Browsing by Author "Mohammed, Kabiru"

Now showing 1 - 4 of 4
  • Thumbnail Image
    ItemOpen Access
    An Efficient Multidimensional k-Anonymisation Strategy Using Self-Organising Maps
    (De Montfort University, 2022-11) Mohammed, Kabiru
    Data mining techniques are highly efficient in sifting through big data to extract hidden knowledge and assist evidence-based decisions. The commercial benefits of these techniques have led to their successful adoption in different domains. However, there is an evidential concern that data mining could potentially be exploited to infer sensitive information, which raises a number of ethical issues, including those relating to privacy rights. Therefore, it is essential to enforce privacy constraints during mining processes in order to maintain a certain degree of privacy on the data to prevent inferences. As such, it is imperative to develop novel privacy techniques that can safeguard individual privacy during mining processes, while yet enabling the use of the data. This thesis centers primarily on privacy-preserving data mining, and specifically focuses on a proposed hybrid framework that combines data transformation methods in conjunction with anonymisation algorithms to derive more data utility during mining processes. The framework encompasses Self-Organising Maps for data transformation, two prominent clustering-based k-anonymisation algorithms to guarantee a certain degree of privacy, selective privacy and data quality metrics to validate our methods, and classification tasks to study the impact of our methods in data mining. The experiments of the study reveal that the proposed hybrid framework produces the most desirable properties for subsequent data mining. It is effective in meeting the desired privacy requirement and captures relevant patterns in microdata that are beneficial for classification tasks. In addition to this, the transformed data produced by self-organising maps conceals the input set into a 1-dimensional set of data, therefore preserving the true values of the original set. Results obtained from the experiments show that this unified approach has better overall performance in classification tasks than conventional methods. The outcome of the study concludes that anonymisation and data mining techniques are highly interdependent. Complicating their combination is the fact that both parties are attempting to achieve contradicting objectives. Therefore, the inclusion of additional techniques serves as a viable starting point for subsequent enhancements to promote compatibility between both parties. This would lead to reduced fluctuations in data mining performance and produce more consistent outcomes. The proposed model is effective in meeting this goal of adequately satisfying a privacy requirement while enhancing utility for subsequent data mining problems.
  • Thumbnail Image
    ItemOpen Access
    Complementing Privacy and Utility Trade-Off with Self-Organising Maps
    (MDPI, 2021-08-17) Mohammed, Kabiru; Ayesh, Aladdin; Boiten, Eerke Albert
    In recent years, data-enabled technologies have intensified the rate and scale at which organisations collect and analyse data. Data mining techniques are applied to realise the full potential of large-scale data analysis. These techniques are highly efficient in sifting through big data to extract hidden knowledge and assist evidence-based decisions, offering significant benefits to their adopters. However, this capability is constrained by important legal, ethical and reputational concerns. These concerns arise because they can be exploited to allow inferences to be made on sensitive data, thus posing severe threats to individuals’ privacy. Studies have shown Privacy-Preserving Data Mining (PPDM) can adequately address this privacy risk and permit knowledge extraction in mining processes. Several published works in this area have utilised clustering techniques to enforce anonymisation models on private data, which work by grouping the data into clusters using a quality measure and generalising the data in each group separately to achieve an anonymisation threshold. However, existing approaches do not work well with high-dimensional data, since it is difficult to develop good groupings without incurring excessive information loss. Our work aims to complement this balancing act by optimising utility in PPDMprocesses. To illustrate this, we propose a hybrid approach, that combines self-organising maps with conventional privacy-based clustering algorithms. We demonstrate through experimental evaluation, that results from our approach produce more utility for data mining tasks and outperforms conventional privacy-based clustering algorithms. This approach can significantly enable large-scale analysis of data in a privacy-preserving and trustworthy manner.
  • No Thumbnail Available
    ItemMetadata only
    Robustness of k-Anonymization Model in Compliance with General Data Protection Regulation
    (IEEE, 2023-03-28) Abubakar, Ibrahim Bio; Yagnik, Tarjana; Mohammed, Kabiru
    The advancement in technology and the emergence of big data and the internet of things (IoT), individuals (data subjects) tend to suffer from privacy breach of various types that has led to a lot of damages to both data subjects and brands. These and other issues about data privacy breach led the European Union to come up with a much stringent regulations that will serve as a deterrent to businesses or organizations that handle data. This gave birth to the General Data Protection Regulation (GDPR) in 2018 which replaced the previous 1995 Data Protection Directive in Europe. This research examined the robustness of k-anonymity in compliance with GDPR regulations at varying k-values (5,10,50, and 100) using the 1994 USA Census Bureau Data referred to as the adult dataset. Various measures were used to determine which k-value meets the GDPR criteria and the findings revealed the best anonymizing threshold complies with the GDPR criteria that prevents information loss (which determines data utility), prosecutor re-identification risk percentage and attacker models (prosecutor, journalist and marketer model).
  • Thumbnail Image
    ItemOpen Access
    Utility Promises of Self-Organising Maps in Privacy Preserving Data Mining
    (2020-09-14) Mohammed, Kabiru; Ayesh, Aladdin; Boiten, Eerke Albert
    Data mining techniques are highly efficient in sifting through big data to extract hidden knowledge and assist evidence-based decisions. However, it poses severe threats to individuals’ privacy because it can be exploited to allow inferences to be made on sensitive data. Researchers have proposed several privacy-preserving data mining techniques to address this challenge. One unique method is by extending anonymisation privacy models in data mining processes to enhance privacy and utility. Several published works in this area have utilised clustering techniques to enforce anonymisation models on private data, which work by grouping the data into clusters using a quality measure and then generalise the data in each group separately to achieve an anonymisation threshold. Although they are highly efficient and practical, however guaranteeing adequate balance between data utility and privacy protection remains a challenge. In addition to this, existing approaches do not work well with high-dimensional data, since it is difficult to develop good groupings without incurring excessive information loss. Our work aims to overcome these challenges by proposing a hybrid approach, combining self organising maps with conventional privacy based clustering algorithms. The main contribution of this paper is to show that, dimensionality reduction techniques can improve the anonymisation process by incurring less information loss, thus producing a more desirable balance between privacy and utility properties.