A WebProtege Plugin for Attesting to the Provenance of Ontologies on the Ethereum Blockchain

Abstract

Ontologies are shared, formal conceptualizations of a domain that are consumed by human and machine agents alike. Trust in ontologies is a central issue for their application. For example, machine learning algorithms for medical diagnosis may rely on the correctness of ontologies and could potentially deliver false results. For enhancing trust, we developed a WebProtege plugin for the decentralized attestation and veri cation to the integrity and validity of ontologies using the Ethereum blockchain. Blockchains are an immutable, tamper-resistant and decentralized storage where all transactions are digitally signed. Thus, they permit tracing the provenance of concepts and identify responsible actors. For a rst experimental evaluation, we evaluated the transaction costs for attesting to the provenance of ontologies.

Keywords

• Ontology
• Attestation
• Blockchain
• Ethereum
• WebProtege

-

With the recent integration of linked data, knowledge representations and machine learning in the semantic web, it has become essential both for human and machine agents to know about the provenance of data and derived information [ 7,5 ]. Thereby, ontologies play a central role as a formal knowledge resource. Through blockchains as append-only, immutable, decentralized and distributed data stores, trust is achieved through full transparency of the recorded, digitallysigned transactions that are veri ed through peer-to-peer consensus protocols.

In this work we therefore present a plugin for the WebProtege collaborative ontology editor which enables the attestation to the provenance of ontologies using the Ethereum blockchain. In contrast to the full storage of ontologies on blockchains, which is infeasible due to the limited storage space and comparatively high transaction costs, attestations permit decentralized, veri able, and Copyright © 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). transparent proofs of the existence and integrity of information without storing the information itself [ 6 ]. In addition to its adequacy for blockchain-based applications, attestation further prevents the disclosure of the underlying information, while still allowing for conducting proofs of existence via so-called zero-knowledge-proofs if required [ 4 ]. 2

Related Work

Traditionally, digital signatures of ontologies have been used for so-called policybased trust. As a foundation for deriving signatures for RDF graphs, Carroll proposed a canonicalization of RDF graphs without changing their semantics [ 3 ]. An approach for computing a digest of RDF graphs for content identi ers without the need for canonicalization was discussed by Sayers and Karp [ 11 ]. Based on this, Kasten et al. described the signing of individual sub-graphs [ 10 ]. However, digital signatures typically rely on a centralized public-key infrastructure whereas blockchains o er a decentralized, distributed, peer-to-peer architecture [ 14 ]. Multiple bene ts have been previously identi ed for applying blockchains in the semantic web [ 2 ], e.g., for using RDF as the data storage format on blockchains and thus providing a decentralized, immutable, tamper-proof data storage for RDF graphs [ 12 ]. Another approach has been proposed in the form of knowledge blockchains for the transparent monitoring of ontology evolution and proving the existence of concepts without disclosing them using so-called zero-knowledge proofs [ 4 ]. Tuan et al. presented a hybrid approach for storing RDF triplets for use in edge networks, where triplets are stored in a distributed o -chain RDF store but access is controlled by smart contracts [ 13 ]. 3

Extension of WebProtege for Enabling Ontology Attestations The cloud-based, collaborative WebProtege editor is a well-established platform for ontology authoring, which can be extended through a plugin system in the form of portlets [ 9 ]. Therefore we chose it as foundation for realizing our implementation for attesting to the provenance of ontologies. The implementation is composed of three major components. A plugin (i) for WebProtege o ers the UI for attesting to or verifying a loaded ontology. For that purpose, a digest of the ontology is calculated by reverting to the OWL API [ 8 ]. Further, a transaction containing the attestation information (IRI, version IRI, digest, name of the signer) to the Ethereum blockchain is initiated. The connection and interaction with the chain network is provided by Metamask4 (ii), a browser wallet extension. Transactions are submitted to the network from a user account, i.e. users are prompted to login with their Ethereum account and authorize the transaction. Thereby, the transaction is digitally signed and sent to a smart contract 4 Metamask - https://metamask.io/ (iii) on the chain. The contract stores the received data, thus irrevocably persisting the attestation information. Finally, a user may query the smart contract to verify if an ontology has been attested to or was changed. The result of the veri cation presents the user with information on the time and date of the attestation as well as on who attested to the origin of the ontology. In this way it can for example be inferred in a decentralized fashion and without someone having to host the ontology centrally that an expert has validated and attested to the provenance of an ontology, thus increasing the trust that can be placed in that ontology. In Ethereum, size and computational complexity of transactions are measured in an energy unit, called Gas. This is a fee payed by the originator in the cryptocurrency Ether and depends on the complexity of commands to be executed [ 1 ]. The Ether price of a unit of gas is in uenced by the transaction volume. Figure 2 shows historical transaction fees in USD for an attestation in comparison to a baseline, a contract storing a 256bit integer value. E.g., on 1. June 2021, an ontology attestation would have cost USD 8.62 (vs. baseline of USD 3.25). Both gas and ether price are highly volatile. As such, the transaction cost may change signi cantly in a short period of time.

However, Ethereum is in the process of adopting the proof-of-stake consensus mechanism5, enabling higher transaction throughput and better energy efciency. Alternatively, our approach may be adapted to other blockchain-based systems, e.g., Avalanche6, an Ethereum compatible proof-of-stake blockchain, 5 See https://ethereum.org/en/developers/docs/consensus-mechanisms/pos/ 6 Avalanche - https://www.avax.network/ 1000 100 ) 0 1 g o l ( ]D 10 S U [ t s o C iton 1 c a s n a r T 0.1 Mean cost for attestation Baseline transaction cost or the Bloxberg infrastructure7 for decentralized services for the scienti c community. We provide the experimental dataset and prototype implementation as open source8. The attestation approach, in form of a portlet plugin, has been integrated in a fork of the original WebProtege distribution. Submitting authorized transactions to the chain network is not handled by the plugin itself, but instead delegated to the required browser wallet extension (Metamask). Thus, users are required to install this extension prior to using the plugin. In return, users retain full control of chain interactions.

Cost of ontology attestation in USD since Jan. 2020 7 Bloxberg infrastructure - https://bloxberg.org/ 8 Repository - https://github.com/curtys/webprotege-attestation-base Future work will include the investigation of alternative hashing procedures for ontologies for enabling zero-knowledge proofs on a more ne granular level and the extension of the smart contract implementation towards supporting attestations by multiple users.

Acknowledgments

The research on this paper has been partially Science Fund grant number 196889. nanced by the Swiss National

References

1. Antonopoulos , A.M. , Wood , G. : Mastering ethereum: building smart contracts and dapps . O'reilly Media ( 2018 )

2. Cano-Benito , J. , Cimmino , A. , Garc a-Castro, R. : Towards blockchain and semantic web . In: Business Information Systems Workshops . pp. 220 { 231 . Springer ( 2019 )

3. Carroll , J.J. : Signing RDF Graphs . In: ISWC. pp. 369 { 384 . Springer ( 2003 )

4. Fill , H.G. : Applying the Concept of Knowledge Blockchains to Ontologies . In: AAAI 2019 Spring Symposium . CEUR-WS.org ( 2019 )

5. Fill , H., Harer, F.: Supporting trust in hybrid intelligence systems using blockchains . In: AAAI 2020 Spring Symposium. CEUR-WS.org ( 2020 )

6. Harer, F. , Fill , H.: Decentralized Attestation of Conceptual Models Using the Ethereum Blockchain . IEEE CBI Conference 01 , 104 { 113 ( 2019 )

7. van Harmelen , F. , ten Teije , A.: A boxology of design patterns for hybrid learning and reasoning systems . J. Web Eng . 18 ( 1-3 ), 97 { 124 ( 2019 )

8. Horridge , M.: OWL API main repository . https://github.com/owlcs/owlapi ( 2020 )

9. Horridge , M. , Goncalves , R.S. , Nyulas , C.I. , Tudorache , T. , Musen , M.A. : WebProtege: A Cloud-Based Ontology Editor . In: World Wide Web Conference . pp. 686 { 689 . ACM ( 2019 )

10. Kasten , A. , Scherp , A. , Schau , P. : A Framework for Iterative Signing of Graph Data on the Web . In: ESCW Conference . pp. 146 { 160 . Springer ( 2014 )

11. Sayers , C. , Karp , A. : Computing the digest of an RDF graph . Tech. rep., HP Laboratories Palo Alto ( 2004 ), https://www.hpl.hp.com/techreports/2003/ HPL-2003 -235R1.pdf (last access 2021-04-09)

12. Sopek , M. , et al.: GraphChain: A Distributed Database with Explicit Semantics and Chained RDF Graphs . In: The Web Conference 2018 . pp. 1171 { 1178 . ACM ( 2018 )

13. Tuan , A. , Hingu , D. , Hauswirth , M. , Le-Phuoc , D. : Incorporating Blockchain into RDF Store at the Lightweight Edge Devices . In: Int. Conf. on Semantic Systems . pp. 369 { 375 . Springer ( 2019 )

14. Yakubov , A. , Shbair , W.M. , Wallbom , A. , Sanda , D. , State , R.: A blockchain-based PKI management framework . In: IEEE/IFIP Network Operations and Management . pp. 1 { 6 . IEEE ( 2018 )