Graph networks are extensively used as an essential framework to analyse the interconnections between transactions and capture illicit behaviour in Bitcoin blockchain. Due to the complexity of Bitcoin transaction graph, the prediction of illicit transactions has become a challenging problem to unveil illicit services over the network. Graph Convolutional Network, a graph neural network based spectral approach, has recently emerged and gained much attention regarding graph-structured data. Previous research has highlighted the degraded performance of the latter approach to predict illicit transactions using, a Bitcoin transaction graph, so-called Elliptic data derived from Bitcoin blockchain. Motivated by the previous work, we seek to explore graph convolutions in a novel way. For this purpose, we present a novel approach that is modelled using the existing Graph Convolutional Network intertwined with linear layers. Concisely, we concatenate node embeddings obtained from graph convolutional layers with a single hidden layer derived from the linear transformation of the node feature matrix and followed by Multi-layer Perceptron. Our approach is evaluated using Elliptic data, wherein efficient accuracy is yielded. The proposed approach outperforms the original work of same data set.

Elliptic data—one of the largest Bitcoin transaction graphs—has admitted promising results in many studies using classical supervised learning and graph convolutional network models for anti-money laundering. Despite the promising results provided by these studies, only few have considered the temporal information of this dataset, wherein the results were not very satisfactory. Moreover, there is very sparse existing literature that applies active learning to this type of blockchain dataset. In this paper, we develop a classification model that combines long-short-term memory with GCN—referred to as temporal-GCN—that classifies the illicit transactions of Elliptic data using its transaction’s features only. Subsequently, we present an active learning framework applied to the large-scale Bitcoin transaction graph dataset, unlike previous studies on this dataset. Uncertainties for active learning are obtained using Monte-Carlo dropout (MC-dropout) and Monte-Carlo based adversarial attack (MC-AA) which are Bayesian approximations. Active learning frameworks with these methods are compared using various acquisition functions that appeared in the literature. To the best of our knowledge, MC-AA method is the first time to be examined in the context of active learning. Our main finding is that temporal-GCN model has attained significant success in comparison to the previous studies with the same experimental settings on the same dataset. Moreover, we evaluate the performance of the provided acquisition functions using MC-AA and MC-dropout and compare the result against the baseline random sampling model.

With the advance of Bitcoin technology, money laundering has been incentivised as a den of Bitcoin blockchain, in which the user's identity is hidden behind a pseudonym known as address. Although this trait permits concealing in the plain sight, the public ledger of Bitcoin blockchain provides more power for investigators and allows collective intelligence for anti-money laundering and forensic analysis. This fascinating paradox arises in the strength of Bitcoin technology. Machine learning techniques have attained promising results in forensic analysis, in order to spot suspicious behaviour in Bitcoin blockchain. This paper presents a comparative analysis of the performance of classical supervised learning methods using a recently published data set derived from Bitcoin blockchain, to predict licit and illicit transactions in the network. Besides, an ensemble learning method is utilised using a combination of the given supervised learning models, which outperforms the given classical methods. This experiment is performed using a newly published data set derived from Bitcoin blockchain. Our main contribution points out that using ensemble learning approach outperforms the performance of the classical learning models used in the original paper, using Elliptic data set, a time series of Bitcoin transaction graph with node transactions and directed payments flow edges. Using the same data set, we show that we are able to predict licit/illicit transactions with an accuracy of 98.13% and F1 score equals to 83.36% using the proposed method. We discuss the variety of supervised learning methods, and their capabilities of assisting forensic analysis, and propose future work directions.

The past few years have witnessed the resurgence of uncertainty estimation generally in neural networks. Providing uncertainty quantification besides the predictive probability is desirable to reflect the degree of belief in the model’s decision about a given input. Recently, Monte-Carlo dropout (MC-dropout) method has been introduced as a probabilistic approach based Bayesian approximation which is computationally efficient than Bayesian neural networks. MC-dropout has revealed promising results on image datasets regarding uncertainty quantification. However, this method has been subjected to criticism regarding the behaviour of MC-dropout and what type of uncertainty it actually captures. For this purpose, we aim to discuss the behaviour of MC-dropout on classification tasks using synthetic and real data. We empirically explain different cases of MC-dropout that reflects the relative merits of this method. Our main finding is that MC-dropout captures datapoints lying on the decision boundary between the opposed classes using synthetic data. On the other hand, we apply MC-dropout method on dataset derived from Bitcoin known as Elliptic data to highlight the outperformance of model with MC-dropout over standard model. A conclusion and possible future directions are proposed.

We propose a novel method to capture data points near decision boundary in neural network that are often referred to a specific type of uncertainty. In our approach, we sought to perform uncertainty estimation based on the idea of adversarial attack method. In this paper, uncertainty estimates are derived from the input perturbations, unlike previous studies that provide perturbations on the model's parameters as in Bayesian approach. We are able to produce uncertainty with couple of perturbations on the inputs. Interestingly, we apply the proposed method to datasets derived from blockchain. We compare the performance of model uncertainty with the most recent uncertainty methods. We show that the proposed method has revealed a significant outperformance over other methods and provided less risk to capture model uncertainty in machine learning.