Enhancing the security of financial transactions in Blockchain by using machine learning techniques: towards a sophisticated security tool for banking and finance
Dalila Boughaci, Abdullah A. K. Alkhawaldeh
Abstract
Blockchain is an innovative technology that can be used for processing and sharing data in a safe manner, via an unreliable network. This paper starts with a background of blockchains. We give some principal concepts of this new technology. Then we deal with machine learning as intelligent techniques that may be used for analyzing huge datasets and detecting eventual malicious transactions that can be occured in the untrusted network. We show the importance of such intelligent techniques when combining together to make good decisions in banking and finance. Our idea is applied to the Bitcoin system where the public Elliptic dataset from Kaggle is used as a benchmark. Since the latter is not all labeled, we use the k-means algorithm to partition the unlabeled data into two main clusters while the labeled data are moved to their corresponding clusters. Then, four machine learning techniques are used to classify all the data. The proposed system shows promising results in particular when combining k-means with the random forest classifier.