In the context of the rapid development of blockchain technology, smart contracts have also been widely used in the Internet of Things, finance, healthcare, and other fields. There has been an explosion in the number of smart contracts, and at the same time, the security of smart contracts has received widespread attention because of the financial losses caused by smart contract vulnerabilities. Existing analysis tools can detect many smart contract security vulnerabilities, but because they rely too heavily on hard rules defined by experts when detecting smart contract vulnerabilities, the time to perform the detection increases significantly as the complexity of the smart contract increases. In the present study, we propose a novel hybrid deep learning model named CBGRU that strategically combines different word embedding (Word2Vec, FastText) with different deep learning methods (LSTM, GRU, BiLSTM, CNN, BiGRU). The model extracts features through different deep learning models and combine these features for smart contract vulnerability detection. On the currently publicly available dataset SmartBugs Dataset-Wild, we demonstrate that the CBGRU hybrid model has great smart contract vulnerability detection performance through a series of experiments. By comparing the performance of the proposed model with that of past studies, the CBGRU model has better smart contract vulnerability detection performance.
Rabia Musheer Aziz, Mohammed Farhan Baluch, Sarthak Patel, Pavan Kumar
Ethereum smart contracts have recently received new commercial applications and a lot of attention from the scientific community. Ethereum eliminates the requirement for a trusted third party by allowing untrusted parties to expose contract details in computer code. Nonetheless, as online commerce grows, plenty of fraudulent activities, such as money laundering, bribery, and phishing, emerge as major threats to trade security. For correctly recognizing fraudulent transactions, this paper developed a Light Gradient Boosting Machine (LGBM) technique-based model. The modified LGBM model optimized the parameters of Light GBM using the Euclidean distant structured estimation approach. This paper also examines the performance of different popular models such as Random Forest (RF), Multi-Layer Perceptron (MLP), Logistic Regression, k-Nearest Neighbors (KNN), XGBoost, Support Vector Classification (SVC), and ADAboost with limited features and compares their performance metrics with the proposed model for Ethereum fraudulent activity classification. A comparative performance evaluation matrices scores of different popular models along with the proposed model demonstrated the applicability of the proposed approach. The modified LGBM algorithms and RF models demonstrate the best performance compared to other models with the highest accuracies, while the modified LGBM algorithm has a slightly superior performance of 99.17 percent compared to the RF model's 98.26 percent.
Cryptocurrency blockchain data encounter a class-imbalance problem due to only a few known labels of illicit or fraudulent activities in the blockchain network. For this purpose, we seek to compare various resampling methods applied to two highly imbalanced datasets derived from the blockchain of Bitcoin and Ethereum after further dimensionality reductions, which is different from previous studies on these datasets. Firstly, we study the performance of various classical supervised learning methods to classify illicit transactions or accounts on Bitcoin or Ethereum datasets, respectively. Consequently, we apply various resampling techniques to these datasets using the best performing learning algorithm on each of these datasets. Subsequently, we study the feature importance of the given models, wherein the resampled datasets directly influenced on the explainability of the model. Our main finding is that undersampling using the edited nearest-neighbour technique has attained an accuracy of more than 99% on the given datasets by removing the noisy data points from the whole dataset. Moreover, the best-performing learning algorithms have shown superior performance after feature reduction on these datasets in comparison to their original studies. The matchless contribution lies in discussing the effect of the data resampling on feature importance which is interconnected with explainable artificial intelligence (XAI) techniques.
In recent years, phishing scams have become the most serious type of crime involved in Ethereum, the second-largest blockchain platform. The existing phishing scams detection technology on Ethereum mostly uses traditional machine learning or network representation learning to mine the key information from the transaction network to identify phishing addresses. However, these methods adopt the last transaction record or even completely ignore these records, and only manual-designed features are taken for the node representation. In this paper, we propose a Temporal Transaction Aggregation Graph Network (TTAGN) to enhance phishing scams detection performance on Ethereum. Specifically, in the temporal edges representation module, we model the temporal relationship of historical transaction records between nodes to construct the edge representation of the Ethereum transaction network. Moreover, the edge representations around the node are aggregated to fuse topological interactive relationships into its representation, also named as trading features, in the edge2node module. We further combine trading features with common statistical and structural features obtained by graph neural networks to identify phishing addresses. Evaluated on real-world Ethereum phishing scams datasets, our TTAGN (92.8% AUC, and 81.6% F1-score) outperforms the state-of-the-art methods, and the effectiveness of temporal edges representation and edge2node module is also demonstrated.
While blockchain technology triggers new industrial and technological revolutions, it also brings new challenges. Recently, a large number of new scams with a "blockchain" sock-puppet continue to emerge, such as Ponzi schemes, money laundering, etc., seriously threatening financial security. Existing fraud detection methods in blockchain mainly concentrate on manual feature and graph analytics, which first construct a homogeneous transaction graph using partial blockchain data and then use graph analytics to detect anomaly, resulting in a loss of pattern information. In this paper, we mainly focus on Ponzi scheme detection and propose HFAug, a generic Heterogeneous Feature Augmentation module that can capture the heterogeneous information associated with account behavior patterns and can be combined with existing Ponzi detection methods. HFAug learns the metapath-based behavior characteristics in an auxiliary heterogeneous interaction graph, and aggregates the heterogeneous features to corresponding account nodes in the homogeneous one where the Ponzi detection methods are performed. Comprehensive experimental results demonstrate that our HFAug can help existing Ponzi detection methods achieve significant performance improvement on Ethereum datasets, suggesting the effectiveness of heterogeneous information on detecting Ponzi schemes.
Cryptocurrencies have dramatically increased adoption in mainstream applications in various fields such as financial and online services, however, there are still a few amounts of cryptocurrency transactions that involve illicit or criminal activities. It is essential to identify and monitor addresses associated with illegal behaviors to ensure the security and stability of the cryptocurrency ecosystem. In this paper, we propose a framework to build a dataset comprising Bitcoin transactions between 12 July 2019 and 26 May 2021. This dataset (hereafter referred to as BABD-13) contains 13 types of Bitcoin addresses, 5 categories of indicators with 148 features, and 544,462 labeled data, which is the largest labeled Bitcoin address behavior dataset publicly available to our knowledge. We also propose a novel and efficient subgraph generation algorithm called BTC-SubGen to extract a${k}$-hop subgraph from the entire Bitcoin transaction graph constructed by the directed heterogeneous multigraph starting from a specific Bitcoin address node. We then conduct 13-class classification tasks on BABD-13 by five machine learning models namely${k}$-nearest neighbors algorithm, decision tree, random forest, multilayer perceptron, and XGBoost, the results show that the accuracy rates are between 93.24% and 97.13%. In addition, we study the relations and importance of the proposed features and analyze how they affect the effect of machine learning models. Finally, we conduct a preliminary analysis of the behavior patterns of different types of Bitcoin addresses using concrete features and find several meaningful and explainable modes.
Francesco Zola, Lander Segurola-Gil, Jan L. Bruse, Mikel Galar Ā· 5 authors
Abstract Classification of Bitcoin entities is an important task to help Law Enforcement Agencies reduce anonymity in the Bitcoin blockchain network and to detect classes more tied to illegal activities. However, this task is strongly conditioned by a severe class imbalance in Bitcoin datasets. Existing approaches for addressing the class imbalance problem can be improved considering generative adversarial networks (GANs) that can boost data diversity. However, GANs are mainly applied in computer vision and natural language processing tasks, but not in Bitcoin entity behaviour classification where they may be useful for learning and generating synthetic behaviours. Therefore, in this work, we present a novel approach to address the class imbalance in Bitcoin entity classification by applying GANs. In particular, three GAN architectures were implemented and compared in order to find the most suitable architecture for generating Bitcoin entity behaviours. More specifically, GANs were used to address the Bitcoin imbalance problem by generating synthetic data of the less represented classes before training the final entity classifier. The results were used to evaluate the capabilities of the different GAN architectures in terms of training time, performance, repeatability, and computational costs. Finally, the results achieved by the proposed GAN-based resampling were compared with those obtained using five well-known data-level preprocessing techniques. Models trained with data resampled with our GAN-based approach achieved the highest accuracy improvements and were among the best in terms of precision, recall and f1-score. Together with Random Oversampling (ROS), GANs proved to be strong contenders in addressing Bitcoin class imbalance and consequently in reducing Bitcoin entity anonymity (overall and per-class classification performance). To the best of our knowledge, this is the first work to explore the advantages and limitations of GANs in generating specific Bitcoin data and āattackingā Bitcoin anonymity. The proposed methods ultimately demonstrate that in Bitcoin applications, GANs are indeed able to learn the data distribution and generate new samples starting from a very limited class representation, which leads to better detection of classes related to illegal activities.
In recent years, blockchain technology has been developing rapidly. More and more traditional industries are using blockchain as a platform for information storage and financial transactions, mainly because of its new characteristics of non-tamperability and decentralization compared with the traditional systems. As a representative of blockchain 2.0, Ethereum has gained popularity upon its introduction. However, because of the anonymity of blockchain, Ethereum has also attracted the attention of some unscrupulous people. Currently, millions of contracts are deployed on Ethereum, many of which are fraudulent contracts deployed by unscrupulous people for profit, and these contracts are causing huge losses to investors worldwide. Ponzi contracts are typical of these contracts, which mainly reward the funds invested by later investors to early investors, and later investors will have no gain. However, although there are some studies for identifying Ponzi contracts on Ethereum, there is some room for progress in the research. Therefore, we propose a method to detect Ponzi scheme contracts on Ethereum-CTRF. This method forms a dataset by extracting the word features and sequence features of the smart contractās code and the features of transactions. The dataset is divided into a training set and a test set. Oversampling is performed on the training set to deal with the problem of positive and negative sample imbalance. Finally, the model is trained on the training set and tested on the test set. The experimental results show that the model has significantly improved recall compared with existing Ponzi contract detection methods.
While transactions with cryptocurrencies such as Ethereum are becoming more prevalent, fraud and other criminal transactions are not uncommon. Graph analysis algorithms and machine learning techniques detect suspicious transactions that lead to phishing in large transaction networks. Many graph neural network (GNN) models have been proposed to apply deep learning techniques to graph structures. Although there is research on phishing detection using GNN models in the Ethereum transaction network, models that address the scale of the number of vertices and edges and the imbalance of labels have not yet been studied. In this paper, we compared the model performance of GNN models on the actual Ethereum transaction network dataset and phishing reported label data to exhaustively compare and verify which GNN models and hyperparameters produce the best accuracy. Specifically, we evaluated the model performance of representative homogeneous GNN models which consider single-type nodes and edges and heterogeneous GNN models which support different types of nodes and edges. We showed that heterogeneous models had better model performance than homogeneous models. In particular, the RGCN model achieved the best performance in the overall metrics.
Fraud prevention in cryptocurrency transactions is paramount because the proportion of fraudulent transactions in this sector is increasing.Artificial intelligence (AI) can increase the ability to identify fraudulent transactions through patterns and irregularities that are hardly noticeable through other approaches.This paper focuses on AI approaches, such as machine learning algorithms, LightGBM, and fraud detection.Some notable works are comparing AI strategies, methods to overcome difficulties when using imbalanced data sets, and the practical utility of these models.
The key issue in the field of smart contract security is efficient and rapid vulnerability detection in smart contracts. Most of the existing detection methods can only detect the presence of vulnerabilities in the contract and can hardly identify their type. Furthermore, they have poor scalability. To resolve these issues, in this study, we developed a smart contract vulnerability detection model based on multi-task learning. By setting auxiliary tasks to learn more directional vulnerability features, the detection capability of the model was improved to realize the detection and recognition of vulnerabilities. The model is based on a hard-sharing design, which consists of two parts. First, the bottom sharing layer is mainly used to learn the semantic information of the input contract. The text representation is first transformed into a new vector by word and positional embedding, and then the neural network, based on an attention mechanism, is used to learn and extract the feature vector of the contract. Second, the task-specific layer is mainly employed to realize the functions of each task. A classical convolutional neural network was used to construct a classification model for each task that learns and extracts features from the shared layer for training to achieve their respective task objectives. The experimental results show that the model can better identify the types of vulnerabilities after adding the auxiliary vulnerability detection task. This model realizes the detection of vulnerabilities and recognizes three types of vulnerabilities. The multi-task model was observed to perform better and is less expensive than a single-task model in terms of time, computation, and storage.
Manuel FebreroāBande, Wenceslao GonzĆ”lezāManteiga, Brenda Prallon, Yuri F. Saporito
This paper proposes a classification model for predicting the main activity of bitcoin addresses based on their balances. Since the balances are functions of time, we apply methods from functional data analysis; more specifically, the features of the proposed classification model are the functional principal components of the data. Classifying bitcoin addresses is a relevant problem for two main reasons: to understand the composition of the bitcoin market, and to identify addresses used for illicit activities. Although other bitcoin classifiers have been proposed, they focus primarily on network analysis rather than curve behavior. Our approach, on the other hand, does not require any network information for prediction. Furthermore, functional features have the advantage of being straightforward to build, unlike expert-built features. Results show improvement when combining functional features with scalar features, and similar accuracy for the models using those features separately, which points to the functional model being a good alternative when domain-specific knowledge is not available.
Jan 1, 2022Ā·Proceedings of the ... Annual Hawaii International Conference on System Sciences/Proceedings of the Annual Hawaii International Conference on System Sciences
Over 3,739 apps on average are published per day on the Google Play store [1]. A handful of the applications contain advertisement malware referred to as malvertising. As a result, Android advertisement malware has been a growing multi-billion-dollar problem. It constantly assaults many of the major advertising libraries such as the Google, Facebook, and Amazon. This paper presents an effective strategy for countering advertising malware using dynamic and static analysis techniques and the Soot compiler framework. Our research aims to detect malvertising click fraud in Android applications using the Soot compiler framework and blockchain technology. But the approach and the framework can be used to counter mobile malware families.
Blockchain technology, smart contracts, and asset tokenization have relevant implications for the auditing environment. This paper evaluates the current stage of blockchain application in auditing, analyzing scientific publications and identifying the impact of what is already a reality and the potential effects of its improvements in audit professionalsā activities performance. The article considers the proposals and suggestions on the leading research indexed by the Scopus and Web of Science databases. We analyzed 374 papers on the topic of blockchain and provide a summary and analysis of the current state of auditing research. The bibliometric analysis was performed using the Bibliometrix R Package and the VOSviewer software. After a systematic study of abstracts and a general review of the papers to only include those directly related to our workās objectives, we found 78 papers. The work results in a framework of potential and effective implications of blockchain technology for auditing, pointing out several new challenges in terms of skills and knowledge needed in this new reality of audit professionals.
Tahmid Hasan Pranto, Kazi Tamzid Akhter Md Hasib, Tahsinur Rahman, AKM Bahalul Haque Ā· 6 authors
Financial fraud cases are on the rise even with the current technological advancements. Due to the lack of inter-organization synergy and because of privacy concerns, authentic financial transaction data is rarely available. On the other hand, data-driven technologies like machine learning need authentic data to perform precisely in real-world systems. This study proposes a blockchain and smart contract-based approach to achieve robust Machine Learning (ML) algorithm for e-commerce fraud detection by facilitating inter-organizational collaboration. The proposed method uses blockchain to secure the privacy of the data. Smart contract deployed inside the network fully automates the system. An ML model is incrementally upgraded from collaborative data provided by the organizations connected to the blockchain. To incentivize the organizations, we have introduced an incentive mechanism that is adaptive to the difficulty level in updating a model. The organizations receive incentives based on the difficulty faced in updating the ML model. A mining criterion has been proposed to mine the block efficiently. And finally, the blockchain network is tested under different difficulty levels and under different volumes of data to test its efficiency. The model achieved 98.93% testing accuracy and 98.22% Fbeta score (recall-biased f measure) over eight incremental updates. Our experiment shows that both data volume and difficulty level of blockchain impacts the mining time. For difficulty level less than five, mining time and difficulty level has a positive correlation. For difficulty level two and three, less than a second is required to mine a block in our system. Difficulty level five poses much more difficulties to mine the blocks.
Nowadays, health insurance has become an essential part of people’s lives as the number of health issues increases. Healthcare emergencies can be troublesome for people who can’t afford huge expenses. Health insurance helps people cover healthcare services expenses in case of a medical emergency and provides financial backup against indebtedness risk. Health insurance and its several benefits can face many security, privacy, and fraud issues. For the past few years, fraud has been a sensitive issue in the health insurance domain as it incurs high losses for individuals, private firms, and governments. So, it is essential for national authorities and private firms to develop systems to detect fraudulent cases and payments. A high volume of health insurance data in electronic form is generated, which is highly sensitive and attracts malicious users. Motivated by these facts, we present a systematic survey for Artificial Intelligence (AI) and blockchain-enabled secure health insurance fraud detection in this paper. This paper presents a taxonomy of various security issues in health insurance. We proposed a blockchain and AI-based secure and intelligent system to detect health insurance fraud. Then a case study related to health insurance fraud is presented. Finally, the open issues and research challenges in implementing the blockchain and an AI-empowered health insurance fraud detection system is presented.
Cryptocurrencies continue to captivate businesses and investors despite market fluctuations. The number of crypto users have risen rapidly in the last few years, and alarmingly, many appear to be unaware of the risks involved. These risks aren't confined to market hazards but include very sophisticated cybercrimes related to cryptocurrencies. As cryptocurrencies have become a breeding ground for a variety of cybercrimes, resulting in enormous financial losses, it hinders user adoption limiting the utility of the blockchain technology. It has become crucial to spot such scams and devise intelligent techniques to make this technology a safer place for investors. This study proposes a classification model to handle fake account problem over Ethereum blockchain, and its contribution is multi-faceted; firstly, available imbalanced Ethereum dataset has been balanced to enhance the accuracy of the classification model. Secondly, correlation-based feature selection technique has been applied to retain best discriminating features. Thirdly, an effective machine learning based model has been presented for the identification of fake accounts over the Ethereum system. A comparative study of ten machine learning techniques has been presented consisting of both individual and ensemble classifiers. Experimental results showed that ensemble classifiers appear to yield better performance measures over individual classifiers and among all, LightGbm-based classification model outperformed with 99.2% accuracy.
Blockchain technology supports the generation and record of transactions, and maintains the fairness and openness of the cryptocurrency system. However, many fraudsters utilize smart contracts to create fraudulent Ponzi schemes for profiting on Ethereum, which seriously affects financial security. Most existing Ponzi scheme detection techniques suffer from two major restricted problems: the lack of motivation for temporal early warning and failure to fuse multi-source information finally cause the lagging and unsatisfactory performance of Ethereum Ponzi scheme detection. In this paper, we propose a dual-channel early warning framework for Ethereum Ponzi schemes, named Ponzi-Warning, which performs feature extraction and fusion on both code and transaction levels. Moreover, we represent a temporal evolution augmentation strategy for generating transaction graph sequences, which can effectively increase the data scale and introduce temporal information. Comprehensive experiments on our Ponzi scheme datasets demonstrate the effectiveness and timeliness of our framework for detecting the Ponzi contract accounts.
Blockchain has widespread applications in the financial field but has also attracted increasing cybercrimes. Recently, phishing fraud has emerged as a major threat to blockchain security, calling for the development of effective regulatory strategies. Nowadays network science has been widely used in modeling Ethereum transaction data, further introducing the network representation learning technology to analyze the transaction patterns. In this paper, we consider phishing detection as a graph classification task and propose an end-to-end Phishing Detection Graph Neural Network framework (PDGNN). Specifically, we first construct a lightweight Ethereum transaction network and extract transaction subgraphs of collected phishing accounts. Then we propose an end-to-end detection model based on Chebyshev-GCN to precisely distinguish between normal and phishing accounts. Extensive experiments on five Ethereum datasets demonstrate that our PDGNN significantly outperforms general phishing detection methods and scales well in large transaction networks.
In privacy-preserving machine learning, it is less obvious to verify correct behavior of participants because they are not supposed to reveal their inputs in cleartext to other participants. It is hence important to make federated machine learning robust against data poisoning and related attacks. While input data can be related to a distributed ledger (blockchain), a less studied input is formed by the random sampling parties perform. In this paper, we describe strategies based on zero knowledge proofs to allow parties to prove they perform sampling (and other computations) correctly. We sketch a number of alternative ways to implement our idea and provide some preliminary experimental results.
Abstract Decentralized Finance (DeFi) is a system of financial products and services built and delivered through smart contracts on various blockchains. In recent years, DeFi has gained popularity and market capitalization. However, it has also been connected to crime, particularly various types of securities violations. The lack of Know Your Customer requirements in DeFi poses challenges for governments trying to mitigate potential offenses. This study aims to determine whether this problem is suited to a machine learning approach, namely, whether we can identify DeFi projects potentially engaging in securities violations based on their tokensā smart contract code. We adapted prior works on detecting specific types of securities violations across Ethereum by building classifiers based on features extracted from DeFi projectsā tokensā smart contract code (specifically, opcode-based features). Our final model was a random forest model that achieved an 80% F-1 score against a baseline of 50%. Notably, we further explored the code-based features that are the most important to our modelās performance in more detail by analyzing tokensā Solidity code and conducting cosine similarity analyses. We found that one element of the code that our opcode-based features can capture is the implementation of the SafeMath library, although this does not account for the entirety of our features. Another contribution of our study is a new dataset, comprising (a) a verified ground truth dataset for tokens involved in securities violations and (b) a set of legitimate tokens from a reputable DeFi aggregator. This paper further discusses the potential use of a model like ours by prosecutors in enforcement efforts and connects it to a wider legal context.
The manuscript presents a study of the possibility of use of Benfordās law conformity test, a well proven tool in the accounting fraud discovery, on a new domain: the discovery of anomalies (possibly fraudulent behaviour) in the the cryptocurrency transactions. Blockchain-based currencies or cryptocurrencies have become a global phenomenon known to most people as a disruptive technology, and a new investment vehicle. However, due to their decentralized nature, regulating these markets has presented regulators with difficulties in finding a balance between nurturing innovation, and protecting consumers. The growing concerns about illicit activity have forced regulators to seek new ways of detecting, analyzing, and ultimately policing public blockchain transactions. Extensive research on machine learning, and transaction graph analysis algorithms has been done to track suspicious behaviour. However, having a macro view of a public ledger is equally important before pursuing a more fine-grained analysis. Benfordās law, the law of first digit, has been extensively used as a tool to discover accountant frauds (many other use cases exist). The basic motivation that drove our research presented in this paper was to test the applicability of the well established method to a new domain, in this case the identification of anomalous behavior using Benfordās law conformity test to the cryptocurrency domain. The research focused on transaction values in all major cryptocurrencies. A suitable time-period was identified that was long enough to present sufficiently large number of observations for Benfordās law conformity tests and was also situated long enough in the past so that the anomalies were identified and well documented. The results show that most of the cryptocurrencies that did not conform to Benfordās law had well documented anomalous incidents, the first digits of aggregated transaction values of all well known cryptocurrency projects were conforming to Benfordās law. Thus the proposed method is applicable to the new domain.