Analyzing Malicious Activities and Detecting Adversarial Behavior in Cryptocurrency based Permissionless Blockchains: An Ethereum Usecase

Abstract

Different malicious activities occur in cryptocurrency-based permissionless blockchains such as Ethereum and Bitcoin. Some activities are due to the exploitation of vulnerabilities which are present in the blockchain infrastructure, some activities target its users through social engineering techniques, while some activities use it to facilitate different malicious activities. Since cryptocurrency-based permissionless blockchains provide pseudonymity to its users, bad actors prefer to carry out transactions related to malicious activities on them. Towards this, we aim at automatically flagging blockchain accounts as suspects that indulge in malicious activities, thus reducing the unintended support that cryptocurrency-based permissionless blockchains provide to malicious actors. We first use the cosine similarity (CS) metrics to study the similarities between the feature vector of accounts associated with different malicious activities and find that most of the malicious activities associated with the Ethereum blockchain behave similarly. We then use the K-Means clustering algorithm to check if accounts associated with similar malicious activities cluster together. We also study the effect of bias on the performance of a machine learning (ML) algorithm, due to the number of accounts associated with malicious activity. We then compare the different state-of-the-art models and identify that Neural Networks (NNs) are resistant to bias associated with a malicious activity and are also robust against adversarial attacks. The previously used ML algorithms for identifying malicious accounts also show bias towards an over-represented malicious activity.

Anomaly Detection in Blockchain Networks Using Unsupervised Learning: A Survey

“…might impose strong capabilities in detecting/predicting malicious behaviors. Temporal features can be extracted from graph-based [175,190] or tabular-based [191] representation of the Ethereum data. In [175], the temporal features were transformed into vectors to provide a time series facet based on the assumption that this choice can assist the detection of malicious accounts considering past attacks analysis.…”

“…The approach followed the above-mentioned AutoML-based synergy between supervised learning and k-Means. In [191,192], several attacks on the Ethereum (e.g., ransomware payments, phishing, scamming, upbit hack, spam token, Ponzi schemes, EtherDelta Hack, etc.) were considered, while addressing issues related to the effective application of k-Means and GANs, where their results were preprocessed by various supervised ML structures such as the Extra-Tree classifier and a neural network in the form of Multiple-Layer Perceptron (MLP).…”

A New Framework for Fraud Detection in Bitcoin Transactions Through Ensemble Stacking Model in Smart Cities

“…The authors proposed a novel approach using ML for detecting malicious activities and adversarial behavior in permission-less blockchains in [15]. neural Network (NN) can acquire large recall value and detect adversarial feature vectors.…”

“…The existing techniques for detecting frauds on the Bitcoin network have drawbacks (i.e., misclassification) Proposed a collective anomaly detection technique [13] F1-score, Accuracy, Precision, and Recall Existing techniques are not ideal for fraud detection Used RF using K fold method [14] Accuracy, F1-score Scalability issue in existing techniques Used K mean clustering and NN [15] F1-score and Accuracy Rule-based techniques do not detect new and unseen fraud Used KNN, SVM and Isolation Forest [16] Accuracy, Precision and Recall Data imbalance issue and ML techniques are not a good choice for big data as it leads to overfitting Used Random Oversampling Used GAN to improve classification accuracy [17] Accuracy, Precision, Recall, F1-score and AUC Existing techniques do not employ deep data analysis TDA is used [18] Accuracy, Precision and Recall Data imbalance issue and existing techniques do not detect fraud accurately SMOTE and GRU are used [20] ROC-AUC and PR-AUC Existing techniques are not good choice for big data as they lead to overfitting Used SMOTE and ML techniques (DT, RF, KNN and SVM) [21] Accuracy, Precision, Recall and F1-score Existing techniques are bit challenging and complex for handling big data Used SNN and optimizable DT [22] Accuracy, Precision, Recall, AUC, and F1-score Data imbalance problem Used random undersampling and corelation for feature extraction [23] Accuracy and F1-score…”

Analyzing the Threats to Blockchain-Based Self-Sovereign Identities by Conducting a Literature Survey

“…Dabrowski et al [40] suggest that hardware wallets should be inspected for the packaging, comparing the printed circuit board (PCB) with the reference picture, and verifying the software with hashes or signatures. Agarwal et al [96] detect phishing, spam, and scams by applying Etherscan, a block explorer and analytics platform for Ethereum. Apostolaki et al [86] focus on the countermeasures of deanonymization with encrypted traffic, fake peers, obfuscating the client's state, routing-aware transactions' requests and advertisements, and applying Tor or a virtual private network (VPN).…”