Blockchain Papers

Mikhail Urinson

ThemeThe convergence of Artificial Intelligence (AI), Quantitative Finance, and Blockchain technologies is reshaping how capital is analyzed, deployed, and optimized across both traditional and decentralized markets.

Huaiyu Jia, Jieshun You, Jingyu Liu, Yizhi Luo · 5 authors

Automated Market Makers (AMMs), as a core infrastructure of decentralized finance (DeFi), uniquely drive on-chain asset pricing through a deterministic reserve ratio mechanism. Unlike traditional markets, AMM price dynamics is triggered largely by on-chain events (e.g., swap) that change the reserve ratio, rather than by continuous responses to off-chain information. This makes event-level analysis crucial for understanding price formation mechanisms in AMMs. However, existing research generally neglects the micro-structural dynamics at the AMMs level, lacking both a comprehensive dataset covering multiple protocols with fine-grained event classification and an effective framework for event-aware modeling. To fill this gap, we construct a dataset containing 8.9 million on-chain event records from four representative AMMs protocols: Pendle, Uniswap v3, Aave and Morpho, with precise annotations of transaction type and block height timestamps. Furthermore, we propose an Uncertainty Weighted Mean Squared Error (UWM) loss function, which incorporates the block interval regression term into the traditional Temporal Point Process (TPP) objective function by weighting the uncertainty with homoscedasticity. Extensive experiments on eight advanced TPP architectures across four representative DeFi protocols demonstrate that this loss function reduces the time prediction error by an average of 31.17% while maintaining the accuracy of event (transaction) type prediction, establishing a robust benchmark for event-aware prediction in the AMMs ecosystem. This work provides the necessary data foundation and methodological framework for modeling the discreteness and event-driven characteristics of on-chain price discovery. All datasets and source code are publicly available. https://github.com/finbrain-lab-hkustgz/Deep-AMM-Events

Gaurav Kokane, Dr. Pratibha V. Kashid, Prathamesh Pandit, Hitesh Patil · 5 authors

ABSTRACT Cryptocurrency trading has rapidly evolved into a highly dynamic and technology-driven financial domain, attracting significant attention from investors, researchers, and institutions worldwide. This paper presents a comprehensive review of modern cryptocurrency trading platforms by combining blockchain technology, artificial intelligence, and advanced trading mechanisms to create a secure, efficient, and scalable trading ecosystem. The study highlights the use of deep learning approaches such as Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and attention-based models for improving cryptocurrency price prediction. These techniques utilize technical indicators, trading patterns, and social media data to enhance prediction accuracy. In addition, reinforcement learning strategies are explored to optimize trading decisions and improve performance under highly volatile market conditions. Furthermore, the paper discusses real-time data integration using APIs, secure authentication mechanisms, and scalable system architectures required for continuous trading operations. It also examines the regulatory landscape of cryptocurrency, particularly in the Indian context, including taxation policies and emerging concepts like Central Bank Digital Currencies (CBDCs). Overall, this review provides insights into the development of intelligent, secure, and user-friendly cryptocurrency trading platforms such as CryptoVista. Keywords: Cryptocurrency, Blockchain, Deep Learning, Reinforcement Learning, Smart Contracts, Real-Time Data, Trading Platforms, Security, Scalability, CBDC

Wisam Bukaita, Xinrui Li

This study examines dynamic interdependencies and risk transmission among major cryptocurrencies and traditional financial assets, including Bitcoin, Ethereum, U.S. equities, and gold, over the period 2017–2024. Particular attention is given to the structural shift associated with the 2024 U.S. spot Bitcoin exchange-traded fund (ETF) approval, which marked a significant milestone in the institutionalization of cryptocurrency markets. Using daily data, the analysis distinguishes volatility-driven co-movement from structural spillover effects across markets. Dependence structures are modeled using tail-sensitive Student-t copulas applied to GARCH-filtered returns to capture nonlinear and extreme co-movements, while a vector autoregressive framework combined with generalized impulse response functions and Diebold–Yilmaz connectedness measures is employed to evaluate order-invariant shock transmission dynamics across pre- and post-ETF regimes. The results reveal three main findings. First, cryptocurrencies display strong internal dependence and short-horizon contagion, with Bitcoin consistently acting as the dominant transmitter of shocks to Ethereum over an approximately three-day transmission window. Second, linkages between cryptocurrencies and equity markets remain moderate and largely regime-dependent rather than indicative of persistent structural spillovers. Third, gold remains weakly connected throughout the sample, maintaining its role as a diversification asset. Portfolio analysis further indicates that including Bitcoin can reduce portfolio variance by 4–7% and Value-at-Risk by up to 5%, although economic gains are sensitive to transaction costs. Overall, the findings suggest that cryptocurrencies function as a partially segmented asset class, offering conditional diversification benefits despite increasing institutional adoption.

Abdullah Albanyan, Hassen Louati, Ali Louati

The rapid convergence of artificial intelligence and blockchain technologies has increased the demand for efficient and accurate methods to detect fraudulent behavior in smart contract–driven systems. Smart contracts automate digital transactions in decentralized environments, yet they remain vulnerable to fraud while operating under strict computational and scalability constraints. In this study, we propose an evolutionary-guided CNN compression framework tailored for Convolutional Neural Networks (CNNs) aimed at improving fraud detection in smart contract analysis while significantly reducing model complexity. The proposed approach uses evolutionary optimization to guide structured model compression, enabling the removal of redundant parameters without compromising predictive performance. Experimental evaluations demonstrate up to a 50% reduction in model parameters while maintaining 97.8–97.9% classification accuracy, making the resulting models suitable for deployment in resource-constrained environments. By combining evolutionary optimization with CNN-based fraud detection, this work provides an efficient and interpretable solution for smart contract analysis, supporting scalable and practical deployment in blockchain-related security applications.

Prabakaran Raghavendran, Yamini Parthiban

This research paper demonstrates how to manage Caputo fractional neutral integro-differential equations which include both integral and nonlinear elements through a unified framework that models dynamic systems with memory-based dynamics. The research establishes sufficient conditions for controllability through fixed point theory in a Banach space framework which requires particular assumptions while the study focuses on the K1<1 condition which leads to the existence of a controllable solution. The proposed criteria are demonstrated through a numerical example which tests the theoretical results. The real-world case study uses artificial neural network (ANN) technology to predict Litecoin prices through the application of the fractional controllability model which analyzes historical financial data. The hybrid framework enables precise forecasting of nonlinear time series because it combines fractional calculus mathematical principles with ANN learning abilities. The proposed method demonstrates its predictive efficiency. The method shows robust performance through experimental results using cross-validation and performance metrics. The proposed model demonstrates competitive performance while providing additional advantages such as incorporation of memory effects and theoretical controllability. The research establishes a novel connection between fractional dynamical systems and machine learning which serves as an essential tool for studying complicated systems in theoretical research and practical applications.

Ali Yeganeh, Mohammad Sadjad, Jean‐Claude Malela‐Majika, Sollie Millard

No abstract is available for this record.

Abdulilah I. Mubarak

Type of the article: Research ArticleAbstractCryptocurrency markets are highly volatile, making price prediction a complex yet essential task for investors, financial engineers, and institutions. The purpose of this study is to evaluate whether Bayesian optimization of technical indicator parameters significantly improves the forecasting performance of Long Short-Term Memory (LSTM) models compared to baseline configurations. The study used daily Bitcoin and Ethereum price data from January 2016 to September 2025. Six technical indicators representing trend, momentum, volatility, and volume-based technical indicators are constructed and dynamically optimized through Bayesian optimization. The optimized indicators are then used as inputs to an LSTM forecasting framework. The study found that the baseline LSTM model achieved moderate predictive accuracy, where Ethereum outperformed Bitcoin. After optimization, both models exhibited improved performance, reducing the forecasting error for Bitcoin by 36.4% and for Ethereum by 12.2%. LSTM model with Bayesian optimized indicators showed a higher forecasting accuracy as compared to the baseline model, with 32% and 18.6% improvements for Bitcoin and Ethereum, respectively. These findings suggest that combining optimized technical indicators with LSTM models enhances predictive power in cryptocurrency markets. The approach offers a robust forecasting framework for traders, analysts, and algorithmic systems in high-volatility environments.Acknowledgment“This work was funded by the Deanship of Scientific Research, Vice Presidency for Graduate Studies and Scientific Research, King Faisal University, Saudi Arabia [Project No. KFU261690].”

Bobba Pavan Santosh, Bommeneni Pavan Madhav, Dr. J. R. Jayavelu, Dr. P. Dhivya

Cryptocurrencies have found their way into contemporary financial systems as a significant component of modern-day financial systems because of their decentralized nature, their ease of adoption and uptake. Ether is considered to be one of the most actively traded currencies and its value tends to be highly volatile. It is not easy to forecast the market price trend of Ethereum due to the influence that technical trends, investor behavior, and external factors have over the market. In this project, the researcher will use machine learning to assess the future price direction of Ethereum the following day through the use of Python. The past trends of prices are analyzed and augmented with various technical indicators in order to reflect the market trends and momentum. The best potential machine learning model was selected after training and evaluating many models using Logistic Regression. The results demonstrate that machine learning may be used to provide rational insights into the price movement of Ethereum and to aid in decision-making using these insights.

Mr. Heet Vipulkumar Chaudhary

Stock markets in emerging economies are shaped by a combination of global integration and domestic financial drivers. In recent years, modern variables such as cryptocurrencies have drawn attention as potential new determinants of equity performance. This study evaluates the comparative influence of traditional variables-Foreign Institutional Investor (FII) flows, USD/INR exchange rate, and NIFVIX-and a modern variable, Bitcoin returns, on the Nifty50 index. Monthly data spanning January 2015 to January 2025 were collected from Investing.com and Moneycontrol. Nifty50, Bitcoin, and USD/INR series were converted into log returns, while FII flows and NIFVIX were used in their original form. Correlation analysis and simple linear regression were done by using Microsoft Excel to measure associations and explanatory power. The results indicate a clear hierarchy of explanatory strength. USD/INR log returns emerged as the most influential determinant, explaining 26% of Nifty50 return variation with a strong negative relationship. NIFVIX explained 14% of the variation, also with a negative and highly significant effect. Bitcoin returns exhibited a modest but statistically significant positive effect, explaining around 8% of the variance. In contrast, both FII equity and total flows were statistically insignificant. The findings suggest that traditional variables-particularly exchange rates and volatility indices-remain dominant drivers of Indian equity returns, while modern variables such as Bitcoin are new but not yet central. The study contributes by showing one of the first systematic comparisons between traditional and modern variables in the Indian equity market context. Keywords: Nifty50, Bitcoin Returns, Foreign Institutional Investors (FII), USD/INR Exchange Rate, NIFVIX, Traditional vs. Modern Variables, Indian Stock Market

Drissia Ennagoura, Kamal El Kehal, Abdelhamid Berdai, Safae Merzouk · 8 authors

Cryptocurrency price prediction is challenging due to strong nonlinearity and high volatility. This paper comparatively evaluates three forecasting models for Ethereum (ETH): SARIMAX with exogenous technical indicators, Long Short-Term Memory (LSTM) networks, and Facebook Prophet. Relative Strength Index (RSI), Moving Average Convergence Divergence (MACD), and Exponential Moving Average (EMA) are incorporated to enhance signal quality. Empirical results reveal clear trade-offs between predictive accuracy, profitability, and risk. SARIMAX achieves the highest directional accuracy (75.00%) with limited profitability, while LSTM yields the highest cumulative profit (23.84%) at the cost of higher drawdown. Prophet provides a balanced compromise between accuracy and risk. The study contributes by jointly evaluating statistical forecasting accuracy and trading-oriented performance metrics, offering practical insights into model suitability for different investor risk profiles.

Zekai Liu, Xiaoqi Li, Wenkai Li, Zongwei Li

Traditional consensus mechanisms, such as Proof of Stake (PoS), increasingly reveal an excessive dependency on large liquidity providers. Although the Proof of Liquidity (PoL) mechanism serves as a critical paradigm for incentivizing sustained liquidity provision and ensuring market stability, its transition from asset staking to active liquidity management significantly increases the complexity of underlying smart contract economic models and interaction logic. This renders hidden liquidity logic flaws difficult to detect via traditional methods, seriously threatening the system stability and user asset security of mainstream DeFi and emerging PoL ecosystems. To address this, we propose the LiquiLM framework, which integrates Large Language Models (LLMs) with a Dynamic Co-Attention Network (DCN). By establishing a dynamic interaction between liquidity-critical contracts and flaw descriptions, the framework effectively bridges the semantic gap between underlying code implementations and high-level liquidity intents. We evaluate the performance of LiquiLM on 1,490 validation contracts (covering precision, recall, specificity, and F1-score). The results show that it achieves significant effectiveness in auditing and explaining liquidity flaws: in experiments using Gemini 3 Pro and GPT-4o as backbone models, respectively, the F1-scores both exceed 90%. Furthermore, through an in-depth audit of 1,380 real-world PoL and Ethereum economic contracts, LiquiLM successfully identifies 238 high-risk contracts and assists in discovering 10 vulnerabilities that have received CVE certification.

Imtinokshang, Anand Stalin, Muhammad Ziyan, J Thangakumar

No abstract is available for this record.

Ningyu Zhou

Cryptocurrency prices often move with narratives and investor sentiment. This paper builds a multilingual crypto sentiment index, ML-CryptoSI, using daily news text in six languages and Binance market data for BTC and ETH. We first aggregate language-level daily sentiment and then use PCA to extract the common component across languages. Next, we test whether ML-CryptoSI predicts next-day returns and volatility proxies after controlling for lagged market conditions, liquidity, and day-of-week fixed effects. The results show that ML-CryptoSI has incremental information for returns, especially for ETH, and the effect is stronger on high news-intensity days. In contrast, the evidence for volatility prediction is weak in this short sample. Overall, the findings suggest that the common factor in multilingual news sentiment matters for short-run crypto pricing and is state dependent.

Nrusingha Tripathy, Sarbeswara Hota, Debabrata Singh, Subrat Kumar Nayak

No abstract is available for this record.

Saurabh Jain, Kamal Kishor Choure

As blockchain continues to saturate into a technological infrastructure for decentralized trust and cost-effective data processing, existing consensus mechanisms remain burdened by immutable issues with scalability, energy inefficiency, and adaptive security. This is an alternative to conventional Proof of Work (PoW), Proof of Stake (PoS), and Byzantine Fault Tolerance (BFT) algorithms that only provide deterministic agreement, but are still expensive, inflexible, and weak under dynamically stable networks. Over the recent years, there has been a lot of development in Machine Learning (ML) and Artificial Intelligence (AI), which have introduced intelligent self-learning consensus mechanisms to increase adaptability, efficiency, and resilience. It highlights the architectural aspects and the operational entities of various ML-powered and hybrid consensus protocols, including PoW–PoS, DPoS–PBFT, and PoCASBFT, as well as their performance implications. Supervised, unsupervised, reinforcement, and federated learning methods are surveyed to provide insights for predictive validation, anomaly detection, energy optimization, and node trust management in blockchain networks [12]. It then compares the performance of its state-of-the-art protocols, demonstrating that ML-assisted hybrids provide 15–40% throughput gains and up to 35% energy savings over the corresponding protocols when trained with traditional models. Ultimately, the paper notes scalability, interpretability, and adversarial ML as the main risks of research in this space, and glances at future directions toward cognitive consensus architectures, also noting that these architectures should be self-healing, context-aware, and able to balance decentralization, performance, and security themselves.

Nada Dammak, Néjib Hachicha

Abstract: This study investigates the dynamic relationships between Bitcoin, oil prices, and the US dollar (USD) using a Vector Autoregressive (VAR) model. Utilizing daily data from 2018 to 2023, the analysis reveals that both Bitcoin and oil prices exert significant short-term impacts on the USD, though these effects diminish over the long term. Bitcoin, characterized by its high volatility and safe-haven attributes, serves as an alternative asset during periods of economic uncertainty, while oil prices influence the dollar through trade flows and inflationary pressures. The findings highlight the transient nature of these interactions, with Bitcoin and oil acting as short-term pressure factors on the USD. These insights are crucial for investors and policymakers in managing risks and optimizing strategies in a volatile financial environment. This study contributes to the literature by providing empirical insights into the interconnectedness of cryptocurrencies, commodities, and currencies, offering valuable implications for financial decision-making. Keywords: Bitcoin, Oil Prices, US Dollar (USD), Vector Autoregressive (VAR) Model, Cryptocurrencies, Exchange Rates, Safe-Haven Assets JEL Classification Number: C32, E44, G15, Q43

Ozan Nadirgil

Abstract Introduction of cryptocurrency markets offered unique confidentiality and hedging benefits to investors, since they have been rapidly integrated into the financial system through forming substantial long- and short-term interdependencies with equity markets. Previous studies present contradicting and ambigious conclusions about the financial contagion between the equity and crypto markets, nonetheless they do not adequately pose light on the impacts of global events on the relationship between these markets. To address these concerns, this research examines the volatility spillovers between the equity and Decentralized Finance (DeFi) markets through simulating the dependencies with the aim of eliminating the bias and drawbacks of the traditional models by appling an optimized and original analytical framework composed of Deep Neural Network (DNN) and Time Varying Parameters Vector Auto Regression (TVP-VAR) models. Results identify substantial transmission from Bitcoin (BTC), Nasdaq 100 (NASDAQ), Shanghai Stock Exchange (SSE), and New York Stock Exchange (NYSE) to BNB, Euronext, Tether (TET), and Ethereum (ETH) and reflect the significant impact of the global important events on the financial spillovers. Model assessment results confirm the robust accuracy, fitness, reliability, and validity of the model, as well as the success of key parameter optimization.

Simona-Vasilica Oprea, Adela BÂRA

Abstract This research presents a regime-aware hybrid forecasting framework for the Bitcoin market’s nonlinear, nonstationary and regime-switching behavior. The architecture integrates econometric models, neural forecasting and meta-learning, unified under a regime-detection mechanism using probabilistic inference. Central to the approach is a Hidden Markov Model (HMM) trained on log returns, which infers latent market regimes, bull, bear and sideways, based on statistical characteristics rather than arbitrary thresholds. Each detected regime triggers a specialized forecasting model: ARIMAX for volatile bear markets, SARIMAX for cyclical sideways periods and NeuralProphet for nonlinear bullish dynamics. These models leverage historical returns (Jan. 2012-Jun. 2025) and external signals, including technical indicators (RSI, MACD, Bollinger bands) and volatility metrics. A meta-learning layer, implemented via XGBoost, dynamically selects the optimal model at each time step based on the regime. This enables real-time adaptation to evolving market conditions. Predictions are made on log returns and translated into price forecasts through exponentiation. The framework’s performance is evaluated using R 2 , Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). The regime-aware model outperforms the no-regime model significantly across all metrics, especially in error reduction (MAE cut by ~ 56%) and higher explanatory power (R² increased from 0.82 to 0.91). Ablation results confirm the structural validity of the proposed framework, with the regime–model assignment (ARIMAX for bear, SARIMAX for sideways, NeuralProphet for bull) achieving the lowest forecasting error (MAE = 736, R 2 = 0.93) at the yearly level and outperforming alternative configurations. The inferred regimes exhibit economically meaningful persistence (average durations 14.8–22.4 days) and transition stability (diagonal probabilities 0.91–0.94). The meta-learning component shows coherent and interpretable behavior, with regime labels and recent model errors explaining nearly 70% of decision weight and regime-consistent model selection exceeding 80%. These forecasting gains translate into tangible economic benefits: in a six-month backtest, the proposed strategy delivers the highest return (19%), lowest drawdown (19%) and highest Sharpe ratio (1.01), outperforming all benchmarks.

Batuhan Karabay

Bu çalışma, Bitcoin fiyat tahmininde mevsimsel ARIMA (SARIMA) modelinin öngörü performansını, basit bir Naive kıyas modeliyle açık biçimde karşılaştırarak yeniden değerlendirmeyi amaçlamaktadır. Analiz, 12 Mart 2021 ile 12 Mart 2026 dönemini kapsayan günlük Bitcoin kapanış fiyatlarına dayanmaktadır. Seri logaritmik forma dönüştürülmüş ve durağanlık özellikleri fark alma işlemleriyle incelenmiştir. İlk aşamada mevsimsel olmayan ARIMA modelleri tahmin edilmiş, ardından mevsimsel dinamikleri içeren alternatif SARIMA modelleri değerlendirilmiştir. Model seçiminde parametre anlamlılığı ile Ljung-Box tanı istatistikleri dikkate alınmış ve mevsimsel hareketli ortalama bileşeninin kısmen anlamlı olduğu görülmüştür. Bu çerçevede SARIMA(0,1,1)(0,1,1)[30] nihai mevsimsel aday model olarak belirlenmiştir. Tahmin performansı değerlendirmesi, yalnızca model uyumuna değil, örneklem dışı tahmin doğruluğuna odaklanmaktadır. Bu amaçla SARIMA modelinin performansı, ortalama mutlak hata (MAE) ve hata kareler ortalamasının karekökü (RMSE) ölçütleri kullanılarak Naive model ile karşılaştırılmıştır. Bulgular, örneklem dışı dönemde Naive modelin SARIMA modeline göre belirgin biçimde daha düşük tahmin hataları ürettiğini göstermektedir. Naive model için MAE 0.016011 ve RMSE 0.023173 iken, SARIMA modeli için bu değerler sırasıyla 0.23172 ve 0.28931’dir. Sonuçlar, Bitcoin gibi yüksek oynaklığa sahip finansal zaman serilerinde daha karmaşık mevsimsel yapıların her zaman daha üstün tahmin performansı sağlamadığını göstermektedir.

Wenhao Zhang, Zhenpeng Tang, Xiaowen Zhuang, Yi Cai · 5 authors

The cryptocurrency market has attracted significant attention from global investors, with Cardano (ADA) ranking among the top cryptocurrencies by market capitalization. However, predicting ADA returns remains challenging due to the complex, multi-scale dynamics influenced by Federal Reserve policies, geopolitical events, and high-frequency trading. This study proposes a “Sliding EMD–Multi Variables” framework for cryptocurrency return prediction, leveraging Empirical Mode Decomposition’s multi-scale fractal properties to capture nonlinear dynamics at different time scales. The sliding window decomposition method addresses data leakage issues while incorporating key economic and policy variables at the component level. The empirical results demonstrate that the Sliding EMD system significantly outperforms univariate and multivariate benchmarks. Compared to the univariate system, it improves MSE, RMSE, SMAPE, and DSTAT by 0.83%, 0.42%, 5.23%, and 0.43%, respectively, while enhancing investment metrics (maximum drawdown, Sharpe ratio, Sortino ratio, Calmar ratio) by 0.19, 0.36, 0.95, and 0.15. Against the multivariate system, improvements reach 5.52%, 3.14%, 5.74%, and 17.62% in prediction accuracy, with investment performance gains of 0.47, 1.69, 4.27, and 0.31. Incorporating economic variables at the component level yields additional improvements of 0.94%, 0.47%, and 0.78% in MSE, RMSE, and MAE. These findings offer valuable insights for cryptocurrency portfolio optimization using fractal-based decomposition methods.

Tahsin Galip TEKİN, Sait Patır

In this study, it is aimed to compare quantitative forecasting methods (traditional and learning based) in cryptocurrency market. For his purpose the daily prices between 16 September 2017 – 15 September 2022 of Bitcoin, Ethereum, Binance Coin and Monero were analyzed with five different methods: ARIMA, exponential smoothing, artificial neural networks, RNN and LSTM.In the results it is indicated that exponential smoothing method is the most successful method at forecasting daily prices. The method has high performance in forecasting BTC, ETH and BNB daily prices. But at forecasting daily XMR prices, artificial neural networks method was the most successful one.The other point which was detected in this study is deep learning based methods made some unsuccessful forecasts. This is thought to be due to the fact that deep learning methods require more data. In future studies, using other quantitative methods (e.g. GRU, XGBoost, transformer models) on other cryptocurrencies will contribute to the literature.

Tim Weingärtner

This paper introduces a novel architecture that brings the Algorithmic Contract Types Unified Standards (ACTUS) to Ethereum by combining on-chain contract definition with off-chain deterministic computation and on-chain, oracle-mediated settlement. The design encodes ACTUS terms and life-cycle state in an ERC-20 stablecoin with balance locks, while a whitelisted multi-oracle computes cash-flows and settles them without manual intervention. This approach gives DeFi instruments bank-grade, standardized cash-flow semantics and auditable automation at low cost. In a three-year time-warp over a portfolio we measure \(\approx\) 60,000 gas per settlement ( \(\approx\$0.08\) ), with oracle throughput around 240 contracts/s on commodity hardware. Initial deployment is \(\approx\$3.74\) , and per-contract setup \(\approx\$0.48\) .

Youssef Said, Al Mahdi Khaddar, Lahcen Hassine, Ahmed Eddaoui · 5 authors

Gas Gas consumption is a critical factor influencing the efficiency, scalability, and operational cost of Ethereum smart contracts.As contract complexity grows, identifying structurally gas-inefficient patterns becomes essential for improving development workflows and preventing costly deployment decisions.This study presents a graph-based deep learning framework for detecting gas-inefficiency risk patterns at the function level, leveraging multi-relational Graph Attention Networks (GAT) applied to function-level contract graphs.By modeling call dependencies, control-flow interactions, and storage-based data dependencies, the model learns structural indicators associated with excessive gas consumption while explicitly excluding direct gas metrics from the feature space to prevent data leakage.Experimental results under a strict contract-level data split protocol demonstrate strong classification performance and stable generalization across held-out contracts under the main split protocol, and consistent behavior under an additional time-forward temporal robustness check.Ablation analysis confirms the contribution of dependency-aware edges and semantic features to predictive accuracy, highlighting the importance of modeling cross-function interactions rather than isolated code metrics.Beyond predictive performance, the proposed approach provides interpretable attention weights that identify structurally influential functions, supporting predeployment analysis and developer-guided manual refactoring decisions.By framing gas inefficiency as a global structural property emerging from function interactions, this work contributes aa scalable and explainable methodology for structural gas-inefficiency detection in smart contracts.The proposed model performs structural detection only and does not automatically modify or optimize smart contract code.