# Museum Chain of Custody and Air Gap Security Research Guide

Original research framework for The Museum Protocol and EAM LLC  
Version 1.0 | November 1, 2026

## Purpose and audience

This guide helps collections staff, conservators, museum leadership, IT teams, and security researchers investigate how physical custody and digital security can support each other. It provides an original research protocol, proposed record templates, and practical acceptance criteria. It does not certify a system, determine title to an object, or establish compliance with a court order.

The central question is: can an independent reviewer reconstruct who was authorized to act, what happened, where the object or data went, and which evidence supports that account?

## 1 Key distinctions

| Concept | Research meaning | Important limit |
|---|---|---|
| Provenance | Documented history of an object's origin, attribution, ownership, and context | A recent handoff cannot fill gaps in historical ownership |
| Custody | Responsibility for possession, handling, or safeguarding over a defined period | Custody does not establish lawful ownership |
| Digital integrity | Evidence that bytes match an identified baseline | Matching hashes do not prove the original information was true |
| Authorization | Recorded permission for an action within a defined scope | Permission does not prove the action occurred |
| Authentication | Establishing confidence in an actor or credential | An account event does not conclusively identify the human at the keyboard |
| Air gap | Separation between specified systems with no physical connection and no automated logical transfer across that interface [2] | A label, screen image, or disconnected cable is insufficient evidence |
| Zero Trust Architecture | An access-oriented architecture requiring explicit decisions rather than presumed trust | It is distinct from an air gap and cannot substitute for a missing custody record |

Collections Trust treats location and movement control as maintaining current object locations and movement history. Its suggested procedure includes authorization, condition checks, handling requirements, and receipt of custody [1]. Use that source with the applicable Spectrum licence; the templates below are original research proposals rather than reproduced Spectrum forms.

## 2 Define the research boundary

Before testing, document the research sponsor, permissions, locations, dates, systems, and participants. Use fictional collection records in public demonstrations. Specify whether the study concerns a physical object, a digital surrogate, a removable medium, an isolated workstation, or their relationship.

Write a bounded claim such as: “During the recorded test interval, workstation LAB-01 had no identified automated communication path to network NET-01 under the inspected configuration.” Avoid “unhackable,” “permanently isolated,” or “proves provenance.”

Research questions:

1. Can a reviewer follow an object's approved movement from request through receipt?
2. Can a digital record be linked to that movement without obscuring changes?
3. Which physical, wireless, peripheral, management, and removable-media paths cross the proposed isolation boundary?
4. Can a file be transferred through a controlled process while preserving a reviewable record?
5. Can records be recovered after disruption and reconciled with the last verified state?

Record assumptions explicitly. A monitor is an output device, not evidence of network isolation. A powered-off device and an operational isolated workstation have different test conditions.

## 3 Assign responsibilities

| Role | Responsibility |
|---|---|
| Museum sponsor | Approves scope, priorities, resources, and residual risk |
| Registrar or collections lead | Defines identifiers, movement authority, custody, and reconciliation |
| Conservator | Defines condition checks and safe handling restrictions |
| System owner | Maintains inventory, configurations, accounts, and restoration arrangements |
| Authorized tester | Executes the approved protocol and preserves observations |
| Evidence custodian | Controls originals, manifests, access, and transfer receipts |
| Independent reviewer | Checks findings against evidence and challenges unsupported conclusions |
| MBISO coordinator | Connects responsibilities, actions, reporting, and educational interpretation |

Smaller teams may combine roles. Record that concentration of authority and add a second-person review for consequential movements or transfers where practical. An AI assistant may summarize approved records; it must not silently authorize movement, alter originals, or supply missing facts.

## 4 Physical custody workflow

1. Identify the object using its museum identifier and an appropriate description. Photograph identifying features only when permitted.
2. Record its current location and custodian. Keep exact storage locations out of public reports.
3. Obtain movement approval and define purpose, destination, timeframe, and handling requirements.
4. Check condition and prepare packing with the responsible collections professionals.
5. Record release, carrier or handler, and each intermediate handoff.
6. Confirm arrival, condition, receipt, and any discrepancies.
7. Update the authoritative location record promptly and reconcile it with temporary paperwork.
8. Retain corrections with author, time, reason, and prior value rather than concealing the earlier entry.

If identity, authority, condition, or destination is uncertain, pause the move and escalate. Emergency evacuation follows the museum's approved emergency procedure; document exceptions and reconcile records afterward.

### Proposed custody event template

| Field | Entry guidance |
|---|---|
| Event ID and object ID | Stable unique identifiers; never recycle an event ID |
| Event type and purpose | Release, transport, receipt, inspection, return, or exception |
| From and to locations | Controlled location identifiers |
| Authorization | Approver, approval reference, scope, and approval time |
| Released by and received by | Identified actors and acknowledgment method |
| Event time and entry time | Include timezone; distinguish observed time from later entry |
| Condition and packing | References to reports, images, and seals where applicable |
| Related digital package | Package ID and manifest reference |
| Exception or uncertainty | What is unknown, who was informed, and current status |
| Review | Reviewer, review date, and correction references |

## 5 Digital evidence and integrity

For each export, photograph set, log collection, or report, assign a package ID. Preserve the original collected bytes and analyze working copies. NIST SP 800-86 provides an IT-forensics foundation for incident investigation; it is not a complete legal evidence procedure [3].

Create a manifest containing relative filename, byte size, cryptographic hash algorithm and value, collection source, collector, collection method, collection time with timezone, and tool/version. Use a modern algorithm such as SHA-256 under the institution's approved policy. Document device clock drift and any uncertainty.

Store the manifest in a controlled repository with access records and version history. A hash saved next to a file on the same writable drive can be changed along with the file. Where supported, use protected storage and authenticated signing; protect signing keys separately. A hash chain can help reveal changes relative to a trusted checkpoint, but an attacker able to rewrite the whole chain may recreate it.

At every transfer, verify against the accepted manifest and record the result. For a mismatch, preserve both versions, suspend reliance on the disputed copy, and investigate. Do not “fix” evidence by overwriting it.

An ordinary application event log should be described as an activity record unless stronger preservation, provenance, and independent verification support a stronger claim. Absence of a logged event does not establish absence of activity.

## 6 Air gap threat model

Identify what needs protection, from whom, and against which plausible paths. Consider accidental reconnection, compromised transfer media, maintenance access, malicious peripherals, credential misuse, physical access, and loss of records. Consider sophisticated covert channels only when the institution's threat model justifies specialist assessment.

Inventory the host, operating system, storage, Ethernet ports, Wi-Fi, Bluetooth, cellular radios, USB adapters, docks, network-capable peripherals, out-of-band management, and power/maintenance arrangements. Record firmware and configuration versions where obtainable.

For HPAOC2016 demonstrations, note that the LV290AA is a wireless peripheral system. Examine its receiver and radio behavior in the threat model. A peripheral radio is not automatically an internet connection, but it is relevant to claims of complete radio isolation. Preserve the unopened LV290AA#ABA box and use separate loose peripherals and test equipment.

Use accurate labels: “offline during this session,” “network segmented,” “isolated from a specified network,” and “air-gapped interface” are different claims. Explain any manual transfer boundary.

## 7 Verification protocol

Use institution-approved test equipment and methods. Do not probe third-party systems or reconnect production equipment merely to prove a point.

| Test | Method | Evidence and decision criterion |
|---|---|---|
| Boundary inventory | Inspect configuration, ports, adapters, radios, docks, and management interfaces | Inventory and diagram complete; unexplained paths mean not verified |
| Physical inspection | Inspect actual equipment and cable endpoints | Record connections and photos when permitted; investigate unknown cables |
| Interface inspection | Review OS and firmware settings and device listings | Reconcile all discovered interfaces with inventory |
| Communication observation | Observe authorized local interfaces and the controlled test network | Unexpected traffic investigated; silence alone cannot prove every path absent |
| Reboot persistence | Restart and repeat relevant checks | No unapproved interface or service re-enabled |
| Maintenance exception | Exercise an approved simulated maintenance change | Authorization, temporary boundary change, and re-verification documented |
| Controlled file transfer | Execute the media protocol with harmless sample data | Source and destination hashes agree; receipts and exceptions retained |
| Restoration | Restore approved data to separate test equipment | Files, access, dependencies, and custody links reconcile |

For every test record: test ID, objective, preconditions, equipment, authorized operator, steps, expected result, actual result, evidence references, limitations, reviewer, and disposition.

Use four outcomes: pass for the stated criterion; fail; inconclusive; not tested. Report individual results and unresolved paths rather than collapsing the study into a misleading overall score. A configuration change invalidates conclusions dependent on the old configuration until reviewed.

## 8 Controlled removable media transfer

Specify permitted direction, data types, approver, transfer device, destination, and retention before transfer. Use dedicated, inventoried media and a controlled inspection station. Avoid media shared casually between public or personal devices and the protected environment.

1. Approve the transfer request and classify the information.
2. Prepare the minimum required files at the source and create a manifest.
3. Inspect media and files using approved tools and applicable content restrictions. Scanning reduces risk; it cannot guarantee clean content.
4. Record media ID, operator, source, destination, inspection result, and release approval.
5. Transfer through the documented manual process.
6. Verify destination hashes and record acceptance or rejection.
7. Retain receipts and handle media under a defined reuse, quarantine, or disposal policy.

Do not sanitize media subject to an incident investigation or preservation hold. For authorized sanitization or disposal, use a media-specific procedure informed by NIST SP 800-88 Revision 2 [4]. Deleting a file is not equivalent to verified sanitization.

## 9 Recovery and discrepancies

Define who can declare an incident, pause a movement, approve restoration, and authorize return to service. Maintain an alternative communication method and an offline contact list protected from public access.

If a custody receipt is missing, mark the gap; contact the relevant handlers and preserve contemporaneous records. Later testimony should retain its date and source rather than appearing as an original receipt. If an unexpected connection appears, preserve relevant configuration and logs using approved response procedures. Avoid improvising shutdown or unplugging actions that could destroy evidence or disrupt safety-critical services.

Practice restoring a collections-record export on separate equipment. Check file integrity, readability, permissions, application dependencies, object identifiers, location history, and outstanding discrepancies. A successful backup job is not a successful restoration test.

## 10 Research scenarios

### Exhibition handoff

A fictional object moves from storage to Gallery A. A receiving acknowledgment is absent. Participants identify the evidentiary gap, assign follow-up, and document whether the object location can be independently confirmed. Success means visible uncertainty and an accountable resolution, not an invented signature.

### Maintenance reconnection

A demonstration workstation is serviced using a network-capable dock. Participants discover the new interface, record the changed boundary, and repeat the applicable checks. Success means withdrawing the earlier isolation claim until the new configuration is evaluated.

### Altered digital package

A harmless test copy differs from its manifest. Participants preserve the source and received copy, record rejection, and trace transfer records. Success means detecting the mismatch without asserting which person caused it from the hash alone.

### Lost online service

The museum's fictional ticketing service is unavailable while local collection records remain accessible. Participants prioritize communications and restoration without treating an offline workstation as a substitute for the entire service environment.

## 11 HPAOC2016 research application

The Museum Protocol may use HPAOC2016 as a conceptual installation for discussing human action and boundaries. Its teaching roles can include displaying evidence, selecting an authorized simulated action, and exporting a clearly labelled demonstration record.

Evaluate artwork, software, physical actuator, and governance workflow separately. Do not infer completed actuation, tamper resistance, identity assurance, or isolation from the presence of an AOC monitor and HP peripherals. PatentSplit remains a proposed architecture unless separately demonstrated and verified; a patent claim or application is not a performance test.

The research should welcome disconfirming results. A useful finding may be that the installation teaches a boundary clearly but does not implement that boundary technically. Keep HPAOC2016's Air-Gap focus distinct from HPAOC2022's Zero Trust Architecture focus.

## 12 Research report and release checklist

Publish a report with objective, scope, authorization, system diagram, methods, results, exceptions, limitations, corrective actions, reviewer, and retest triggers. Include evidence references in a restricted appendix. Remove precise storage locations, credentials, personal information, and sensitive network detail from public copies.

Suggested measures: percentage of sampled handoffs with complete acknowledgments; delay between movement and location update; packages with verified hashes; unresolved interface paths; restoration tests completed; and overdue corrective actions. Define denominators and sample periods. These measures describe the study and should not be marketed as damage prevention or guaranteed security.

Before release, confirm that every consequential statement is supported, every demonstration is labelled, vendor results are attributed, and fictional examples are unmistakable. Do not imply a federal case requirement, court finding, institutional endorsement, or legal ownership determination without the relevant documentation.

## References

Sources consulted November 1, 2026. Original procedures and templates in this guide are proposed research methods; they are not official NIST or Collections Trust forms.

1. Collections Trust. Location and movement control; Spectrum standard and suggested procedure. https://collectionstrust.org.uk/spectrum/primary-procedures/location-and-movement-control-spectrum-5-0-primary-procedures/ and https://collectionstrust.org.uk/resource/location-and-movement-control-suggested-procedure/
2. NIST CSRC glossary. Air gap; definition attributed to CNSSI 4009-2022. https://csrc.nist.gov/glossary/term/air_gap
3. NIST SP 800-86. Guide to Integrating Forensic Techniques into Incident Response. August 2006. https://doi.org/10.6028/NIST.SP.800-86
4. NIST SP 800-88 Revision 2. Guidelines for Media Sanitization. September 2025. https://doi.org/10.6028/NIST.SP.800-88r2
