Trust-AnchoredBlockchainProvenanceSystem(TABPS): AHierarchicalEvidenceFrameworkforAncientManuscript Authentication

Founded by Simpy Hope Limited. Natura Humanitas Journal Homepage: https://www.naturahumanitas.com/ ISSN 3106-731X Original Research Trust-Anchored Blockchain Provenance System (TABPS): A Hierarchical Evidence Framework for Ancient Manuscript Authentication Zhang Haoyang a*, Zhang Luokuan b aSchool of Fine Arts, Sichuan University of Science & Engineering, Zigong, China bFaculty of Business and Economic, University of Malaya, Kuala Lumpur, Malaysia *Corresponding author: zhanghaoy[email protected] Article History Received: 18 October 2025 Revised: 30 October 2025 Accepted: 26 November 2025 Published: 5 December 2025 Keywords Blockchain Provenance; Ancient Manuscripts; Trust Anchor Graph; Verifiable Credentials; Proof-of-Authority Abstract We propose the Trust-Anchored Blockchain Provenance System (TABPS), a hierarchical evidence framework for authenticating ancient manuscripts by integrating domain-specific trust constructs into a blockchain-based provenance model. Existing systems often treat provenance as monolithic metadata, which fails to capture the nuanced and multi-layered nature of historical verification. TABPS addresses this limitation by decomposing trust into a structured taxonomy of evidence tiers, direct physical analysis, contextual scholarly attributions, and procedural custody logs, each formalized as W3C Verifiable Credentials with tailored cryptographic schemas. The primary innovation is found in the Trust Anchor Graph (TAG), a directed acyclic graph that assigns dynamic weights to evidence contributions for blockchain consensus, which permits validators to determine composite trust scores for provenance updates. Furthermore, the system implements an adjusted Proof-of-Authority mechanism in which validator authority increases based on their past correctness, thereby aligning with the specialized knowledge of cultural institutions. For instance, Tier-1 credentials are produced by certified laboratories employing IoT-based spectrometers, whereas Tier-2 attestations safeguard scholar anonymity by means of zero-knowledge proofs. The proposed method not only substitutes traditional curator logs with cryptographically grounded trust anchors but also exports GraphQL-queryable TAGs for transparent evidence tracing. TABPS connects artifact scholarship with decentralized systems, creating a novel framework for provenance tracking that is academically sound and computationally verifiable. https://doi.org/10.5281/zenodo.18005577 1 Introduction Determining the authenticity of ancient manuscripts has traditionally depended on interdisciplinary approaches, where material examination, historical research, and provenance records are jointly applied to verify genuineness. Traditional approaches, however, face challenges in scalability, interoperability, and tamper resistance, particularly as digital archives and cross-institutional collaborations expand. Recent progress in blockchain technology presents a viable approach by -1- Natura Humanitas (2025), 1(1): 35-48 Zhang Haoyang & Zhang Luokuan establishing unalterable, distributed systems for tracing origins. Current blockchain-based systems frequently overlook the hierarchical and context-dependent aspects of trust in manuscript authentication, as they regard all evidence as uniformly weighted inputs to the consensus mechanisms. This paper introduces the Trust-Anchored Blockchain Provenance System (TABPS), a framework that structures the multi-layered trust hierarchies intrinsic to manuscript authentication and embeds these into a verification process grounded in blockchain technology. In contrast to Previous studies that narrowly concentrate on cryptographic immutability or handle provenance metadata as a single entity, TABPS introduces a flexible trust model based on weighted evidence. The system classifies authentication evidence into distinct tiers, which include direct physical analysis (e.g., carbon dating, ink composition) as well as scholarly attributions and custody logs, and converts each into a W3C Verifiable Credential with customized validation rules. These credentials are subsequently compiled into a Trust Anchor Graph (TAG), a directed acyclic graph which dynamically modifies the impact of each evidence type on blockchain consensus. The primary innovation of TABPS is its capacity to connect domain-specific authentication methods with decentralized systems. For example, Tier-1 credentials, produced by certified laboratories with IoT-based analytical instruments hold greater trustworthiness compared to Tier-3 attestations derived from unvalidated historical data. Validators, selected based on their expertise and historical accuracy, employ a modified Proof-of-Authority protocol to compute composite trust scores for provenance updates. This guarantees consensus adheres to scholarly rigor instead of relying solely on computational metrics. Moreover, TABPS safeguards the anonymity of participating scholars by means of zero-knowledge proofs, thereby resolving privacy issues without compromising verifiability. The proposed framework improves upon current methodologies in three notable aspects. First, it introduces a structured taxonomy of trust anchors, which supports fine-grained provenance tracking to reflect the nuanced reality of manuscript authentication. Second, it embeds domain expertise within blockchain consensus, guaranteeing that validation adheres to established scholarly standards. Third, it establishes a transparent, queryable interface for tracing evidence contributions, which supports both academic scrutiny and public accountability. The remainder of this paper is organized as follows: Section 2 reviews related work in blockchain-based provenance tracking and manuscript authentication. Section 3 introduces the foundational concepts of W3C Verifiable Credentials and their applicability to manuscript provenance. Section 4 describes the Trust-Anchor Framework, which encompasses the TAG structure and consensus mechanism. Section 5 presents experimental results evaluating the system’s integrity and validator performance. Section 6 discusses limitations and future directions, and Section 7 concludes the paper. 2 Related Work The tracing of origins for cultural artifacts has seen notable progress With the emergence of blockchain technology, yet current methods Frequently do not meet the specialized needs of verifying manuscripts authenticity. Prior work can be broadly categorized into three areas: generic blockchain provenance frameworks, specialized solutions for cultural heritage, and cryptographic trust management systems. 2.1 Blockchain Provenance Frameworks Multiple investigations have examined blockchain’s capacity for unalterable provenance monitoring in diverse fields. The SOK paper on blockchain provenance [1] presents a systematic classification of requirements and emphasizes the necessity of tamper-evident logs alongside decentralized consensus. However, such Frameworks typically treat provenance metadata as homogeneous data structures without distinguishing between evidence types or their relative trustworthiness. A method based on blockchain technology for open data systems [2] illustrates how Proof-of-Stake (PoS) Protocols can increase scalability, yet their verification process does not account for the semantic interpretation of provenance assertions. In the same way, the modular structure suggested for big data science [3] emphasizes collaboration among institutions, yet fails to embed domain-specific trust metrics within its consensus framework. 2.2 Cultural Heritage Applications In the cultural heritage domain, blockchain has been applied to artifacts tracking with varying degrees of specialization. The digital papyrology platform [4] employs blockchain for version control of manuscript transcriptions, though it lacks mechanisms to validate the underlying physical evidence. A clear system for archaeological participation [5] employs cryptocurrency rewards to crowdsource the examination of artifacts, yet its trust framework depends exclusively on token-based economics instead of academic verification. The IBDCP protocol [6] introduces adaptive consensus for art authentication, yet its risk-distributed approach does not formalize the hierarchical relationships between different evidence categories as done in our Trust Anchor Graph. -2- Natura Humanitas (2025), 1(1): 35-48 Zhang Haoyang & Zhang Luokuan 2.3 Trust Management Systems Recent developments in cryptographic trust management present foundational elements for verifying provenance. The W3C Verifiable Credentials framework [7] supports decentralized claim attestation, which we modify for hierarchical evidence structuring. Current implementations, such as the Universal Authentication Framework concentrate mainly on verifying identity instead of comprehensive historical authentication. Interdisciplinary approaches to merging diverse datasets [8] illustrate the algorithmic assignment of weights to varied forms of evidence, yet their pseudo-labeling mechanisms fall short of the methodological precision demanded in textual analysis. The multi-factor authentication strategy for academic institutions [9] highlights the importance of tiered trust levels, but it’s centralized governance model contrasts with our decentralized validator network. The proposed TABPS framework progresses beyond these studies by introducing three new elements: (1) a hierarchical evidence taxonomy structured as verifiable credentials with domain-specific schemas, (2) a Trust Anchor Graph which dynamically modifies consensus weights depending on the evidence type and provenance, and (3) an adjusted PoA protocol where validator authority stems from proven skill in the manuscript authentication instead of token ownership or computational power. This evidence-aware approach addresses a critical gap in existing systems, which either treat all provenance data equally or rely on trust models divorced from scholarly practice. 3 Preliminaries: W3C Verifiable Credentials and Manuscript Provenance To establish a foundation for the TABPS framework, this section introduces the core concepts of W3C Verifiable Credentials and their Application to manuscript provenance tracking. These standards establish the structural and cryptographic foundation for expressing authentication evidence in a manner that is both machine-verifiable and cognizant of domain-specific requirements. 3.1 W3C Verifiable Credentials The W3C Verifiable Credentials (VC) specification [10] establishes a uniform structure for cryptographically signed attestations, designed for issuance, storage, and verification in decentralized systems. A VC consists of three primary components: 1. Metadata: Includes issuance date, expiration, and issuer identifier. 2. Claims : Statements with a subject-predicate-object structure Concerning the credential holder, such as ‘Laboratory X attests Manuscript Y possesses a 95% likelihood of originating in the 12th century.’ 3. Proof: A digital signature or zero-knowledge proof binds the issuer to the claims. The VC data model accommodates extensible schemas, which permits the embedding of domain-specific properties. For manuscript authentication, we extend this model with scholarly-specific claim types such as material analysis results, stylistic attributions, and custody chain entries. 3.2 Provenance as a Trust Hierarchy Manuscript authentication relies on heterogeneous evidence types with varying degrees of trustworthiness. Methods such as radiocarbon dating and multispectral imaging generally possesses greater epistemic credibility compared to approaches relying on stylistic analysis or historical documentation. TABPS formalizes this hierarchy through a tiered credential system: •Tier-1 (Physical Evidence): Machine-generated credentials from calibrated instruments (e.g., spectrometer readings), signed by accredited laboratories. •Tier-2 (Scholarly Attributions): Expert opinions with cryptographic anonymity preservation via zk-SNARKs [11]. •Tier-3 (Procedural Logs): Custody records and exhibition histories, often derived from institutional archives. This stratification aligns with the Chain of Trust model [12], where each tier inherits verification rules from its superior tiers. For instance, an attribution at the second tier concerning ink composition necessitates citation of at least One primary material analysis to attain complete verifiability. -3- Natura Humanitas (2025), 1(1): 35-48 Zhang Haoyang & Zhang Luokuan 3.3 Blockchain Anchoring for Immutable Timestamping Although venture capitalists furnish portable attestations, the enduring integrity of these documents depends on being anchored to an unchangeable ledger. TABPS employs blockchain not as a storage layer but as a timestamping service for credential digests, thereby guaranteeing the integrity of the data. 1. Non-repudiation: Via periodic checkpointing of credential Merkle roots to the blockchain (Equation 1). Root𝑡=𝐻(VC1∥VC2∥ · · · ∥ VC𝑛)(1) where 𝐻is a cryptographic hash function and ∥denotes concatenation. 2. Temporal Consistency : Each checkpoint contains a link to the previous root, thereby establishing an interlocked sequence of evidence states. This approach balances decentralization with scalability, as only compact digests (typically 32 bytes per VC batch) require on-chain storage. The complete credentials continue to reside outside the blockchain, retrievable by means of content-addressable systems such as IPFS [13]. TABPS’s trust architecture is established by uniting verifiable credentials for evidence documentation, tiered trust hierarchies for academic verification, and blockchain anchoring to prevent tampering. The next section elaborates on how these elements are synthesized into The Trust Anchor Graph and consensus protocol. 4 Trust-Anchor Framework for Blockchain-Anchored Manuscript Credentials The Trust-Anchor Framework implements the tiered evidence model by employing a directed acyclic graph (DAG) structure, which dynamically assigns weights to authentication inputs according to their academic foundation. As illustrated in Figure 1, the Trust Anchor Graph (TAG) functions as the foundational structure for cryptographic provenance tracking, with nodes denoting verifiable credentials and edges indicating logical dependencies among evidence tiers. This structure permits validators to determine trust scores that account for context while retaining the inherent hierarchical connections essential to manuscript verification. Figure 1: Trust Anchor Graph (TAG) and Blockchain Integration 4.1 Construction of the Trust Anchor Graph (TAG) for Manuscripts The Trust Anchor Graph (TAG) formalizes manuscript authentication as a directed acyclic graph 𝐺=(𝑉, 𝐸), where vertices 𝑉represent verifiable credentials (VCs) and edges 𝐸denote logical dependencies between evidence tiers. For a manuscript 𝑀, each vertex VC𝑖∈𝑉belongs to one of three tiers: •Tier-1 (Direct Evidence): Vertices containing machine-generated attestations from calibrated instruments (e.g., VCcarbon with radiocarbon dating results 𝑡±Δ𝑡). •Tier-2 (Contextual Evidence): Vertices encoding scholarly attributions (e.g., VCstyle asserting paleographic matches with confidence score 𝑝). -4- Natura Humanitas (2025), 1(1): 35-48 Zhang Haoyang & Zhang Luokuan •Tier-3 (Procedural Evidence): Vertices documenting custody chains (e.g., VCcustody listing ownership transfers). Edges (VC𝑖→VC𝑗) ∈ 𝐸enforce dependency constraints where lower-tier credentials must reference higher-tier ones. For instance, stylistic analysis at Tier-2 level ( VC_style ) necessitates a connection to at least one Tier-1 material analysis ( VC_ink ) to verify consistency in ink composition. The graph’s acyclic property prevents circular reasoning in evidence validation. The TAG construction algorithm processes raw authentication data as follows: 1. Vertex Initialization: For each evidence item, generate a VC with tier-specific schema: VC𝑖={type :𝜏, claims :𝐶, proof :𝜎}(2) where 𝜏∈ {Tier1,Tier2,Tier3},𝐶contains domain-specific claims (e.g., 𝐶Tier1 ={dating :𝑡±Δ𝑡}), and 𝜎is a cryptographic signature or zero-knowledge proof. 2. Edge Formation : For Tier-2 and Tier-3 VCs, establish connections to all cited Tier-1 and Tier-2 VCs correspondingly. The dependency check ensures: ∀VC𝑗∈𝑉Tier𝑘,∃(VC𝑖→VC𝑗) ∈ 𝐸where VC𝑖∈𝑉Tier(𝑘−1)(3) 3. Graph Validation : Exclude any TAG that breaches the tier hierarchy or includes cyclic dependencies. This framework permits validators to navigate evidence chains and determine composite trust scores, with further explanation provided in Section 4.2. 4.2 Evidence-Aware Consensus Mechanism in TABPS The consensus mechanism in TABPS differs from traditional blockchain protocols by including tier-weighted validation of authentication evidence. A composite trust score (S(T)) is defined for every transaction (T), aggregating inputs from all linked verifiable credentials within the Trust Anchor Graph (TAG). The score computation follows: 𝑆(𝑇)= 𝑛 ∑︁ 𝑖=1 𝑤𝑖·𝑓(VC𝑖)(4) where VC𝑖denotes the 𝑖-th credential in the TAG, 𝑤𝑖is its tier-specific weight, and 𝑓(VC𝑖)maps the credential to a trust metric. The weights 𝑤𝑖are governance parameters set as 𝑤1=0.6for Tier-1, 𝑤2=0.3for Tier-2, and 𝑤3=0.1for Tier-3, reflecting the relative epistemic Authority of each evidence type. The trust mapping function 𝑓(VC𝑖)differs per tier: 1. Tier-1 Credentials: Binary trust based on instrument calibration status and issuer accreditation: 𝑓(VCTier1)=(1if 𝜎lab valid and device certified 0otherwise (5) Here, 𝜎lab represents the digital signature of an accredited laboratory. 2. Tier-2 Credentials: Probabilistic confidence derived from scholarly consensus or Bayesian analysis: 𝑓(VCTier2)=𝑝where 𝑝∈ [0,1]is the claim’s confidence score (6) For example, a paleographic match may carry 𝑝=0.85 if supported by three independent experts. 3. Tier-3 Credentials: Binary validity checks on procedural continuity: 𝑓(VCTier3)=(1if CIDprev matches prior record 0otherwise (7) where CIDprev is the Content Identifier of the preceding custody entry. A transaction 𝑇is accepted only if 𝑆(𝑇) ≥ 𝜃, where 𝜃is a dynamically adjusted threshold. The threshold responds to network conditions, increasing when conflicting claims arise (e.g., 𝜃=0.8during disputes versus 𝜃=0.6for routine updates). Validators compute 𝑆(𝑇)by traversing the TAG to verify all credential dependencies, ensuring no Tier-2/Tier-3 claim is evaluated without its requisite Tier-1/Tier-2 anchors. -5- Natura Humanitas (2025), 1(1): 35-48 Zhang Haoyang & Zhang Luokuan 4.3 Domain-Specific VC Schemas and Cryptographic Extensions The TABPS system’s verifiable credentials demand specialized schemas tailored to specific domains to accurately address the complexities of authenticating manuscripts. We define three credential types with distinct claim structures and validation rules, each extending the W3C VC data model with scholarly-relevant fields. Tier-1 VC Schema (Physical Evidence) For direct material analysis, the schema embeds scientific measurements with their uncertainty ranges: VCTier1 ={type :MaterialAnalysis,claims :{parameter :𝜇±𝛿, method :𝑚},proof :𝜎issuer}(8) Here, 𝜇±𝛿represents a measured value with confidence interval (e.g., radiocarbon date 1150 ±25 CE), 𝑚specifies the analytical technique (e.g., “FT-IR spectroscopy”), and 𝜎issuer is a signature from an accredited lab. The schema requires machine-readable units and standardized method codes to support automated validation. Tier-2 VC Schema (Contextual Evidence) Scholarly attributions incorporate confidence scores derived from Bayesian consensus networks: VCTier2 ={type :Attribution,claims :{assertion :𝑎, confidence :𝑝, references :[VCTier1]},proof :𝜋zk}(9) where 𝑎is a scholarly claim (e.g., “script matches 12th-century Byzantine style”), 𝑝∈ (0,1)quantifies confidence via methods like [14], and 𝜋zk is a zero-knowledge proof concealing the scholar’s identity while verifying their credentials. The references field establishes connections to Tier-1 VCs, thereby upholding the TAG’s dependency requirements. Tier-3 VC Schema (Procedural Evidence) Custody documentation employs hash-chains to establish chronological consistency. VCTier3 ={type :Custody,claims :{event :𝑒, timestamp :𝜏, previous :CIDprev},proof :𝜎owner}(10) Here, 𝑒describes a transfer event (e.g., “Donated to Museum X”), 𝜏is the event time, and CIDprev is the IPFS Content Identifier of the prior custody VC. The signature 𝜎owner comes from the relinquishing party. Cryptographic Extensions To address privacy and scalability: 1. Selective Disclosure : Tier-2 VCs employ BBS+ signatures [15] for redaction, which grants scholars the capability to disclose only particular subsets of claims. 2. Batch Anchoring: Multiple VCs are hashed into a single Merkle tree root 𝑅for efficient blockchain checkpointing: 𝑅=𝐻(VC1∥𝐻(VC2∥𝐻(. . . ))) (11) where 𝐻is SHA-256 and ∥denotes concatenation. Validators request full VCs only when disputing a claim. These schemas guarantee that each evidence tier keeps its domain-specific validation logic intact while staying interoperable within the TAG framework. The next subsection details how validator reputation further refines this trust calculus. 4.4 Proof-of-Authority (PoA) with Reputation Tracking TABPS improves the traditional Proof-of-Authority (PoA) consensus mechanism by introducing dynamic validator reputation weights based on historical authentication accuracy. In contrast to static PoA systems With fixed authority node influence, our approach dynamically modifies validator voting power (𝑣𝑝) according to a reputation score (𝑟∈ [0,1]) that is regularly refreshed. 𝑣𝑝=𝑟·𝑠base (12) where 𝑠base is a base stake amount set by governance. The reputation score 𝑟for validator 𝑉is computed as: 𝑟𝑉=𝛼·acc𝑉+ (1−𝛼) · part𝑉(13) Here, acc𝑉measures 𝑉’s historical decision accuracy, calculated as the fraction of past validations that aligned with final consensus (weighted by time decay factor 𝜆): acc𝑉=Í𝑛 𝑖=1𝜆𝑡now −𝑡𝑖·I(dec𝑖=cons𝑖) Í𝑛 𝑖=1𝜆𝑡now −𝑡𝑖(14) -6- Natura Humanitas (2025), 1(1): 35-48 Zhang Haoyang & Zhang Luokuan where dec𝑖is 𝑉’s decision on transaction 𝑖, cons𝑖is the final network consensus, and Iis the indicator function. The participation factor part𝑉incentivizes consistent engagement: part𝑉=1−𝑒−𝛽·𝑁𝑉(15) with 𝑁𝑉being 𝑉’s participation count in recent epochs and 𝛽a sensitivity parameter. The reputation contract enforces three key behaviors: 1. Progressive Decay: Unused validator influence diminishes via 𝑟←𝛾·𝑟per epoch (𝛾=0.99). 2. Misconduct Penalties: Provably incorrect validations trigger 𝑟←𝑟/2. 3. Tier-Specialization: Validators declare expertise domains (e.g., radiocarbon dating), and their 𝑟weights higher for matching TAG tiers. This system guarantees institutions such as the British Museum or the Louvre, which possesses established manuscript expertise, inherently gain greater consensus authority compared to general node operators, while upholding accountability by means of clear reputation monitoring. 4.5 Substitution of Traditional Provenance Inputs and GraphQL-Encoded TAG Queries TABPS substitutes traditional provenance documentation with cryptographic trust anchors, guaranteeing that every historical assertion is verifiably connected to its corroborating evidence. Traditional museum catalog entries, often unstructured text or spreadsheet records are transformed into machine-readable credential chains. For a custody event 𝐸 previously documented as “Manuscript X transferred from Collector Y to Museum Z in 1923”, TABPS generates: 1. A Tier-3 Verifiable Credential VCcustody containing: {type :Transfer,prevCID :𝐻(VCprior),newOwner :did:museum:z,timestamp :𝜏}(16) where 𝐻(VCprior)is the hash of the preceding custody VC and 𝜏is the notarized transfer time. 2. A Tier-1 Material Analysis ( VCpaper ) Conducted by the conservation laboratory of the receiving institution, which establishes the manuscript’s physical continuity. {fiberMatch : 0.92,inkType :IronGall,confidence : 0.95}(17) 3. A Tier-2 Attribution ( VCstyle ) derived from paleographers, with an unidentified source stating: {period :Late Byzantine,confidence : 0.87,refs :[VCpaper]} (18) These credentials form a subgraph in the Trust Anchor Graph, where edges enforce that VCcustody depends on VCpaper, which in turn anchors VCstyle. To enable scholarly interrogation of this structure, TABPS implements a GraphQL interface that exposes TAG traversals as queryable paths. A curator investigating provenance gaps could submit: 1query { 2manuscript(id: "X") { 3custodyEvents { 4eventType 5timestamp 6supportedBy { 7... on Tier1VC { 8parameter 9method 10 } 11 ... on Tier2VC { 12 assertion 13 confidence 14 } 15 } 16 } 17 } 18 } -7- Natura Humanitas (2025), 1(1): 35-48 Zhang Haoyang & Zhang Luokuan The query engine resolves this by: 1. Fetching the root VCcustody from IPFS using its Content Identifier (CID). 2. Recursively retrieving linked credentials via their refs or prevCID fields. 3. Applying tier-specific filters to return only requested claim types. 4. Validating the cryptographic signatures of the traversed credentials to ensure the VCcustody chain remains unbroken. This substitutes traditional provenance research workflows, which often require manual cross-referencing of lab reports, auction catalogs, and scholarly papers, with a cryptographically verifiable evidence graph. The system’s temporal consistency checks additionally identify anomalies such as custody gaps or contradictory material analyses, thereby indicating possible forgeries or documentation inaccuracies when: ∃VC𝑖,VC𝑗∈𝐺where claims(VC𝑖) ∩ claims(VC𝑗)≠∅ ∧ VC𝑖.VC𝑗(19) Here ≡denotes semantic equivalence under domain-specific rules (e.g., differing radiocarbon dates beyond instrument error margins). These conflicts prompt a reassessment of the consensus protocol for the impacted TAG subgraphs, which is detailed in Section 4.2. The GraphQL schema reflects the TAG hierarchy, which grants the ability to query intricate evidence chains with precision comparable to conventional academic citations while capitalizing on blockchain’s unalterable assurances. This dual-layered approach, structured credentialing for machines and flexible querying for scholars, bridges The gap between cryptographic verification and humanities research practices. 5 Experimental Evaluation of Provenance Integrity and Validator Consensus To assess the performance of TABPS, we carried out thorough experiments evaluating three critical dimensions: (1) the system’s capacity to preserve provenance integrity in adversarial scenarios, (2) the consensus protocol’s conformity with academic validation standards, and (3) the computational burden imposed by the Trust Anchor Graph (TAG) framework. Our assessment contrasts TABPS with standard blockchain provenance approaches and conventional authentication techniques, employing both artificially generated and authentic manuscripts collections. 5.1 Experimental Setup and Datasets Test Collections: TABPS was assessed with three different manuscript collections, each presenting unique authentication difficulties. 1. Byzantine Charters (12th-15th century) : 247 documents with known provenance chains and material analyses [16]. This dataset serves as a reliable reference for monitoring custody and verifying tangible evidence. 2. Dead Sea Scroll Fragments : 93 disputed fragments with conflicting scholarly attributions [17]. Employed to assess the system’s capacity to manage disputed assertions of genuineness. 3. Synthetic Forgery Set : Fifty algorithmically produced documents designed to resemble medieval manuscripts, containing deliberate variations in ink composition and handwriting attributes. Serves as controlled test cases for fraud detection. Baseline Systems: We compared TABPS against: 1. Generic Blockchain Provenance (GBP): A standard Ethereum-based implementation treating all metadata equally [18]. 2. Academic Ledger (AL) : A dedicated framework for academic research employing the Proof-of-Stake mechanism [19]. 3. Traditional Curation (TC) : Traditional museum authentication procedures entail both material examination and assessment by specialists. Metrics: Evaluation focused on: 1. Provenance Accuracy (PA): Fraction of correctly authenticated custody events and material analyses. 2. Conflict Resolution Rate (CRR): Percentage of disputed attributions resolved correctly. 3. Trust Score Deviation (TSD): Mean absolute difference between TABPS trust scores and ground truth scholarly consensus. 4. Latency: Time from evidence submission to final validation. -8- Natura Humanitas (2025), 1(1): 35-48 Zhang Haoyang & Zhang Luokuan 5.2 Provenance Integrity Under Adversarial Conditions We tested TABPS’s resilience against three attack vectors common in manuscript authentication: 1. Custody Chain Tampering: Adversaries attempted to insert fake ownership records or alter existing ones. As indicated in Table 1, TABPS identified 98.7% of forgeries by means of its Tier-3 credential validation rules, achieving higher performance than GBP (72.3%) and AL (85.1%). The system’s hash-chained custody VCs (Equation 10) and mandatory Tier-1 anchors prevented undetected modifications. Table 1. Forgery Detection Rates Across Attack Types Table 1: Detection Accuracy Comparison Attack Type TABPS GBP AL TC Fake Custody Record 98.7% 72.3% 85.1% 94.2% Material Misattribution 96.5% 61.8% 78.9% 92.7% Style Forgery 93.2% 54.6% 70.3% 88.4% 2. Material Misattribution: Attackers submitted falsified Tier-1 analyses (e.g., manipulated radiocarbon dates). TABPS’s equipment attestation criteria (Equation 5) and cross-verification with current TAG nodes yielded 96.5% detection accuracy, compared to 61.8% for GBP. Traditional methods performed well (92.7%) but required manual lab verification. 3. Style Forgery: Sophisticated forgeries imitating historical scripts were identified by TABPS’s Tier-2 credential checks (93.2%), owing to its stipulation for multiple independent attributions (Equation 6). The system detected inconsistencies in cases where forgeries were not supported by Tier-1 evidence. The composite trust score mechanism (Equation 4) was notably successful in countering blended assaults that merged partial truths. For instance, a counterfeit document employing genuine parchment from the 12th century yet contemporary ink was identified when the Tier-2 ink analysis score (p) contradicted the Tier-1 material evidence. 𝑆(𝑇)=0.6·1parchment +0.3·0.2ink < 𝜃 =0.7(20) 5.3 Validator Consensus Performance We examined the degree to which TABPS’s adjusted Proof-of-Authority mechanism (Section 4.4) aligns validator choices with authoritative academic evaluations. Figure 2 shows the relationship between the number of validators and composite trust score stability for the Dead Sea Scroll fragments dataset. -9-