Sr IT Developer, First Horizon Bank, Memphis, TN, USA, Rushikesh Anantrao Deshpande
The article examines methods for optimizing PL/SQL queries in distributed banking databases, emphasizing the transition from static rule-based mechanisms to adaptive, learning-driven architectures. The study’s relevance is defined by the increasing complexity of financial data environments that require real-time consistency, fault tolerance, and intelligent workload distribution. The research synthesizes results from seven recent works published between 2021 and 2025, covering neural cost modeling, heuristic algorithms, hybrid plan enumeration, and visualization-based diagnostics. Special attention is devoted to learned cost models and metaheuristic strategies that enhance selectivity estimation, reduce latency, and stabilize throughput in distributed ledger systems. The methodological framework integrates comparative analysis, systematization, and critical evaluation of hybrid, heuristic, and learning-based optimizers. The findings reveal a multi-layered optimization model that combines probabilistic inference, robust plan selection, and heuristic refinement. The conclusions underscore the practical applicability of adaptive PL/SQL optimization for high-volume banking infrastructures and data-intensive financial analytics.
Hengxin Lei, Thein Lai Wong, Tong Ming Lim, XiangFu Zhao · 6 authors
Smart contracts manage billions of dollars' worth of digital assets. Once vulnerabilities are exploited, they may lead to fund theft, transaction rollback, or asset freezing. The current machine learning based smart contract vulnerability detection methods have poor performance and high time complexity in detecting sparse labeled data. We propose a smart contract vulnerability detection model for supervised Bigram-Principle Component Analysis (Bi-PCA). Supervising Bi-PCA can utilize labeled vulnerability data for supervised dimensionality reduction. The supervised Bi-PCA model can utilize the information from these additional labels to accurately extract more interpretable potential structures. This detection model is universal and can be used in industrial scenarios on all labeled datasets. The experimental results show that it has high accuracy and recall while maintaining the feature of the original label data.
Sheikh Sharfuddin Mim, Doina Logofătu, Gabriel Guerrero-Contreras, Inmaculada Medina‐Bulo
Federated learning (FL) offers a compelling solution to the privacy and compliance challenges that plague the financial industry by enabling decentralized machine learning without the need for raw data sharing. As regulations like GDPR and the GLBA enforce strict data protection requirements, financial institutions are increasingly exploring FL as an avenue for collaborative intelligence. This paper presents a comprehensive analysis of state-of-the-art FL algorithms tailored for finance, evaluates their performance across realistic tasks such as credit scoring, fraud detection, and customer segmentation, and identifies the trade-offs among performance, fairness, privacy, and communication cost. We benchmark twelve prominent FL algorithms, highlight their privacy implications with differential privacy and secure aggregation, and provide practical, strategic recommendations for deploying FL systems in financial environments. The study closes with a discussion on regulatory alignment, deployment challenges, and future directions for ethical and robust FL adoption in finance.
Smart contracts have emerged as a core innovation within blockchain technology, enabling automated, trustless transactions without the need for intermediaries. While they offer a secure and transparent means of transferring assets and enforcing agreements, smart contracts are still software, and thus susceptible to bugs and vulnerabilities. Given their role in managing high-value digital assets, these flaws have been exploited to devastating effect, resulting in the loss or theft of billions of dollars. Vulnerabilities can lead to serious consequences, including unauthorized fund transfers, permanent loss or locking of assets, and the circumvention of contract logic. Critically, once deployed to the blockchain, smart contracts are immutable. Machine learning solutions in this field suffer from the lack of real accurate datasets to develop models that surpass static analyzers. In this paper we provide a methodological approach to create a curated vulnerable smart contract dataset leveraging open-source tools. A weighted ensemble mechanism is used to label contracts. Secondly, dataset preprocessing is demonstrated to create a classification friendly dataset. Finally, we train several machine learning models on a generated dataset for Solidity showing its effectiveness and strong performance in vulnerability detection.
S Chowdhury, Md Hadiur Rahman Hamim, S Joy, Md. Shakil Ahmed · 7 authors
Credit allowances for individuals who are seeking financial opportunities are vital for their growth and sustainability. Traditional credit scoring systems don’t allow them to get personal loans, as individuals lack collateral or have a limited financial history. This paper approaches a novel solution that makes use of blockchain technology to overcome these issues and improve individuals’ access to financing. The main benefits of blockchain technology are transparency, security, and decentralization, which have the potential to completely transform the credit rating system. We aim to develop a decentralized credit scoring system that incorporates a wide range of parameters by combining international standard scoring systems like FICO and EQUIFAX to make it more efficient. Moreover, the decentralized credit scoring system ensures data integrity and security, reducing the risk of fraud and manipulation in the credit scoring process.
Detecting fraudulent transactions in Ethereum is challenging due to the evolving tactics of malicious actors and the complexity of blockchain transactions. This work constructs a proximity-aware graph (PAG) for each Ethereum account, where each node is connected to its transactional neighbours within two hops to capture financial interactions effectively. For better performance, EthAegis integrates both local graphical features and global transactional properties of each proximity-aware graph (PAG).To effectively capture the graphical and transactional relationships between Ethereum accounts, Graph Sample and Aggregate (GraphSAGE) are deployed. It is a Graph Neural Network (GNN) based model that is optimised for inductive learning on large-scale graphs. It is observed that EthAegis could effectively capture hidden patterns and anomalies in Ethereum transactions by considering both the transactional and graphical characteristics. Experimental evaluations on real Ethereum transaction data demonstrate that this method significantly improves fraud detection accuracy while ensuring scalability, providing a robust and adaptive solution to secure blockchain transactions. The proposed model achieves an accuracy of 0.9908, precision of 0.9922, recall of 0.9894, and F1-score of 0.9908 on the Ethereum transactional dataset.
Smart contracts control critical financial assets on blockchains, with potential weaknesses risking substantial losses. Thus, smart contract vulnerability detection is essential for maintaining blockchain ecosystem stability. Traditional methods depend extensively on expert-driven patterns, resulting in poor scalability. Although deep learning-based approaches have made significant progress, they still suffer from issues such as inflexible representations, insufficient feature modalities, and limited model capabilities. In this paper, we propose FSGDec, a novel smart contract vulnerability detection framework that fuses sequential information and structural features at the bytecode level. Firstly, an efficient node embedding method is developed for contract control flow graphs, flexibly processing node sequences and incorporating node-specific semantic information associated with weaknesses. Then, by modeling node features as time series signals, an adaptive graph wave network is introduced to automatically capture vulnerability-related structural features. Finally, a classifier is deployed to perform bug detection utilizing the extracted graph-level features that integrate semantic information. Evaluated on two real-world smart contract datasets, the experimental results demonstrate that FSGDec achieves superior performance compared to state-of-the-art baselines.
Gyuyeon Na, Minjung Park, Hyeonjeong Cha, Soyoun Kim · 9 authors
Blockchain transaction networks are complex, with evolving temporal patterns and inter-node relationships. To detect illicit activities, we propose a hybrid GCN-GRU model that captures both structural and sequential features. Using real Bitcoin transaction data (2020-2024), our model achieved 0.9470 Accuracy and 0.9807 AUC-ROC, outperforming all baselines.
Teoman Berkay Ayaz, Muhammet Furkan Özara, Ahmet Erkan Çelik, Akhan Akbulut
Blockchain systems promote transparency, decentralization, and reliability. Nevertheless, they remain vulnerable to more sophisticated fraudulent actors, particularly on newly launched platforms with little or nonexistent transaction histories. This paper introduces an innovative hybrid learning model that integrates few-shot learning with active learning to address two fundamental issues in fraud detection within financial systems: the limited availability of annotated fraud data and the ongoing evolution of illegal behavior. The study provides a comprehensive evaluation of two complementary datasets: a public dataset comprising real-life transactions obtained from Kaggle to establish a generalizable benchmark (9,841 transactions, 22.1% fraudulent) and a custom synthetic dataset designed for the PointXchange platform (197,458 transactions, 0.16% fraudulent). The approach we use generates balanced training sets and minimizes annotation costs by sampling as few as 8 to 128 samples per class and iteratively querying an oracle for useful labels. Benchmarks conducted across four families of algorithms: gradient boosting machines (XGBoost, LightGBM, CatBoost), boosting (AdaBoost), ensemble learners (Random Forest, Extra Trees), and neural networks (MLP, XNet), demonstrate the effectiveness of the proposed approach with recall scores reaching up to 0.9906.
With the shift from Centralized Finance (CeFi) to Decentralized Finance (DeFi), financial transactions have become trustless and self-executing through blockchain platforms, creating new opportunities while exposing the ecosystem to significant fraud risks. However, due to the lack of centralized oversight and the vulnerabilities in the blockchain platforms, DeFi transactions still face several security challenges, including fraud, identity theft, insider threats, and data breaches. Various methods, including regulatory frameworks, machine learning (ML), and deep learning (DL) techniques, are employed to detect these threats, particularly fraud, in DeFi transactions. Although these approaches help identify fraudulent activities, they face challenges related to accuracy and zero-day attacks due to insufficient data and the complexity of emergingfraud patterns. This study presents a novel approach for detecting and profiling fraud attacks, including zero-day ones in DeFi transactions, thereby eliminating the reliance on wallet transaction history, a limitation that previous research has heavily depended on. The proposed approach leverages two key components: a novel analyzer named DeFiTransLyzer (V1.0) and an Advanced Genetic Algorithm (AGA) for fraud transaction profiling. DeFiTransLyzer extracts 79 features from transaction and wallet data. At the same time, the AGA incorporates advanced techniques, including Penalized Fitness Evaluation, Elite Retention Strategy, Dynamic Mutation Rate, and dynamic generation, to create precise fraud profiles. By focusing solely on transaction features, the model ensures that all fraudulent activities, including zero-day ones, initiated within the first transaction of a new account can be effectively detected, without relying on prior wallet activity. To address the scarcity of comprehensive validation datasets, we introduce BCCCDeFiFraudTrans-2025, which comprises 1,026,867 annotated Ethereum transaction samples from the DeFi ecosystem. Additionally, the study establishes two taxonomies for systematic classification, covering the literature on fraud detection and profiling methods. Experimental results demonstrate that the proposed method achieves superior accuracy, precision, and efficiency while offering interpretability through its profiling mechanism. These promising outcomes highlight the potential of AGA profiling to enhance the detection and identification of fraudulent activities, including zero-day ones within DeFi transactions, contributing to the security and resilience of blockchainbased financial systems.
Ethereum's distributed ledger technology operates on decentralized principles that have transformed how digital assets are exchanged. Yet the platform's underlying design contains security weaknesses that expose users to multiple forms of malicious activity, such as fraudulent wallet schemes, network identity manipulation, pyramid financial structures, and service interruption attacks. This study aims to address the cyber-attacks on the Ethereum platform by developing an innovative ensemble machine-learning framework based on a refined feature set derived from real-time Ethereum transactions. The proposed framework employed a method that encapsulates the feature selection process, leveraging the advantages of machine learning algorithms to derive the refined feature set. The conducted experiment employs Random Forest and XGBoost, to identify the significant features. To mitigate the overfitting issue in the proposed machine learning model, the SMOTEENN (Synthetic Minority Oversampling Edited Nearest Neighbors) technique, has applied. The experimental results reveal that the proposed model exceeds the performance of existing fraud detection approaches, achieving a 99.4% accuracy and a Matthews Correlation Coefficient (MCC) of 94.9%.
The global financial ecosystem is currently undergoing an unprecedented digital metamorphosis, characterized by the rapid adoption of decentralized finance, mobile banking, and real-time payment systems. While these advancements have significantly enhanced consumer convenience and operational efficiency, they have simultaneously expanded the attack surface for sophisticated fraudulent actors. Traditional fraud detection methodologies, predominantly reliant on static rule-based frameworks and retrospective manual auditing, are increasingly proving inadequate in the face of modern, high-velocity deceptive maneuvers. This research paper provides an exhaustive analysis of the transition from reactive fraud investigation to proactive, AI-powered real-time prevention. By synthesizing a vast array of peer-reviewed research and industrial case studies, the study explores the implementation of advanced machine learning architectures, including Graph Neural Networks (GNNs) for relational intelligence, Long Short-Term Memory (LSTM) networks for temporal sequence analysis, and Isolation Forests for unsupervised anomaly detection. The analysis further delves into the architectural requirements for real-time processing, highlighting the role of cloud-native microservices, stream-processing engines, and edge intelligence. Furthermore, the paper addresses the critical intersection of technical efficacy, regulatory compliance, and ethical accountability through the lens of Explainable AI (XAI) and Federated Learning. The findings suggest that a multi-layered, synergistic approach - integrating AI with cybersecurity protocol is essential for reducing detection latency from hours to milliseconds, thereby safeguarding the integrity of the global financial network.
As Ethereum continues to gain traction as a leading blockchain platform, its open and decentralized nature has also made it an attractive target for fraudulent activities. This paper presents a machine learning-based approach to detect fraudulent Ethereum transactions by analyzing behavioral patterns within transaction data. Using a labeled dataset of Ethereum transactions, various classification algorithms such as Random Forest, XGBoost, and Support Vector Machines were trained and evaluated. The proposed system focuses on identifying anomalies and suspicious transaction behavior by extracting relevant features like gas usage, transaction value, and timing. Experimental results show that the model can achieve high accuracy and precision in distinguishing between legitimate and fraudulent transactions. This work contributes to the growing field of blockchain security by demonstrating the viability of intelligent fraud detection techniques and providing a framework that can be integrated into real-world applications.
Smart contracts play a pivotal role in blockchain ecosystems, and fuzzing remains a critical approach to securing them. However, existing smart contract fuzzers often optimize either seed generation or mutation scheduling in isolation and rely on narrow, fragmented feedback signals, leaving multi-transaction reasoning and stagnation recovery under-explored. In this work, we propose aLarge Language Models(LLMs)-based Multi-feedback Smart Contract Fuzzing framework (LLAMA). Key components of the proposed LLAMA include: (i) a hierarchical prompting strategy that guides LLMs to generate structurally valid, context-aware multi-transaction initial seeds, together with a lightweight pre-fuzzing phase that validates and prioritizes high-potential LLM-generated candidates; (ii) a multi-feedback-guided evolutionary optimization module that jointly optimizes seed selection and mutation scheduling by a group of constraints for driving an LLM-bootstrapped bandit scheduler. (iii) an LLM-guided hybrid fuzzing module that integrates evolutionary fuzzing with a dual-channel recovery mechanism, which concurrently employs asynchronous coverage-stagnation- based LLM reseeding and selective symbolic execution to resolve complex path constraints. Our extensive experiments demonstrate that LLAMA outperforms state-of-the-art fuzzers in both coverage and vulnerability detection. Specifically, it achieves 92% instruction coverage on small contracts and 81% on large contracts, while detecting 132 out of 148 known vulnerabilities across diverse categories. Ablation studies further evidence that the proposed multi-feedback and hybrid recovery strategies have strong impact on LLAMA’s performance. The results explain LLAMA’s effectiveness, adaptability, and practicality in complex smart contract scenarios.
Highly accurate predictions of cryptocurrency prices are of paramount importance to investors and researchers, as they can guide investment strategies and market analysis. However, due to the nonlinear and volatile nature of the cryptocurrency market, it is challenging to assess the distinct characteristics of time-series data, which results in difficulties in generating appropriate and reliable price forecasts. Numerous studies have been conducted on cryptocurrency price prediction using different deep learning-based algorithms, as these techniques have shown promise in capturing the complex patterns and trends in this market. This study proposes three types of recurrent neural networks: Long Short-Term Memory, Gated Recurrent Unit, and Bi-Directional LSTM, for exchange rate predictions of the three major cryptocurrencies in the world, as measured by their market capitalization: Bitcoin, Ethereum, and Litecoin. The experimental results on the three major cryptocurrencies using both Root Mean Squared Error and Mean Absolute Percentage Error demonstrate that the Bi-LSTM model performed better in prediction than LSTM and GRU, and can be considered the most effective algorithm for this task. Bi-LSTM presented the most accurate prediction compared to GRU and LSTM, with MAPE values of 0.036, 0.041, and 0.124 for BTC, LTC, and ETH, respectively. The study suggests that the proposed prediction models are accurate and reliable in forecasting cryptocurrency prices and can be beneficial for investors, traders, and researchers in the cryptocurrency market.
Survival modeling predicts the time until an event occurs and is widely used in risk analysis; for example, it's used in medicine to predict the survival of a patient based on censored data. There is a need for large-scale, realistic, and freely available datasets for benchmarking artificial intelligence (AI) survival models. In this paper, we derive a suite of 16 survival modeling tasks from publicly available transaction data generated by lending of cryptocurrencies in Decentralized Finance (DeFi). Each task was constructed using an automated pipeline based on choices of index and outcome events. For example, the model predicts the time from when a user borrows cryptocurrency coins (index event) until their first repayment (outcome event). We formulate a survival benchmark consisting of a suite of 16 survival-time prediction tasks (FinSurvival). We also automatically create 16 corresponding classification problems for each task by thresholding the survival time using the restricted mean survival time. With over 7.5 million records, FinSurvival provides a suite of realistic financial modeling tasks that will spur future AI survival modeling research. Our evaluation indicated that these are challenging tasks that are not well addressed by existing methods. FinSurvival enables the evaluation of AI survival models applicable to traditional finance, industry, medicine, and commerce, which is currently hindered by the lack of large public datasets. Our benchmark demonstrates how AI models could assess opportunities and risks in DeFi. In the future, the FinSurvival benchmark pipeline can be used to create new benchmarks by incorporating more DeFi transactions and protocols as the use of cryptocurrency grows.
The rapid evolution of financial fraud in digital finance applications—such as mobile banking, cryptocurrency transactions, and online payment gateways—has rendered traditional rule-based detection systems increasingly ineffective, leading to heightened financial losses and security vulnerabilities. These systems struggle to adapt to complex fraudulent schemes, resulting in inefficiencies in identifying such activities. This research introduces a Scalable Black Widow-driven Gradient Boosting Machine (SBW-GBM) framework designed to enhance fraud detection through real-time data analysis and machine learning (ML). The framework evaluates performance in both decentralized finance and traditional financial contexts, utilizing two separate datasets. The first dataset is based on an Ethereum Phishing Transaction Network; the second originates from Kaggle and pertains to Credit Risk Assessment. Data preparation involves addressing missing values, normalizing numerical features, and employing outlier detection techniques to improve data quality. For feature extraction, Principal Component Analysis (PCA) reduces data dimensionality while preserving critical information regarding transaction behaviors. The classification employs Gradient Boosting Machine (GBM) for high predictive accuracy, with the SBW algorithm dynamically fine-tuning GBM hyperparameters to enhance efficiency. Inspired by black widow spiders, the SBW algorithm optimizes hyperparameter selection by eliminating weak solutions and reinforcing stronger ones. This results in an adaptive fraud detection model trained on labeled transaction data. Experimental results confirm that SBW-GBM achieves an F1-score of 0.899, an accuracy of 0.948, an error rate of 0.128, and an experimental runtime of 0.40 seconds on Dataset 1, outperforming the baseline FFSVM. For Dataset 2, SBW-GBM attains a precision of 0.875, a recall of 0.660, an F1-score of 0.750, and an AUC of 0.932, surpassing benchmark traditional classifiers. By continuously learning from new transaction patterns, the proposed framework ensures adaptability to emerging threats and supports real-time fraud detection.
With the widespread application of blockchain technology, the security issues of smart contracts have become increasingly prominent. Traditional contract vulnerability detection methods mostly depend on source code, making it hard to adapt to the actual situation where only bytecode is available on-chain. This paper presents a novel bytecode-based vulnerability detection method for smart contracts (BC-GNN). By parsing the contract bytecode, it extracts opcode sequences, control flow graphs (CFGs), and data flow graphs (DFGs), integrates them into a unified heterogeneous graph structure, and uses graph neural networks (GNNs) to automatically detect multiple types of vulnerabilities. Experimental results show that BC-GNN achieves excellent performance on multiple actual and enhanced datasets. In particular, it demonstrates strong applicability and generalization ability in scenarios where source code is unavailable. This approach offers new ideas and technical pathways for blockchain smart contract security detection.
Hong Qu, Krzysztof Gogol, Florian Grötschla, Claudio J. Tessone
Decentralized Finance (DeFi) lending enables permissionless borrowing via smart contracts. However, it faces challenges in optimizing interest rates, mitigating bad debt, and improving capital efficiency. Rule-based interest-rate models struggle to adapt to dynamic market conditions, leading to inefficiencies. This work applies Offline Reinforcement Learning (RL) to optimize interest rate adjustments in DeFi lending protocols. Using historical data from Aave protocol, we evaluate three RL approaches: Conservative Q-Learning (CQL), Behavior Cloning (BC), and TD3 with Behavior Cloning (TD3-BC). TD3-BC demonstrates superior performance in balancing utilization, capital stability, and risk, outperforming existing models. It adapts effectively to historical stress events like the May 2021 crash and the March 2023 USDC depeg, showcasing potential for automated, real-time governance.
Abstract— This study integrates blockchain technology and machine learning to enhance credit card fraud detection. Precise fraud prediction is performed using advanced algorithms such as Random Forest, Logistic Regression, XGBoost, and Bayesian models. Tools such as Ganache and MetaMask from Ethereum blockchain facilitate safe and transparent tracking of suspicious transactions. Decentralized and tamper-proof properties of blockchain add reliability, and machine learning adds precision and flexibility. The system is highly accurate and transparent and has the potential to be used to fight financial fraud. Keywords— Credit Card Fraud, Blockchain, Machine Learning, Ethereum, Web3, SMOTE, XGBoost, Streamlit
The rise of digital finance has led to a surge in fraudulent activities, particularly in credit card transactions and cryptocurrency ecosystems. With financial crimes becoming more sophisticated, traditional fraud detection methods often fail to identify complex fraudulent patterns. This research explores the application of machine learning (ML) and artificial intelligence (AI) techniques to enhance the security of digital finance by detecting fraudulent activities in credit card transactions and cryptocurrency wallets within the USA. The study utilizes large-scale transaction datasets containing key financial indicators such as transaction frequency, spending patterns, anomaly scores, and network behaviors. To develop an AI-driven fraud detection framework, we implement and compare six machine learning models: XGBoost, RLightGBM, Decision Trees, K-Nearest Neighbors (KNN), Convolutional Neural Networks (CNNs), and Autoencoders. The models are trained on both structured financial data (e.g., credit card transaction logs) and unstructured blockchain transaction records (e.g., Bitcoin wallet addresses and transaction flows). To address data imbalance, the study applies the Synthetic Minority Over-sampling Technique (SMOTE), ensuring fair representation of fraudulent transactions. Model performance is evaluated using Precision, Recall, F1-score, and ROC-AUC metrics to determine the most effective fraud detection approach. Additionally, the research emphasizes data privacy and security, incorporating anonymization techniques and regulatory compliance measures to safeguard sensitive financial information. This study contributes to the ongoing fight against financial fraud by demonstrating how AI-based solutions can enhance the security and resilience of digital finance systems in the USA.