Mitigating significant risks pertaining to the implementation of cognitive computing

Van Wyk, Jana (2017-03)

Thesis (MAcc)--Stellenbosch University, 2017.

Thesis

ENGLISH SUMMARY : Cognitive computing is recognised as the third era in the evolution of computing. This era is driven by the exponential growth in data, advances in enabling technologies and enterprise’s need to realise significant business value from data resources. The capabilities of cognitive computing creates significant and immediate opportunities for enterprises. The problem is that management are implementing cognitive computing systems without understanding the technology or the risks the enterprise are exposed. The aim of the research is therefore to identify and mitigate significant risks pertaining to the implementation of cognitive computing. The research aims to investigate cognitive computing, identify significant risks and recommend safeguards to mitigate these risks. A literature review was performed to provide a theoretical foundation for the research and focussed on cognitive computing, corporate governance, IT governance, data protection and the use of control frameworks to achieve effective governance. COBIT 5 was selected as the most appropriate control framework to identify significant risk. In order to identify the risks the core components of a cognitive computing system were identified and classified into specific phases based on their function within the cognitive computing system The research found that a cognitive computing system consists of consist of twelve core components and four phases The core components include: unstructured, semi-structured and structured data; data access, metadata, feature extraction, natural language processing and deep learning; corpus and advances analytics; and hypothesis generation and scoring, and machine learning. Based on the understanding of the core components, COBIT 5 was used to identify significant risks. Significant risks were identified at a strategic and operational or technological level. Risks at a strategic level involved inadequate governance and management, as well as insufficient human skills and resource management. Significant risks at an operational or technological level comprised of cost, privacy, security, scalability, integration, interoperability, veracity, ownership and life cycle risks. The research proceeded to formulate appropriate internal control techniques to mitigate the significant risks identified. The internal control techniques include establishing a cognitive computing strategies and policies, implementing human skills and resource controls, data controls, infrastructure controls, supplier controls and life cycle controls. The final product of the findings is a risk matrix, which maps the relevant core components with the significant risk which they introduce and a risk control matrix which maps the risk to the control technique which mitigates the risk.

AFRIKAANSE OPSOMMING : Kognitiewe verwerking (cognitive computing) word erken as die derde era in die evolusie van rekenaar verwerking (computing). Die era word gedryf deur die eksponensiele groei van data, die verbetering van bemagtigende tegnologiee en ondernemings se behoefte om beduidende besigheidswaarde uit data bronne te realiseer. Kognitiewe verwerking is in staat om beduidende en onmiddellike geleenthede vir ondernemings te skep. Die probleem is egter dat bestuur kognitiewe verwerking stelsels (cognitive computing systems) implementeer sonder dat hulle die tegnologie of die risiko’s waaraan die onderneming blootgestel word verstaan. Die doel van die navorsing is dus om wesenlike risiko’s in verband met die implementering van kognitiewe verwerking te identifiseer en aan te spreek. Die navorsing beoog om ‘n dieper begrip te ontwikkel van kognitiewe verwerking, om wesenlike risiko’s te identifiseer en om kontroles aan te beveel wat die risiko’s aanspreek. ‘n Literatuur studie was uitgevoer om ‘n teoretiese basis vir die navorsing te bied en het gefokus op kognitiewe verwerking, korporatiewe beheer, IT beheer, data beskerming en die gebruik van kontrole raamwerke om effektiewe beheer te bewerkstellig. COBIT 5 was geselekteer as die beste kontrole raamwerk om wesenlike risiko’s te identifiseer. Om die risiko’s te identifiseer is die onderliggende komponente van ‘n kognitiewe verwerking stelsel geidentifiseer en geklassifiseer gebaseer op hul funksies in die kognitiewe verwerking stelsel. Die navorsing het gevind dat ‘n kognitiewe verwerking stelsel uit twaalf onderliggende komponente en vier fases bestaan. Die onderliggende komponente sluit in: ongestruktureerde, semigestruktureerde en gestruktureerde data; data toegang, metadata, kenmerk ontrekking (feature extraction), natuurliketaalverwerking (natural language processing) en dieper leer (deep learning); korpus (corpus) en gevorderde ontleding (advances analytics); en hipotese generering en meting (hypothesis generation and scoring), en masjien leer (machine learning). COBIT 5 is gebruik om wesenlike risiko’s te identifiseer, gebaseer op kennis van die onderliggende komponente. Wesenlike risiko’s is op ‘n strategiese en operasionele of tegnologiese geidentifiseer. Risiko’s op ‘n strategiese vlak sluit in onvoldoende beheer en bestuur, sowel as onvoldoende menslike hulpbron bestuur. Wesenlike risiko’s op ‘n operasionele of tegnologiese vlak bestaan uit koste, privaatheid, sekuriteit, skaalbaarheid (scalability), integrasie, interoperasionaliteit (interoperability), geldigheid (veracity), data eienaarskap en lewenssiklus risiko’s. Die navorsing het voortgegaan om interne beheermaatreels te formuleer om die wesenlike risiko’s aan te spreek. Die interne beheermaatreels sluit in die vestiging van kognitiewe verwerking strategiee en beleide, die implementering van menslike hulpbron kontroles, data kontroles, infrastruktuur kontroles, verskaffer kontroles en lewenssiklus kontroles. Die finale produk van die bevindinge is ‘n risiko matriks, wat die relevante onderliggende komponente verbind met die wesenlike risiko’s wat hulle skep en ‘n risiko-kontrole matriks wat die risiko’s verbind met die beermaatreels wat die risiko’s aan spreek.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/101314
This item appears in the following collections: