University of Tasmania

File(s) under embargo

Less than a day

until file(s) become available

Capitalizing knowledge using blockchain : multi blockchain based framework for rule- based knowledge base systems (KBS)

posted on 2023-05-27, 19:47 authored by Ali RazaAli Raza
Traditional Knowledge base systems (KBS) are typically constrained by their ability to acquire new knowledge without the intercession of a technical knowledge engineer. Even though there are several mature, formal methods available for knowledge capture, these techniques are conducted by knowledge engineers. This introduces a fundamental disconnect between the system and domain experts, causing lengthy delays and cogitative disruption for the experts during the attempts to capture their knowledge. Ripple Down Rules (RDR) and Multiple Classification Ripple Down Rules (MCRDR) address this problem as they provide a knowledge acquisition methodology where a domain expert can incrementally provide knowledge to classify case data in the local context in which it occurs. The justification for the classification is only required against cases previously seen by the system, meaning the expert only has to differentiate a case's classification based on what has already been classified. This ripple down methodology allows the knowledge base to be incrementally built and maintained without the need of a knowledge engineer. In contrast with single classifications for a case in RDR, MCRDR provides the potential for multiple classifications thus making MCRDR more suitable for domains where multiple classifications for a case are required. Traditionally, MCRDR KBSs are implemented in a constrained and isolated fashion with the expectation there is only one domain expert to populate the system's rule-base at a given time. Also, the rule-base and associated cornerstone cases that were used in rule formation are themselves stored in traditional databases with centralised access control that are subject to tampering. In addition to security concerns, tampering has the potential to be a contradiction to the fundamental incremental refinement technique MCRDR provides and, in case of large rule-bases, an unauthorised access to tamper any rule without audit-trail would create integrity issues about the inferred knowledge received. These issues are of grave importance when it comes to domains such as medicine where knowledge inferred from rulebased KBS may be critical to health. Similarly, concurrent rule addition from multiple experts may result in rule contradiction thus creating validation issues. Due to the huge success of bitcoin, Blockchain technology has attained widespread attention in industry, academia and research. It is based on Distributed Ledger Technology (DLT) storing data in the form of transactions which reside in hash-linked blocks. The ledger consistency and user security are maintained by using asymmetric cryptography and distributed consensus algorithms thus making blockchain persistent, auditable, decentralised and anonymous. This makes blockchain suitable to conduct the research associated with rulebased KBS. To address the shortcomings of tampering, integrity, single point of failure and validation of knowledge in RDR/MCRDR, this research proposes a decentralised framework for capturing, storing and using knowledge in rule-based knowledge base systems and it validates the proposed approach by evaluating an MCRDR-based KBS that is underpinned by blockchain technology and policies. Until now, blockchain has been used mainly for cryptocurrency and assets transfer. This research adapts and leverages blockchain to capitalise knowledge by providing a framework for rule-based knowledge base systems having the features of simultaneous knowledge acquisition, cross-platform (Ethereum and Hyperledger Fabric) communication and code generation by Domain Experts (DE) who do not necessarily have technical expertise. This research defines how these aspects of blockchain are well suited and advantageous when applied to an MCRDR KBS. This research was conducted in two phases. The first phase proposed a multi-blockchain based framework and methodology for rule-based knowledge base systems to store knowledge in the form of rules in a decentralised manner with role-based access to constrain rule maintenance to domain experts. A prototype system was implemented in Ethereum that leveraged an MCRDR based KBS implementation as it provided multiple classifications and can be readily adopted in different domains. Evaluation results, in terms of blockchain and KBS, showed similar accuracy in terms of the percentage of correct classifications provided when compared to traditional KBS. The second phase extended the framework by adding a consensus algorithm for concurrent rule addition by multiple domain experts while resolving any contention issues. To make the framework platform-agnostic, this phase also defined and implemented a programming API for double-encrypted cross-platform communication between Ethereum and the Hyperledger Fabric for sharing knowledge in the form of rules. Domain Experts are not necessarily equipped with technical knowledge to create smart contracts in multiple programming languages in standard blockchain and this could create significant hurdles in implementing their policies in the blockchain. Thus, the second phase also provided a modeldriven code conversion module to facilitate DEs to write XML-based contract requirements. These requirements are then converted into technology-specific languages for Ethereum and Hyperledger Fabric, Solidity and Golang respectively, and the results undergo verification before they are added to the blockchain. Each contribution in this phase was evaluated separately; the consensus algorithm was evaluated in Ethereum and Hyperledger Fabric, the API was evaluated with varying workloads for throughput and latency along with the effects of encryption, and the code conversion module was evaluated in terms of time and varying number of lines required in Solidity and Golang. To the best of knowledge, this is the first study to conduct initial research on a decentralised rule-based knowledge base system (RDR/MCRDR) with simultaneous knowledge acquisition, providing ground work for exploring this area in depth. The significant contributions to the body of knowledge made by this research are: 1- A framework has been proposed for decentralised capturing and maintaining knowledge in rule-based knowledge base systems. An MCRDR based KBS was adopted and evaluated to validate the proposed approach which provided more than 90% of correct classifications showing that knowledge can be captured and maintained in a tamper-proof decentralised way. 2- The definition and hybridisation of blockchain (heterogenous blockchain) based on seven algorithms to capitalise knowledge by storing and retrieving the rules for inference which are linked together but stored in different blocks unlike independent transactions in traditional blockchain based systems. 3- The definition and evaluation of a consensus algorithm for adding simultaneous knowledge in the form of rules by multiple domain experts and to solve any contention between domain experts for adding rules. 4- The definition and evaluation of a double encrypted communication model between Ethereum and Hyperledger Fabric for transfer and sharing of knowledge. Since both platforms have different transaction formats, this communication model proposes a unified transaction format for exchanging data and knowledge between the two platforms. 5- A novel methodology has been defined and evaluated for non-technical domain experts to write their own policies in different blockchain platforms (Ethereum and Hyperledger Fabric).



School of Information and Communication Technology

Publication status

  • Unpublished

Rights statement

Copyright 2022 the author

Repository Status

  • Restricted

Usage metrics

    Thesis collection


    No categories selected


    Ref. manager