Publications
246 results found
Srinivasan P, Knottenbelt W, 2024, Offline Reinforcement Learning with Behavioral Supervisor Tuning, 33rd International Joint Conference on Artificial Intelligence (IJCAI 2024)
Dong N, Wang Z, Sun J, et al., 2024, Defending against poisoning attacks in federated learning with blockchain, IEEE Transactions on Artificial Intelligence, ISSN: 2691-4581
In the era of deep learning, federated learning (FL) presents a promising approach that allows multi-institutional data owners, or clients, to collaboratively train machine learning models without compromising data privacy. However, most existing FL approaches rely on a centralized server for global model aggregation, leading to a single point of failure. This makes the system vulnerable to malicious attacks when dealing with dishonest clients. In this work, we address this problem by proposing a secure and reliable FL system based on blockchain and distributed ledger technology. Our system incorporates a peer-to-peer voting mechanism and a reward-and-slash mechanism, which are powered by on-chain smart contracts, to detect and deter malicious behaviors. Both theoretical and empirical analyses are presented to demonstrate the effectiveness of the proposed approach, showing that our framework is robust against malicious client-side behaviors.
Sönmez FÖ, Knottenbelt WJ, 2024, ContractArmor: Attack Surface Generator for Smart Contracts, Pages: 8-15
This paper presents an ongoing study of a novel attack surface generator tool for smart contracts developed in Solidity. The tool leverages a rule-based engine and ChatGPT API for security analysis. The rule-based engine provides numerical values and key variables and functions for further analysis, while ChatGPT handles complex queries. However, ChatGPT may generate similar responses for more general questions, irrespective of the given contract code. The tool combines both approaches to identify and mitigate potential security vulnerabilities in Solidity-based smart contracts. The effectiveness of the tool is evaluated on real-world smart contracts, and its potential for detecting and preventing common attack vectors is demonstrated.
Xiong X, Wang Z, Chen X, et al., 2024, Leverage Staking with Liquid Staking Derivatives (LSDs): Opportunities and Risks., CoRR, Vol: abs/2401.08610
Xiong X, Wang Z, Knottenbelt W, et al., 2023, Demystifying Just-in-Time (JIT) liquidity attacks on Uniswap V3, 2023 5th Conference on Blockchain Research & Applications for Innovative Networks and Services (BRAINS), Publisher: IEEE, ISSN: 2835-3021
Uniswap is currently the most liquid DecentralizedExchange (DEX) on Ethereum. In May 2021, it upgraded to the third protocol version named Uniswap V3. The key feature update is “concentrated liquidity”, which supports liquidity provision within custom price ranges. However, this design introduces a new type of Miner Extractable Value (MEV) source called Just-in-Time (JIT) liquidity attack, where the adversary mints and burns a liquidity position right before and after a sizable swap. We begin by formally defining the JIT liquidity attack and subsequently conduct empirical measurements on Ethereum. Over a span of 20 months, we identify 36,671 such attacks, which have collectively generated profits of 7,498 ETH. Our analysis suggests that the JIT liquidity attack essentially represents a whales’ game, predominantly controlled by a select few bots. The most active bot, identified as 0xa57...6CF, has managed to amass 92% of the total profit. Furthermore, we find that this attack strategy poses significant entry barriers, as it necessitates adversaries to provide liquidity that is, on average, 269 times greater than the swap volume. In addition, our findings reveal that the JIT liquidity attack exhibits relatively poor pr ofitability, with an average Return On Investment (ROI) of merely 0.007%. We alsofind this type of attack to be detrimental to existing Liquidity Providers (LPs) within the pool, as their shares of liquidity undergo an average dilution of 85%. On the contrary, this attack proves advantageous for liquidity takers, who secure execution prices that are, on average, 0.139% better than before. We further dissect the behaviors of the top MEV bots and evaluate theirstrategies through local simulation. Our observations reveal that the most active bot, 0xa57...6CF, conducted 27% of non-optimal attacks, thereby failing to capture at least 7,766 ETH (equivalent to 16.1M USD) of the potential attack profit.
Wang Z, Cirkovic M, Le D, et al., 2023, Pay less for your privacy: towards cost-effective on-chain mixers, 5th ACM Conference on Advances in Financial Technologies (AFT 2023), Publisher: Schloss Dagstuhl -- Leibniz-Zentrum fuer Informatik, Pages: 16:1-16:25, ISSN: 1868-8969
On-chain mixers, such as Tornado Cash (TC), have become a popular privacy solution for manynon-privacy-preserving blockchain users. These mixers enable users to deposit a fixed amount ofcoins and withdraw them to another address, while effectively reducing the linkability between theseaddresses and securely obscuring their transaction history. However, the high cost of interactingwith existing on-chain mixer smart contracts prohibits standard users from using the mixer, mainlydue to the use of computationally expensive cryptographic primitives. For instance, the deposit costof TC on Ethereum is approximately 1.1m gas (i.e., 66 USD in June 2023), which is 53× higher thanissuing a base transfer transaction.In this work, we introduce the Merkle Pyramid Builder approach, to incrementally build theMerkle tree in an on-chain mixer and update the tree per batch of deposits, which can thereforedecrease the overall cost of using the mixer. Our evaluation results highlight the effectiveness ofthis approach, showcasing a significant reduction of up to 7× in the amortized cost of depositingcompared to state-of-the-art on-chain mixers. Importantly, these improvements are achieved withoutcompromising users’ privacy. Furthermore, we propose the utilization of verifiable computations toshift the responsibility of Merkle tree updates from on-chain smart contracts to off-chain clients,which can further reduce deposit costs. Additionally, our analysis demonstrates that our designsensure fairness by distributing Merkle tree update costs among clients over time.
Bor J, Casale G, Knottenbelt W, et al., 2023, Fitting with matrix exponential mixtures generated by discrete probabilistic scaling, Performance Evaluation Review, Vol: 51, Pages: 15-17, ISSN: 0163-5999
Matrix exponential (ME) distributions generalize phase-type distributions; however, their use in queueing theory is hampered by the difficulty of checking their feasibility. We propose a novel ME fitting algorithm that produces a valid distribution by construction. The ME distribution used during the fitting is a product of independent random variables that are easy to control in isolation. Consequently, the calculation of the CDF and the Mellin transform factorizes, making it possible to use these measures for the fitting without significant restriction on the distribution order. Trace-driven queueing simulations indicate that the resulting distributions yield highly accurate results.
Srinivasan P, Subramanian R, Knottenbelt W, 2023, Thinking the GOAT: imitating tennis styles, MIT Sloan Sports Analytics Conference, Publisher: MIT Sloan Sports Analytics Conference
A tactically aware coach is key to improving tennis players’ games; a coach analyses past matches with two considerations in mind: 1) the style of the player and how that style translates to real-world shot-making, and 2) the intent of a shot, irrespective of the outcome. Modern Hawk-Eye technology deployed in top-tier tournaments has enabled deeper analysis of professional matches than ever before. The aim of this paper is to emulate and augment the qualities of great coaches using data collected by Hawk-Eye; we develop a deep learning approach to imitate tennis players’ responses, to learn individual player styles efficiently, and we demonstrate this using performance metrics and illustrations.
Forshaw M, Gilly K, Knottenbelt W, et al., 2023, Preface, Communications in Computer and Information Science, Vol: 1786 CCIS, ISSN: 1865-0929
Xiong X, Wang Z, Chen X, et al., 2023, Leverage Staking with Liquid Staking Derivatives (LSDs): Opportunities and Risks., IACR Cryptol. ePrint Arch., Vol: 2023, Pages: 1842-1842
Xiong X, Wang Z, Knottenbelt WJ, et al., 2023, Demystifying Just-in-Time (JIT) Liquidity Attacks on Uniswap V3., Publisher: IEEE, Pages: 1-8
Wang Z, Xiong X, Knottenbelt WJ, 2023, Blockchain Transaction Censorship: (In)secure and (In)efficient?, Publisher: Springer, Pages: 78-94
Dong N, Wang Z, Sun J, et al., 2023, Defending Against Malicious Behaviors in Federated Learning with Blockchain., CoRR, Vol: abs/2307.00543
Wang Z, Cirkovic M, Le DV, et al., 2023, Pay Less for Your Privacy: Towards Cost-Effective On-Chain Mixers., Publisher: Schloss Dagstuhl - Leibniz-Zentrum für Informatik, Pages: 16:1-16:1
, 2023, Mathematical Research for Blockchain Economy: 4th International Conference MARBLE 2023, London, United Kingdom, July 11-13, 2023, Publisher: Springer
Wang Z, Dong N, Sun J, et al., 2023, zkFL: Zero-Knowledge Proof-based Gradient Aggregation for Federated Learning., CoRR, Vol: abs/2310.02554
Matsui T, Knottenbelt WJ, 2023, Optimal Hedge Ratio Estimation for Bitcoin Futures using Kalman Filter, 5th IEEE International Conference on Blockchain and Cryptocurrency (ICBC), Publisher: IEEE, ISSN: 2832-8892
Wang Z, Xiong X, Knottenbelt WJ, 2023, Blockchain Transaction Censorship: (In)secure and (In)efficient?, IACR Cryptol. ePrint Arch., Vol: 2023, Pages: 786-786
Li J, Veneris A, Krishnamachari B, et al., 2022, Message from the ICBC2022 General and Technical Program Chairs, IEEE International Conference on Blockchain and Cryptocurrency, ICBC 2022
Matsui T, Al-Ali A, Knottenbelt WJ, 2022, On the Dynamics of Solid, Liquid and Digital Gold Futures, 4th IEEE International Conference on Blockchain and Cryptocurrency (IEEE ICBC), Publisher: IEEE
Srinivasan P, Knottenbelt WJ, 2022, Time-series Transformer Generative Adversarial Networks., CoRR, Vol: abs/2205.11164
Pan S, Finlay C, Besenbruch C, et al., 2021, Three gaps for quantisation in learned image compression, New Trends in Image Restoration and Enhancement (NTIRE 2021) (CVPR Workshop), Publisher: IEEE
Learned lossy image compression has demonstrated impressive progress via end-to-end neural network training. However, this end-to-end training belies the fact that lossy compression is inherently not differentiable, due to the necessity of quantisation. To overcome this difficulty in training, researchers have used various approximations to the quantisation step. However, little work has studied the mechanism of quantisation approximation itself. We ad-dress this issue, identifying three gaps arising in the quantisation approximation problem. These gaps are visualised, and show the effect of applying different quantisation approximation methods. Following this analysis, we propose a Soft-STE quantisation approximation method, which closes these gaps and demonstrates better performance than other quantisation approaches on the Kodak dataset.
Koutsouri A, Petch M, Knottenbelt W, 2021, Performance of the CoinShares Gold and Cryptoassets Index under different market regimes, Cryptoeconomic Systems, Vol: 1
Regime-switching models are frequently used to explain the tendency of financial markets to change their behavior, often abruptly. Such changes usually translate to structural breaks in the average means and volatilities of financial indicators, and partition their time-series into distinct segments, each with unique statistical properties. In this paper, we address the problem of identifying the presence of such regimes in the constituents of diversified, cryptoasset-containing portfolios, ultimately to define high-risk market conditions and assess portfolio resilience. For each portfolio component, we first consider a Gaussian Hidden Markov Model (HMM) in order to extract intermediate trend-related states, conditional on the weekly returns distributions. We further apply a Markov-switching GARCH model to the demeaned daily returns to describe changes in the conditional variance dynamics and isolate volatility-related states. We combine the former approaches to generate a number of price paths for each constituent, simulate the portfolio allocation strategy and obtain a risk profile for each combination of the trend and volatility regimes. We apply the proposed method to the CoinShares Gold and Cryptoassets Index, a diversified, monthly-rebalanced index which includes two main risk-weighted components; a cryptoassets basket and physical gold. Results demonstrate an overall stable risk-reward profile when compared against the individual components and suggest a superior performance in terms of Omega ratio for investors that target wealth preservation and moderate annual returns. We detect underperformance regions in bear-low volatility market regimes, where diversification is hindered.
Ilie D, Werner S, Stewart I, et al., 2021, Unstable throughput: when the difficulty algorithm breaks, 2021 IEEE Conference on Blockchain and Cryptocurrency (ICBC 2021), Publisher: IEEE, Pages: 1-5
In Proof-of-Work blockchains, difficulty algorithms serve the crucial purpose of maintaining a stable transaction throughput by dynamically adjusting the block difficulty in response to the miners’ constantly changing computational power. Blockchains that may experience severe hash rate fluctuations need difficulty algorithms that quickly adapt the mining difficulty. However, without careful design, the system could be gamed by miners using coin-hopping strategies to manipulate the block difficulty for profit. Such miner behavior results in an unreliable system due to the unstable processing of transactions. We provide an empirical analysis of how Bitcoin Cash’s difficulty algorithm design leads to cyclicality in block solve times as a consequence of a positive feedback loop. In response, we mathematically derive a difficulty algorithm using a negative exponential filter which prohibits the formation of positive feedback and exhibits additional desirable properties, such as history agnosticism. We compare the described algorithm to that of Bitcoin Cash in a simulated mining environment and verify that the former would eliminate the severe oscillations in transaction throughput.
Koutsouri A, Petch M, Knottenbelt W, 2021, Diversification benefits of commodities for cryptoasset portfolios, 2021 International Conference on Blockchain and Cryptocurrency (ICBC 2021), Publisher: IEEE, Pages: 1-9
The aim for balance between risk and reward in investment portfolios often requires studying the diversification contribution of its constituents. This objective requires to specify whether investors can extend their exposure in certain asset classes and benefit their portfolios in a statistically significant way. In this paper, we address this issue of diversification in the context of cryptoasset portfolios and examine whether their risk-adjusted performance can be enhanced through seeking exposure into the commodities class. For an equally-weighted portfolio of five cryptoassets, we first consider the addition of physical gold, as conceptualised by the CoinShares Gold and Cryptoassets Index, a diversified, monthly-rebalanced index that seeks exposure to both asset classes. We further consider modifying the index composition by replacing physical gold with a basket of five commodities. Mean-variance spanning tests reveal that the addition of physical gold in the original cryptoasset portfolio translates to a significant shift in the efficient frontier, both in terms of the global minimum variance and the tangency portfolios. Additionally, expanding the exposure in the commodity side confirms a statistically significant improvement, with the diversification benefit arising from a shift in the tangency portfolio. We further generate a number of price paths for the original index, the modified index and their components, according to a Dynamic Conditional Correlation GARCH specification, to assess the efficiency of the index weighted risk contribution scheme. Results demonstrate a superior performance of the two indices when compared against their constituents in terms of Omega ratio. The modified index appears more appropriate for investors that seek higher annual returns, while the original composition would be more appropriate for individuals with mod
Knottenbelt W, Wolter K, 2021, Message from the Chairs, Performance Evaluation Review, Vol: 48, ISSN: 0163-5999
Jurdak R, Knottenbelt W, Krishnamachari B, 2021, Message from the ICBC2021 general and technical program chairs, IEEE International Conference on Blockchain and Cryptocurrency, ICBC 2021
Werner SM, Perez D, Gudgeon L, et al., 2021, SoK: Decentralized Finance (DeFi)., Publisher: arXiv
Decentralized Finance (DeFi), a blockchain powered peer-to-peer financial system, is mushrooming. One year ago the total value locked in DeFi systems was approximately 600m USD, now, as of January 2021, it stands at around 25bn USD. The frenetic evolution of the ecosystem makes it challenging for newcomers to gain an understanding of its basic features. In this Systematization of Knowledge (SoK), we delineate the DeFi ecosystem along its principal axes. First, we provide an overview of the DeFi primitives. Second, we classify DeFi protocols according to the type of operation they provide. We then go on to consider in detail the technical and economic security of DeFi protocols, drawing particular attention to the issues that emerge specifically in the DeFi setting. Finally, we outline the open research challenges in the ecosystem.
Ntolkeras K, Sharif H, Salmasi SD, et al., 2021, Performance Analysis of a Hyperledger Iroha Blockchain Framework Used in the UK Livestock Industry, Pages: 456-461
This investigation focuses on the performance analysis of one of the most nascent Blockchain frameworks, Hyperledger Iroha. This paper evaluates the performance of the Hyperledger Iroha framework based on three parameters: (a) total requests per seconds (RPS) over time; (b) response times in milliseconds over time; and (c) number of users in the network over time. The results indicate that the integration of Hyperledger Iroha with a major livestock management and trading platform in the UK, can support at least 200 participants with no errors in the network. Additionally, the total requests per second can reach as high as 40.6, and the response times are in the order of a fraction of a second. This research and its results can assist scholars and practitioners regarding selection of an ideal Blockchain framework for their problem setting.
Chong QZ, Knottenbelt WJ, Bhatia KK, 2021, Evaluation of Active Learning Techniques on Medical Image Classification with Unbalanced Data Distributions, 1st Workshop on Deep Generative Models for Medical Image Computing and Computer Assisted Intervention (DGM4MICCAI) / 1st MICCAI Workshop on Data Augmentation, Labelling, and Imperfections (DALI), Publisher: SPRINGER INTERNATIONAL PUBLISHING AG, Pages: 235-242, ISSN: 0302-9743
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.