor stack, dramatically less binder.
| Metric | Before Safeguard | With Safeguard |
|---|---|---|
| TSA audit prep | 8 weeks | 1 day |
| Signalling patch cycle | 30 days | 5 days |
| AV model-attestation prep | 3 weeks | 1 hour |
| Tool consolidation | 8 vendors | 1 |
| Ticketing-vendor concentration mapping | Manual | Automated |
| Alert noise | ~80% | ~5% |
| Passenger-data residency audit | Reactive | Continuous |
Evidence at the speed of your safety case.
Talk to the team about TSA Rail evidence pipelines, AV / ADAS model attestation, and a deployment shape that lives inside your operator's rail-DMZ.