The convergence of artificial intelligence and blockchain technology offers a compelling paradigm for deploying secure, auditable, and decentralisedintelligent systems in distributed environments where trust between participants cannot be assumed. Existing AI deployment frameworks lackimmutable audit trails, transparent model governance, and cryptographic integrity guarantees — requirements increasingly mandated by regulatoryframeworks including the EU AI Act and GDPR. This study presents ChainMind, a blockchain-enabled secure intelligent system frameworkintegrating smart contract-governed model lifecycle management, federated learning with on-chain gradient verification, and zero-knowledge proof(ZKP)-based inference auditing for privacy-preserving accountability. ChainMind was deployed and evaluated across three distributed intelligentsystem applications: a decentralised medical AI consortium (6 European hospitals, 284,000 patient records), a cross-border financial fraud detectionnetwork (4 banks, Germany and Italy), and a smart city data marketplace (Stuttgart urban IoT network, 12,400 sensors). ChainMind achieved modeltampering detection accuracy of 99.97%, federated learning convergence within 18.3% fewer rounds than standard FedAvg under adversarialgradient poisoning, and ZKP inference verification latency of 47.3 ms — compatible with real-time deployment. These results establish ChainMind asa technically viable and regulatory-compliant framework for blockchain-enabled secure AI in distributed environments.
Eunsung Roh, Jinheock Choi, Young-Hoon Park, Seung-Woo Seo
Deep learning networks have rapidly developed across various fields, with several high-performing models already being applied in real-world applications. However, even with identical network architectures, variations in parameter values can lead to significantly different outputs, making the security of these parameters crucial for maintaining network performance. Despite their importance, limited research has focused on securing deep learning network parameters. In recent years, distributed ledger technologies, such as blockchain and hashgraph, have been extensively studied. Their applications extend beyond cryptocurrencies to include public services, voting, supply chains, and insurance. This paper proposes a novel approach to enhancing the security of deep learning network parameters by leveraging Hedera Hashgraph, providing a more secure foundation for their deployment and commercialization. With the security offered by hashgraph, model parameters are safely stored, ensuring the reliability and trustworthiness of deep learning systems. Additionally, when these stored parameters are used for further training, such as in transfer learning, the results are expected to be reliable. We demonstrate that our proposed system secures the parameters effectively, guaranteeing both security and reliability in deep learning.
Secure machine learning paradigms have emerged as compelling solutions to address growing concerns of large-scale data collection in modern Machine Learning (ML) systems. These paradigms leverage secure computation techniques to enable the execution of ML applications without the necessity to share raw data, models or predictions to be shared between parties, offering strong, formal privacy guarantees. Recent advances have significantly enhanced both the scalability and expressiveness of these secure paradigms, facilitating their deployment in real-world scenarios across a variety of privacy-sensitive domains. However, the very mechanisms that provide these privacy guarantees also introduce new challenges to robustness, trust, and accountability. To ensure secrecy, secure ML systems conceal the processes of training and inference, making them difficult to inspect, validate, or audit. This intrinsic opacity creates a fundamental tension between privacy and accountability: hiding data and models to protect users’ privacy can also obscure failures and enable undetectable manipulation. Furthermore, in many secure ML frameworks, multiple, potentially untrusted parties collaboratively contribute to computations, thereby amplifying risks. Traditional threat models in adversarial ML often depend on transparent access to data, models, or outputs—assumptions that do not hold in secure settings. As a result, these systems become vulnerable to new and sometimes more potent attack vectors. Without dedicated integrity mechanisms, these privacy-preserving systems cannot be safely deployed in high-stakes domains such as healthcare, finance, or critical infrastructure. Realizing the full potential of secure ML requires a comprehensive understanding of the unique threats these systems face, the development of new integrity mechanisms, and their integration into these systems in a way that is efficient and preserves the privacy guarantees they provide. This dissertation advances accountability in secure ML through two complementary directions. First, it develops an understanding of the robustness challenges that arise in secure settings. We investigate the role of memorization and system-level dynamics in exposing secure systems to targeted manipulation. Based on these insights, we then introduce new cryptographic building blocks to strengthen the robustness and transparency of secure ML. We present RoFL, a system for privacy-preserving input validation in secure Federated Learning; Arc, the first framework for end-to-end auditing of secure ML pipelines; and Artemis, a new construction for generating efficient zero-knowledge proofs for real-world ML models. Together, these contributions lay the foundation for secure ML systems that are not only private, but also accountable and trustworthy in practice.
This talk given at the 2025 MDIC CM&S Summit on "Credible Models in the AI Age" traces the evolution of trust mechanisms in computational systems, from traditional model validation approaches in mechanistic modeling to emerging cryptographic verification methods for AI. We'll explore how the credibility challenge for regulators has transformed as we've moved from deterministic simulations to probabilistic AI systems, and examine how cryptographic proofs, zero-knowledge techniques, and verifiable computation are creating new pathways for establishing trust in AI outputs. By understanding this historical progression, we can better appreciate both the continuity and fundamental shifts in how we ensure reliability in our computational approaches.
Smart contract security constitutes the foundational cornerstone for ensuring the trusted operational integrity of blockchain ecosystems. In recent years, multi-task learning MTL architectures have been widely adopted in smart contract vulnerability detection, owing to their context-aware optimization and superior generalization capabilities compared to single-task learning STL frameworks. However, MTL-based approaches for smart contract vulnerability detection face two persistent challenges: 1 The negative transfer phenomenon, the mitigation of negative transfer via adaptive loss weighting in smart contract vulnerability detection remains underexplored in existing research. 2 Performance degradation caused by the homogeneous contribution assumption where undifferentiated contract representations impair expert layer learning efficacy. To overcome these limitations, we propose a novel detection framework incorporating adaptive loss weight and entropy-based feature enhancement. Our dual-weighting mechanism introduces: 1 dynamic loss coefficients that automatically balance task-specific optimization objectives based on evolving learning complexity and task significance, and 2 entropy-aware attention weights that prioritize high-information contract features during expert network training. Comprehensive evaluations on real-world smart contract datasets demonstrate the framework's superior detection performance compared to three state-of-the-art adaptive weighting baselines. Experimental results reveal significant improvements in F1-score across multiple vulnerability types, validating the effectiveness of our approach in mitigating negative transfer while maintaining robust concurrent detection capabilities. The experimental code will be systematically organized and made publicly available on GitHub shortly.
This research presents a novel approach to detecting role-based access control vulnerabilities in smart contracts using large language models (LLMs). Smart contracts deployed on blockchain platforms often contain critical security vulnerabilities that can lead to significant financial losses, with improper access control being a common issue. Our methodology combines Abstract Syntax Tree (AST) analysis with strategic context slicing to enable effective vulnerability assessment by LLMs. The system first identifies potentially vulnerable functions through structural analysis, extracts relevant security context, and then leverages LLMs to make the final vulnerability and exploitability determinations. We evaluate our approach using a dataset of 28 smart contracts across five different language models, including GPT-4 and its variants and GPT-o3-mini as well. Results demonstrate strong performance, with the best model achieving 93% accuracy and 100% precision in vulnerability detection. The system not only identifies vulnerable functions but also assesses their practical exploitability, achieving up to 100% accuracy in exploitability determination with GPT-4. The comparative analysis across different models reveals interesting trade-offs between precision and recall, with some models excelling at avoiding false positives while others prioritize catching all vulnerabilities. Our findings suggest that LLMs, when properly guided through context slicing and structured prompting, can effectively reason about complex security properties in smart contracts. This approach offers a promising direction for augmenting traditional static analysis tools with the contextual understanding capabilities of large language models, potentially improving the security of blockchain applications.
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).
Neural network inference in cloud service offers tangible benefits to users, from individuals and small institutions to large companies. However, two crucial concerns must be addressed. The first arises in satisfying the privacy of the model, the input data, and the inference results throughout the inference process. The second pertains to verifying that the inferences are derived from the designated neural network model. Although Secure Multi-Party Computation (MPC) and Zero-Knowledge Proof (ZKP) are typically adopted to mitigate such issues, the major challenge lies in achieving privacy preservation and verifiability simultaneously. In this study, we address both issues by proposing VSecNN, a verifiable and privacy-preserving neural network inference scheme. Specifically, we integrate MPC with the Zero-Knowledge Succinct Non-Interactive Argument of Knowledge (zk-SNARK) protocol to achieve zero-knowledge proof generation for multiple parties. Subsequently, we perform adaptive optimizations on the multi-party proof generation approach to align with the neural network, thereby achieving both privacy-preserving capabilities and verifiability. Experimental results demonstrate an improvement in the efficiency. For example, the computation time for completing our multi-party proof generation could be as low as 1.7 times that of the single-party proof generation, while the verification requires only 169ms on the MNIST dataset.
Vulnerability detection is particularly relevant in smart contracts, where modifying the code after deployment is impossible. Machine learning solutions provide greater efficiency than static analyzers in speed and detection. This study evaluates various classic machine-learning techniques and state-of-the-art neural networks for training a vulnerability detector. We analyze the largest and most reliably labelled dataset of smart contracts currently available, experimenting with six data representations of smart contracts and a multimodal approach. Our experiments show that both deep and traditional machine learning methods excel in different scenarios. Notably, eXtreme Gradient Boosting achieved an F1-score of 0.91 with the multimodal approach, which suggests its potential for more robust classification. At the same time, the results underscore the need for larger datasets to showcase the full potential of the evaluated methods.
Dakai Kang, Junchao Chen, Tien Tuan Anh Dinh, Mohammad Sadoghi
The rise of cryptocurrencies like Bitcoin and Ethereum has driven interest in blockchain database technology, with smart contracts enabling the growth of decentralized finance (DeFi). However, research has shown that adversaries exploit transaction ordering to extract profits through attacks like front-running, sandwich attacks, and liquidation manipulation. This issue affects blockchains where block proposers have full control over transaction ordering. To address this, a more fair transaction ordering mechanism is essential. Existing fairness protocols, such as Pompe and Themis, operate on leader-based consensus protocols, which not only suffer from low throughput caused by the single-leader bottleneck, but also allow adversarial block proposers to manipulate transaction ordering. To address these limitations, we propose a new framework, FairDAG, that runs fairness protocols on top of DAG-based consensus protocols. FairDAG improves protocol performance in both throughput and fairness quality by leveraging the multi-proposer design and validity property of DAG-based consensus protocols. We conducted a comprehensive analytical and experimental evaluation of two FairDAG variants - FairDAG-AB and FairDAG-RL. Our results demonstrate that FairDAG outperforms prior fairness protocols in both throughput and fairness quality.
Abdelrazak A. Yousef Elbunan, Nuradeen K. Emhemed Fethalla, Badriya Abdullah Altarhuni
The rapid proliferation of autonomous robotic systems, ranging from nano-drones to industrial collaborative robots (cobots), is generating massive, distributed datasets. While deep learning (DL) serves as the cornerstone of modern robotic intelligence, the conventional approach of centralizing this data for training poses insurmountable challenges related to privacy, security, bandwidth, and latency. Federated Learning (FL) has emerged as a disruptive paradigm that enables collaborative model training across distributed devices without the need for raw data exchange. However, the integration of FL into real-world robotic swarms—characterized by extreme heterogeneity, dynamic connectivity, and stringent resource constraints—introduces a unique set of complexities that extend far beyond those of conventional edge devices. This survey provides a comprehensive and critical examination of the burgeoning field of FL within robotic and autonomous systems. We move beyond a mere overview to present a novel taxonomy that classifies FL architectures for robotics based on communication topology, learning paradigm, and application criticality. A significant portion of our analysis is dedicated to the potent synergy between FL and Distributed Ledger Technologies (DLTs), particularly blockchain, for achieving decentralized trust, auditability, and robust aggregation in the presence of potentially malicious agents. We extensively review applications across perception, control, and collaborative tasks, highlighting pioneering works in multi-robot SLAM, federated reinforcement learning, and human-robot interaction. Furthermore, we identify and discuss pressing open challenges, including communication efficiency in mobile swarms, energy-aware client selection, personalized learning for non-IID data, and defense mechanisms against sophisticated adversarial attacks. This paper serves as a foundational reference for researchers and practitioners aiming to develop the next generation of private, secure, and collectively intelligent robotic systems.
Zero-knowledge (ZK) proofs have been increasingly popular in privacy-preserving applications and blockchain systems.To facilitate handy and efficient ZK proof generation for normal users, the industry has designed domain-specific languages (DSLs) and ZK compilers.Given a program in ZK DSL, a ZK compiler compiles it into a circuit, which is then passed to the prover and verifier for ZK checking.However, the correctness of ZK compilers is not well studied, and recent works have shown that de facto ZK compilers are buggy, which can allow malicious users to generate invalid proofs that are accepted by the verifier, causing security breaches and financial losses in cryptocurrency.In this paper, we propose MTZK, a metamorphic testing framework to test ZK compilers and uncover incorrect compilations.Our approach leverages deliberately designed metamorphic relations (MRs) to mutate ZK compiler inputs.This way, ZK compilers can be automatically tested for compilation correctness using inputs and mutated variants without requiring manual intervention.We propose a set of design considerations and optimizations to deliver an efficient and effective testing framework.In the evaluation of four industrial ZK compilers, we successfully uncovered 21 bugs, out of which the developers have promptly patched 15.We also show possible exploitations of the uncovered bugs to demonstrate their severe security implications.
Proof-of-Stake (PoS) blockchains often select committees in direct proportion to stake, which makes security sensitive to large validators and stake concentration. In such settings, a purely stake-based lottery can sometimes produce committees whose adversarial share crosses the safety threshold, even if the global adversarial stake remains below one third. This paper introduces a simple mixed-weight rule that combines stake with a bounded baseline distribution through a single mixing parameter λ. The rule leaves committee size, rewards, and VRF-based sortition unchanged, but pulls weight away from highly concentrated positions. Proved that, whenever the adversary is more concentrated than the baseline, the expected adversarial seats fall linearly in λ, while standard concentration bounds show an exponential drop in committee-capture probability. While the mechanism relies on entity-level attribution (or high-cost identities) to prevent Sybil attacks, experiments on ten production PoS networks indicate that modest mixing (around λ=0.3) reduces expected adversarial seats by about one quarter and tightens worst-case guarantees by orders of magnitude.
Open access
Blockchain Technology Applications and Security
Adversarial Robustness in Machine Learning
Infrastructure Resilience and Vulnerability Analysis