With the widespread deployment of Deep-Learning-as-a-Service, secure multi-party computation-based outsourcing neural network (NN) inference has garnered significant attention for its high-security guarantee. Nevertheless, under the dishonest-majority setting with malicious adversaries, prior secure inference works are still costly in terms of communication and run-time. Additionally, existing outsourcing frameworks impose a substantial client-side design, which leads to obstacles in resource-constrained devices. To address the above challenges, we propose MD-SONIC, an online efficient and maliciously-secure framework for outsourcing NN inference with a dishonest majority. We first construct communication-efficient n-party protocols for the basic primitives such as fixed-point multiplication and most significant bit extraction by combining mask-sharing and TinyOT-sharing with SPD$\mathbb {Z}_{2^{k}}$seamlessly. Then, we build fast secure blocks for the widely used NN operators, including matrix multiplication, ReLU, and Maxpool, on top of our basic primitives. To enable an arbitrary number of users to outsource the secure inference task to n computing servers, we propose a lightweight-client and fast$\Sigma $paradigm named SPIN, stemming from zero-knowledge proofs. Our SPIN can be instantiated into a set of efficient outsourcing protocols over multiple algebraic structures (e.g., finite field and ring). We also conduct extensive evaluations of MD-SONIC on various neural networks. Compared to the work by Damgård et al. (IEEE S&P’19) and MD-ML (USENIX Security’24), we achieve up to$594.4\times $and$45.1\times $online communication improvements, and improve the online execution time by at most$14.3\times $(resp.$20.5\times $) and$1.8\times $(resp.$2.3\times $) in LAN (resp. WAN).
Mantri Christ Elison, Martin Victor K, Gifton Paul Immanuel
The objective of this research is to develop an R&D (Research and Development) for the hardiness relay alert system, including applying the machine learning, and the fuzzy logic networks for the real time Ethereum transaction 'match failure' detection and the improved Ethereum blockchain security.As an example, the system is computing on the transactions due to the fact the system for transaction analysis corresponds with concrete intrinsic characteristics and thus it mainly takes out suspicious or malicious transactions.The logistic regression, support vector machines (SVM) decision tree and random forests are used in this research and optimized by grid search.Finally, on the other hand, uncertainty problems and false alarms are solved where fuzzy membership functions are used to put transaction attributes into linguistic hobbled variables (such as 'low', 'medium' and 'high').The conclusion of this descriptive research is that fuzzy logic integration with machine learning can improve the approach of anomaly mediation compared to the rules based approach and it is superior to rules based approach.Finally, the effectiveness of the models is detailed and replicated in various graphical representations of the decision making process and membership functions to show that the system can be deployed in real time to secure blockchain networks.
Inderscience is a global company, a dynamic leading independent journal publisher disseminates the latest research across the broad fields of science, engineering and technology; management, public and business administration; environment, ecological economics and sustainable development; computing, ICT and internet/web services, and related areas.
This paper proposes a secure data-sharing model that utilises a blockchain-based secure framework and a deep learning-based intrusion detection model to ensure patient privacy and provide personalised healthcare services. The proposed model consists of two phases: validation and verification. In the validation phase, electronic health record (EHR) data is uploaded to an Ethereum blockchain, encrypted using improved elliptic curve cryptography (Imp-ECC), and stored in an interplanetary file system (IPFS) within the blockchain. In the verification phase, an optimised deep-learning approach, enhanced capsule-BiLSTM, is used to detect unauthorised users in the network. If an attack is detected, access is denied; otherwise, the user is authorised to access the encrypted data. The proposed model is evaluated using two datasets, EHR and UNSW-NB15. The results show that the proposed model achieves a less encryption time of 198 seconds for the EHR dataset and an accuracy of 97.19% for the UNSW-NB15 dataset.
Mosbah Alown, Mehmet Sabır Kiraz, Muhammed Ali Bingöl
Electronic voting (e-voting) systems have significantly improved the traditional voting process by addressing key concerns such as security, public acceptability, and convenience. However, these systems often face unique challenges, such as ensuring voter privacy and verifiability, preventing coercion and double voting, and maintaining scalability while protecting participant confidentiality. This study critically analyses and compares various e-voting schemes and technologies, evaluating their security features, verifiability mechanisms, and potential vulnerabilities. This paper reviews Direct Recording Electronic (DRE) voting, internet voting, and blockchain-based e-voting systems. In so doing, we provide an understanding of cryptographic primitives employed in e-voting systems and how they address specific characteristics and challenges associated with each voting scheme. Furthermore, we examine the applications proposed by previous studies in the context of these voting systems, assessing their strengths, limitations, and impact on democratic procedures. The cryptographic primitives reviewed include techniques like homomorphic encryption, blind signatures, and zero-knowledge proofs, which can enhance voter privacy, verifiability, and resistance to coercion and double voting.
Terrence August, Duy Dao, Kihoon Kim, Marius Florin Niculescu
Cryptocurrencies have prompted a shift away from classic security attacks toward ransomware-based extortion. To better understand the impact of cryptocurrencies on the cybersecurity landscape, we conduct a comparative analysis of cybersecurity metrics prior to and after the adoption of cryptocurrency using a series of connected software-use models in the presence of security externalities. In this framework, we endogenize the actions of both heterogeneous consumers and attackers, with entry of the latter being driven by both the size of the unpatched consumer population and, as a subset of it, the size of the ransom-paying consumer population. We first examine users’ adoption and patching behavior under both security scenarios. We explore how changes in attacker entry costs impact outcomes under both conventional and post-crypto ransomware threat landscapes. We show that ransomware scenarios may be more desirable than conventional ones when attacker entry costs are low, provided that the gains from entering with standard attacks under the ransomware scenario are not too high. However, under such scenarios, social welfare can increase under the same conditions that lead to larger ransoms being demanded and a higher expected total ransom being paid, which presents a conundrum to policymakers. We also examine the impact of market parameters associated with security losses from conventional attacks and residual losses when victims pay in ransomware attacks. This paper was accepted by Kay Giesecke, finance. Funding: This work was partially supported by Insung Research Grant of KUBS, the LG Yonam Foundation (of Korea), and an award from the Georgia Institute of Technology Center of International Business Education and Research as part of its funded research program. Supplemental Material: The online appendices are available at https://doi.org/10.1287/mnsc.2023.00969 .
Decentralized finance (DeFi) protocols are crypto projects developed on the blockchain to manage digital assets. Attacks on DeFi have been frequent and have resulted in losses exceeding $80 billion. Current tools detect and locate possible vulnerabilities in contracts by analyzing the state changes that may occur during malicious events. However, this victim-only approaches seldom possess the capability to cover the attacker’s interaction intention logic. Furthermore, only a minuscule percentage of DeFi protocols experience attacks in real-world scenarios, which poses a significant challenge for these detection tools to demonstrate practical effectiveness. In this paper, we propose DeFiTail, thefirstframework that utilizes deep learning technology for access control and flash loan exploit detection. Through feeding the cross-contract static data flow, DeFiTail automatically learns the attack logic in real-world malicious events that occur on DeFi protocols, capturing the threat patterns between attacker and victim contracts. Since the DeFi protocol events involve interactions with multi-account transactions, the execution path with external and internal transactions requires to be unified. Moreover, to mitigate the impact of mistakes in Control Flow Graph (CFG) connections, DeFiTail validates the data path by employing the symbolic execution stack. Furthermore, we feed the data paths through our model to achieve the inspection of DeFi protocols. Comparative experiment results indicate that DeFiTail achieves the highest accuracy, with 98.39% in access control and 97.43% in flash loan exploits. DeFiTail also demonstrates an enhanced capability to detect malicious contracts, identifying 86.67% accuracy from the CVE dataset. By monitoring existing contracts, we identified five distinct categories of vulnerabilities: repetition abuse, unsafe unintended exploitation, signature violated exploitation, insecure interfaces exploitation, and unrestricted token transfer.
Zubaida Rehman, Iqbal Gondal, Hai Dong, Mengmeng Ge · 6 authors
Eclipse attacks, which isolate victim nodes by monopolizing their peer connections, remain a critical threat to Ethereum’s consensus mechanism. To address this, we present a principled framework for detecting Eclipse attacks in Ethereum peer-to-peer networks, grounded in a formal adversarial model. Existing defenses are either ad-hoc or lack provable guarantees, leaving open questions about their reliability under adaptive adversaries. Our work aims to bridge this gap by formally defining eclipse attack detection as a security property. We specify soundness, completeness, and robustness theorems under bounded adversarial drift, and derive formal guarantees within false positive and false negative bounds, resilience to adversarial manipulation, and multi-node compositional reliability. We then instantiate a lightweight detection framework that maps packet-level traffic features to predictions using ensemble classifiers (Random Forest, XGBoost). The system was validated using a controlled Ethereum testbed and extended with CTGAN-generated synthetic traces to emulate networks of up to 100 nodes. Empirical evaluation shows that our framework achieves up to 96% F1-score with sub-second inference latency, well within Ethereum’s 12-second Proof-of-Stake validator time slots. These findings demonstrate that lightweight statistical features, when coupled with formal analysis, enable accurate, efficient, and scalable detection of network-level partitioning attacks. Our work establishes a deployable and theoretically grounded defense foundation for securing modern blockchain systems against eclipse adversaries.
Zheng Che, Meng Shen, Zhehui Tan, Hanbiao Du · 9 authors
With the rapid evolution of Web3.0, cryptocurrency has become a cornerstone of decentralized finance. While these digital assets enable efficient and borderless financial transactions, their pseudonymous nature has also attracted malicious activities such as money laundering, fraud, and other financial crimes. Effective detection of malicious accounts is crucial to maintaining the security and integrity of the Web 3.0 ecosystem. Existing malicious account detection methods rely on large amounts of labeled data and suffer from low generalization. Label-efficient and generalizable malicious account detection remains a challenging task. In this paper, we propose ShadowEyes, a framework for detecting malicious accounts by leveraging interaction feature learning with only a small labeled dataset. Specifically, We first propose a generalized account representation named TxGraph, which captures the universal interaction features of Ethereum and Bitcoin. Then we carefully design an account representation augmentation method tailored to simulate the evolution of malicious accounts to generate positive pairs. We conduct extensive experiments using public datasets to evaluate the performance of ShadowEyes. The results demonstrate that it outperforms state-of-the-art (SOTA) methods in four typical scenarios. Specifically, in the scenario of acrossplatform malicious account detection, ShadowEyes maintains an F1 score of around 90%, which is 10% higher than the SOTA method. In the zero-shot learning scenario, it can achieve an F1 score of 79.56% for detecting gambling accounts, surpassing the SOTA method by 10.44%.
Blockchain is a prominent and leading decentralized ledger technology that has gained global attention and adoption across various industries. Long-range attacks (LRAs) are when an adversary attempts to rewrite the blockchain’s history from a point far back in time. Since PoS Blockchain relies on validators’ stakes as a form of security, LRAs can potentially undermine the network’s security if not detected and prevented. In order to protect against long-range attacks, this research suggests a high-performance explainable neural network model that can accurately categorize nodes as malicious or non-malicious while maintaining interpretability. The proposed explainable neural network model includes Residual Auto Encoder (RAE) guided generalized additive models with incorporating structured interactions (RAE-GAMI-Net) for LRA detection in PoS Blockchain In this work, a wrapper-based Binary Orchard Algorithm (W-BOA) is used to find the best features to lessen the dimensionality of extracted Characteristics, and a global feature extraction has been implemented based on multi-scale Densenet (MDensenet) that assures early convergence and optimal performance by providing global optimal solution. Then, the transformed features are used to train the RAE-GAMI-Net-based model to detect the LR attack. The included RAE learns a compressed representation (latent) of the input features. Then, the latent features are classified with GAMI-Net, balancing the model interpretability and accuracy. The effectiveness of our proposed method is assessed using the Proof of Stake blockchain dataset and benchmarked against other deep learning techniques. Our approach yields significant enhancements in accuracy (0.962), precision (0.9614), and recall 0.9604, accompanied by a notably low Brier score of 0.038.
Developers and users are drawn to Ethereum due to its rapidly growing asset count. However, potential vulnerabilities and malicious behaviors during the execution of smart contracts have led to an increasing demand for security detection technology. Conventional static and dynamic analysis methods are less useful in the case of complex opcode sequences and multiple execution paths. To tackle this problem, this paper proposes an Ethereum intrusion detection method based on Bidirectional Long Short-Term Memory (Bi-LSTM) network with multi-head attention. It examines the opcode execution paths generated from the intra-and-inter-function Control Flow Graphs (CFGs) using the EPP algorithm and captures the rich feature representations and long dependencies. This combination increases the precision and efficacy of detecting malicious activity and smart contract vulnerabilities while simultaneously enhancing the model’s robustness and interpretability and handling variable-length sequences. For the five selected vulnerabilities, the precision, recall and F1-score of this model are above 89.9%, 87.3%, and 88%, respectively.
Hanbiao Du, Meng Shen, Yang Liu, Zheng Che · 7 authors
Ethereum serves as the cornerstone for value transfer in Web 3.0, providing a decentralized and efficient trust mechanism for global connectivity. However, the anonymity of Ethereum undermines market regulatory capabilities, leading to frequent malicious behaviors such as Ponzi Scheme, Money Laundering, and Phishing. Therefore, in the face of the diverse and continuously emerging malicious behaviors, implementing fine-grained detection is crucial for maintaining the prosperous development of the blockchain ecosystem. In this paper, we propose FiMAD, a fine-grained and class-incremental malicious account detection framework based on dynamic graph learning. Specifically, we first propose a general graph structure calledDynamic Account Relation Graph (DARG), which dynamically models Ethereum accounts from a continuous-time perspective. Then, we design a cascade graph feature extraction method to capture deep temporal evolution patterns and neighbor interaction features in DARG. Next, we construct a pre-training universal encoder to transform account features into high-dimensional embeddings, followed by fine-tuning the model classifier with a few labeled samples, enabling accurate fine-grained detection and rapid updates for incremental classes. We conduct extensive experiments using real Ethereum data. The results demonstrate that FiMAD outperforms state-of-the-art (SOTA) methods in fine-grained detection across five typical scenarios: class-incremental, full data, new malicious accounts, imbalanced data, and binary classification. In the class-incremental scenario, FiMAD improves the Macro-F1 by up to 26.4% compared to SOTA methods.