Abstract

ABSTRACT

A device can be handed to a new owner while the old owner’s reach into it survives. Changing a device owner, a memory permission, or a security-domain label does not prove that the old owner’s direct memory access (DMA), cached address translations, queued commands, pending interrupts, and debug or sideband paths have stopped mattering. If the new owner starts issuing consequential commands before those paths are closed, the old owner can still affect the result. Who can safely activate the reassigned device?

Based on Indian Provisional Filing: 202631136555, filed 15 November 2026, and disclosed with the Mothership (PCT/IB2026/055615) on 16 July 2026.

This patent-pending technical concept proposes a complementary mechanism for multi-path trust-domain transitions. It covers reassigning peripherals, accelerators, secure devices, and physical actuators between domains. Its core principle is that the successor’s first consequential command is not authorized until every mandatory old-owner path has been proven closed.

The mechanism works in six steps:

  1. A transition contract names the resource, the old and new authority, the required path closures, the safety conditions, and the exact successor operation.
  2. An inventory of the prior owner’s paths lists the access and effect routes that apply under the current verified topology and threat model.
  3. An evidence matrix records, for each path, trustworthy proof of completion, invalidation, drain, reset, or containment.
  4. A unique closure epoch ties that evidence to this one transition, so stale evidence cannot be reused.
  5. The successor’s command is held in a non-effective state, and new old-owner commands are restricted. A stop request is never treated as a completed drain.
  6. A transition closure barrier checks that all mandatory evidence is current and satisfied. If any path is failed or unknown, the successor stays disabled and a protected reset or recovery may produce new evidence. If all pass, a narrowly scoped authorization is bound to the resource, the successor, the current epoch, the exact effect, and the target enforcement point.

At that final enforcement point, a stale command from the old owner is rejected and only the successor’s bounded command can take effect. Evidence comes from a hardware completion event, a protected firmware readback, or another sufficiently trusted source. In the software and firmware variant, the enforcement point sits at a trusted device-assignment controller or protected driver dispatcher, and only the covered interfaces are assured. In the stronger hardware variant, it sits at a command-enable latch, a device control register, or an actuator-admission boundary.

Prospective industry relevance (AI-generated; see notice below). The mapping shows where the closure check could sit alongside documented products. Each placement is an engineering hypothesis, and none of these vendors is represented as incorporating this concept.

  • Arm Neoverse CPUs and compute subsystems: the check sits at the protected subsystem resource-activation interface. It reads platform-supported current-state signals and rejects activation of a reassigned accelerator until the old-owner paths are closed.
  • Arm CCA, RME, RMM, and Realm device assignment: the check sits at trusted device-assignment or command admission, where exposed. It adds a path-completeness test around the exact device activation, without replacing the mandated protocols, and rejects activation if a mandatory path is open.
  • Arm CoreLink MMU-700 and SMMUv3: the check adds an evidence row for stale translations and requester context, based on the architecture’s invalidation-completion semantics. It rejects enabling the successor’s command while an old translation can still reach the device.
  • Arm Neoverse CMN-700 coherent interconnect: the check covers outstanding coherent requests during reassignment. It admits the target only after supported fabric completion and rejects admission while old-owner requests are in flight.
  • Arm CoreLink GIC-700 and GIC-720AE: a pending interrupt route to the old owner becomes a mandatory closure condition when it is consequential. The successor’s device-control interface stays disabled until the reroute completes.
  • Arm CoreLink DMA-350 and CoreSight: the check collects closure evidence for DMA drains and privileged debug paths. A protected queue or device activation gate rejects the successor’s command while either remains open.
  • Arm Zena, Cortex-R and Cortex-M, and Ethos-U: the check sits at the safety-domain peripheral command or actuator boundary. It holds the successor in a transition-safe state and keeps independently authorized safety functions running.
  • Intel TDX Connect and VT-d: closure evidence across the IOMMU, the trusted execution environment’s I/O, and guest command paths during secure device detach and assignment. It rejects reuse of a device whose detach is not proven.
  • AMD EPYC SEV-SNP and SEV-TIO: path-coverage tests around trusted PCIe device assignment, IOMMU, and DMA isolation. They reject a trusted assignment if a DMA path is uncovered.
  • NVIDIA BlueField DPU and DOCA: DPU-side admission of network and device egress, only where the programmable hooks truly cover the effect.
  • AWS Nitro System: a reclassification gate at supported customer-controlled service interfaces. Nitro’s internal access is not assumed.
  • Microsoft Azure Confidential VMs: device-reconfiguration and effect admission around supported VM, guest, and accelerator boundaries.
  • Qualcomm Snapdragon Ride Flex: bounded domain-change admission for mixed-criticality cockpit and driver-assistance components, with safety continuity preserved.

These are possible complementary integration points, not claims of deficiency in any named product. Existing mechanisms such as Arm CCA and RME, IOMMU and SMMU, DMA isolation, device assignment, interrupt routing, firmware sequencing, and device reset are not claimed as individually new. The proposed contribution is the enforceable combination of mandatory path-closure evidence and a barrier on the successor’s actual effect. If a platform procedure already enforces equivalent closure, it should be reused. The practical test is to map every material old-owner path of one real reclassification protocol to trustworthy completion semantics, then measure transition latency. No universal latency guarantee is claimed.

Where to read the full filing. Search “Sangam Das” on WIPO PATENTSCOPE, or search PCT/IB2026/055615 (pctib202655615). The full PCT file is also available in Technical Disclosure Commons: https://www.tdcommons.org/dpubs_series/12542/

Basis and notice. The underlying invention is human-originated and based on the Mothership application PCT/IB2026/055615, filed 4 June 2026 and publicly disclosed 16 July 2026. This abstract was prepared with AI-assisted tools, including the industry-relevance section, and may contain errors, omissions, or prospective interpretations. Readers must cross-check it against the original PCT filings and the companies’ own public documentation before drawing any conclusion. Company and product references are for explanatory, complementary architectural alignment only. They are not an endorsement, validation, or infringement statement, and imply no adoption, collaboration, licensing, or finding on patent validity or third-party rights. The original specifications and claims are authoritative.

Part 6 of 39. This is Part 6 of a 39-part technical disclosure series: 38 numbered embodiments plus one separately designated Embodiment 12A. Each part presents one embodiment, its problem, the proposed solution, practical implementation mechanisms, and prospective complementary industry relevance.

License: CC BY-NC 4.0.

KEYWORDS

AI Safety; Hardware Security; Trust-Domain Transition; Confidential Computing; Trusted Device Assignment; TEE-IO; IOMMU; SMMU; DMA Isolation; Address Translation Invalidation; Interrupt Routing; Coherent Interconnect; Device Reset and Quiescence; Peripheral Security; Accelerator Security; DPU Security; Arm CCA; Realm Management Extension; Hardware-Rooted Authorization; Runtime Authorization; Consequence-Bound Authorization; Automotive Functional Safety; Mixed-Criticality Systems; Cyber-Physical Security; Multi-Tenant Cloud Isolation; AI Infrastructure Security

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License

Share

COinS