The rapid growth of data-driven applications in healthcare, finance, IoT, and autonomous systems has created a pressing need for privacy-preserving and scalable machine learning methods. Traditional centralized learning, which aggregates data into a single repository, faces challenges related to data privacy, security, communication overhead, and regulatory compliance. Federated Learning (FL) offers a decentralized solution, enabling multiple clients to collaboratively train a global model without sharing raw data. Only model updates are exchanged, preserving privacy while leveraging distributed computational resources. This paper reviews FL architectures— including centralized, decentralized, horizontal, vertical, cross-device, and cross-silo—along with core components such as local clients, central servers, and communication protocols. Privacy- preserving techniques like differential privacy, secure aggregation, homomorphic encryption, and anonymization/pseudonymization are discussed to protect sensitive information. FL applications span healthcare, finance, IoT, smart devices, and autonomous systems, highlighting its transformative potential. Key challenges include data and system heterogeneity, efficient aggregation, personalization, robustness, and regulatory compliance. Future directions focus on enhanced privacy, communication efficiency, model personalization, and integration with edge and IoT environments. FL thus represents a promising paradigm for secure, collaborative, and distributed artificial intelligence.
ABSTRACT Blockchain technology has emerged as a pivotal solution for securing sensitive data across various domains, including artificial intelligence (AI), supply chain management, cloud computing, and healthcare. Its core attributes, confidentiality, decentralization, security, and privacy, offer significant advantages to the healthcare sector. The integration of Internet of Things ( IoT ) devices within healthcare systems enhances interoperability, enabling seamless communication between healthcare software and IT infrastructure. However, traditional healthcare systems face persistent security challenges, including phishing, masquerading, and identity theft. To address these issues, we propose a secure blockchain‐based decentralized application for generating, maintaining, and validating medical certificates. This application facilitates secure interactions among healthcare entities, including hospitals, patients, and IoT devices, while ensuring confidentiality, authentication, and access control through smart contracts. The proposed blockchain architecture enhances data integrity and secure transmission using the zero‐knowledge proof ( ZKP ) mechanism. Additionally, we incorporate the interplanetary file system ( IPFS ) for off‐chain data storage to optimize storage costs and enhance security through Ethereum smart contracts. Performance evaluations demonstrate the effectiveness of our approach in mitigating existing security vulnerabilities, thereby offering a robust and scalable solution for secure healthcare data management.
This paper presents a privacy-preserving intrusion detection architecture tailored for smart home environments, addressing the dual challenge of maintaining data confidentiality while enabling accurate anomaly detection. The proposed system replaces conventional raw data analysis with a proof-driven mechanism leveraging Zero-Knowledge Proofs (ZKPs). Behavioral patterns from smart devices such as motion sensors, door contacts, and environmental monitors are abstracted into cryptographic representations, which are then processed by a zk-SNARK-compatible machine learning model. Inference results are accompanied by cryptographic proofs verifying the correctness of each decision without disclosing the input data. A private blockchain layer, implemented using Ethereum smart contracts, records event hashes, proof metadata, and decision outcomes to ensure tamper-evident logging and automated response handling. Experimental simulations on synthetic home automation datasets demonstrate that the architecture achieves over 92% anomaly detection accuracy while ensuring zero exposure of raw sensor streams. The system also exhibits low-latency proof generation (~400 ms) and end-to-end response time under 1.2 seconds, confirming its suitability for real-time smart home applications.
With the growth of digital financial systems, robust security and privacy have become a concern for financial institutions. Even though traditional machine learning models have shown to be effective in fraud detections, they often compromise user data by requiring centralized access to sensitive information. In IoT-enabled financial endpoints such as ATMs and POS Systems that regularly produce sensitive data that is sent over the network. Federated Learning (FL) offers a privacy-preserving, decentralized model training across institutions without sharing raw data. FL enables cross-silo collaboration among banks while also using cross-device learning on IoT endpoints. This survey explores the role of FL in enhancing financial security and introduces a novel classification of its applications based on regulatory and compliance exposure levels— ranging from low-exposure tasks such as collaborative portfolio optimization [16] to high-exposure tasks like real-time fraud detection [7], [8]. Unlike prior surveys, this work reviews FL’s practical use within financial systems, discussing its regulatory compliance and recent successes in fraud prevention and blockchainintegrated frameworks. However, FL’s deployment in finance is not without challenges. Data heterogeneity, adversarial attacks, and regulatory compliance make implementation far from easy. This survey reviews current defense mechanisms and discusses future directions, including blockchain integration, differential privacy, secure multi-party computation, and quantum-secure frameworks. Ultimately, this work aims to be a resource for researchers exploring FL’s potential to advance secure, privacycompliant financial systems.
Traceable Ring Signatures (TRS) were introduced by Fujisaki and Suzuki~[PKC'07], where a trace algorithm can publicly check if two signatures with the same event label were generated by the same signer (linkability). In addition, if the two signatures correspond to different messages, then the signer's identity is revealed (traceability). Following [PKC'07], most subsequent works adopt the same definitions and consider three security properties, anonymity, linkability, and exculpability. [PKC'07] proved that the latter two properties together imply unforgeability, a fundamental requirement for all signature-like primitives. ~~~~In this work, we identify a gap in the aforementioned proof, which arises from the insufficient consideration of linkability and exculpability in [PKC'07]. To address this, we revisit the syntax and security notions of TRS, and close this gap by defining extended linkability and extended exculpability. Building on these, we design a new framework of TRS from PseudoRandom Functions (PRF) and Zero-Knowledge Proofs of Knowledge (ZKPoK) that supports tracing, provided that both two signatures are valid. This constitutes a substantial improvement over existing approaches---all of which require tracing with the size of the ring---and elevates TRS to a level of practicality and efficiency comparable to Linkable Ring Signatures (LRS), which have already achieved widespread deployment in practice. Finally, we instantiate our generic framework from the DDH assumption and leverage the Bulletproofs [S\&P'18] to construct a TRS scheme with log-size signatures. The proposed scheme achieves highly optimized signature sizes in practice and remains compatible with most existing DLog-based systems. On Curve25519, the signature size is bytes, which to our best knowledge is the shortest LRS scheme for a ring .
Blind signatures have become a cornerstone for privacy-sensitive applications such as digital cash, anonymous credentials, and electronic voting. The elliptic curve variant of the Digital Signature Algorithm (ECDSA) is widely adopted due to its efficiency in resource-constrained environments, such as mobile devices and blockchain systems. Building blind ECDSA is hence a natural goal. One presents the first such construction relying solely on the ECDSA assumption. Despite the inherent complexities in integrating blindness with ECDSA, we design a protocol that ensures both unforgeability and blindness without introducing new computational assumptions and ensuring concurrent security. It involves zero-knowledge proofs based on the MPC-in-the-head paradigm for complex statements combining relations on encrypted elliptic curve points, their coordinates, and discrete logarithms.
Russell W. F. Lai, Monisha Swarnakar, Ivy K. Y. Woo
The Learning with Errors (LWE) problem asks to distinguish noisy samples s^T A + e^T mod q from uniformly random values given the random matrix A. In this work, we show that a variant called Leaky LWE, where the distinguisher receives additionally noisy leakages (s^T, e^T) L + f^T of the LWE secret s and error e for low-norm matrix L chosen adaptively by the distinguisher after seeing A, is not easier than the standard LWE of the same dimensions up to polynomial losses in the noise level and the modulus. More generally, we show that the Leaky LWE problem is hard even if the public matrix A is structured and/or hinted and if the non-leaky parts of the secret and error do not follow Gaussian distributions, as long as the corresponding LWE problem without leakage is hard. Our reduction from LWE to Leaky LWE unifies and extends prior results on the Error-Leakage LWE problem [Döttling-Kolonelos-Lai-Lin-Malavolta-Rahimi, EUROCRYPT'23], where L only acts on the error e and the Hint-MLWE problem [Kim-Lee-Seo-Song, CRYPTO'23], where L is restricted to concatenations of random Gaussian scalar matrices not controlled by the distinguisher. Previously, the Hint-MLWE and Error-Leakage LWE assumptions were used as computational replacements of the statistical noise flooding technique in security proofs which led to improved parameters in lattice-based cryptographic constructions such as zero-knowledge proofs, threshold signatures and registration-based encryption. We provide lemmas which abstract out such computational arguments based on Leaky LWE.
An anonymous credential (AC) system with partial disclosure allows users to prove possession of a credential issued by an issuer while selectively disclosing a subset of their attributes to a verifier in a privacy-preserving manner. In keyed-verification AC (KVAC) systems, the issuer and verifier share a secret key. Existing KVAC schemes rely on computationally expensive zero-knowledge proofs during credential presentation, with the presentation size growing linearly with the number of attributes. In this work, we propose two highly efficient KVAC constructions that eliminate the need for zero-knowledge proofs during the credential presentation and achieve constant-size presentations. Our first construction adapts the approach of Fuchsbauer, Hanser and Slamanig (JoC'19), which achieved constant-size credential presentation in a publicly verifiable setting using their proposed structure-preserving signatures on equivalence classes (SPS-EQ) and set commitment schemes, to the KVAC setting. We introduce structure-preserving message authentication codes on equivalence classes (SP-MAC-EQ) and designated-verifier set commitments (DVSC), resulting in a KVAC system with constant-size credentials (2 group elements) and presentations (5 group elements). To avoid the bilinear groups and pairing operations required by SP-MAC-EQ, our second construction uses a homomorphic MAC with a simplified DVSC. While this sacrifices constant-size credentials (n+2 group elements, where n is the number of attributes), it retains constant-size presentations (2 group elements) in a pairingless setting. We formally prove the security of both constructions and provide open-source implementation results demonstrating their practicality. We extensively benchmarked our KVAC protocols and, additionally, bechmarked the efficiency of our SP-MAC-EQ scheme against the original SPS-EQ scheme, showcasing significant performance improvements.
In this paper, we show for the first time it is practical to privately delegate proof generation of zkSNARKs to a single server for computations of up to 2^20 R1CS constraints. We achieve this by computing zkSNARK proof generation over homomorphic ciphertexts, an approach we call blind zkSNARKs. We formalize the concept of blind proofs, analyze their cryptographic properties and show that the resulting blind zkSNARKs remain sound when compiled using BCS compilation. Our work follows the framework proposed by Garg et al. (Crypto'24) and improves the instantiation presented by Aranha et al. (Asiacrypt'24), which implements only the FRI subprotocol. By delegating proof generation, we are able to reduce client computation time from 10 minutes to mere seconds, while server computation time remains limited to 20 minutes. We also propose a practical construction for vCOED supporting constraint sizes four orders of magnitude larger than the current state-of-the-art verifiable FHE-based approaches. These results are achieved by optimizing Fractal for the GBFV homomorphic encryption scheme, including a novel method for making homomorphic NTT evaluation packing-friendly by computing it in two dimensions. Furthermore, we make the proofs publicly verifiable by appending a zero-knowledge Proof of Decryption (PoD). We propose a new construction for PoDs optimized for low proof generation time, exploiting modulus and ring switching in GBFV and using the Schwartz-Zippel lemma for proof batching; these techniques might be of independent interest. Finally, we implement the latter protocol in C and report on execution time and proof sizes.
The industrial internet of things (IIoT) expanded fast as physical devices and systems were connected to the internet. However, this interconnectedness made IIoT systems vulnerable to hackers. Intrusion detection systems (IDSs) were put in place to detect and prevent such assaults. Nonetheless, attackers might circumvent IDSs by forging identities or interfering with recorded data. The article intended to improve IIoT security by achieving system confidentiality, integrity, availability, scalability, performance, and security. For IIoT security, the article developed a secure federated learning access control framework (SecureFLACF) linked with a blockchain-based IDS. SecureFLACF used blockchain to secure data collected by IDS, AES-256 encryption to secure stored data, zero-knowledge proof (ZKP) to validate user identities and manage data access, and a federated learning access control framework (FLACF) to train a machine learning model for intrusion detection. SecureFLACF developed as a viable solution for improving IIoT security, providing strong assurances for IDS data and access control using blockchain’s tamper-proof structure and AES-256 encryption. Furthermore, FLACF’s design allows private machine learning model training, ensuring data privacy as well as model fidelity. The framework’s usefulness was highlighted by its application in real-world circumstances, making it a cost-effective option for organisations of all sizes. This method not only strengthened IIoT systems against a wide range of cyber threats, but also stressed their dependability as a safeguard. SecureFLACF exhibited considerable promise for improving IIoT security across several dimensions by encapsulating practicability, cost-effectiveness, and dependability.
Eugenio Felipe Merlano, Glenn Parry, Abubakar-Sadiq Shehu, Steve Schneider
Global labour markets face significant disruption from the rapid advance of artificial intelligence (AI) and automation. Digital or gig economy workers, like freelancers and online independent contractors, are more exposed to the disruptive impacts of technological changes due to their flexible working conditions, which often come with flexible contracts, less robust legal agreements and other unstable working conditions. This study explores how gig economy workers benefit from alternative privacy-enhancing decentralised reputation systems and technologies that enable them to manage information like education, certifications, credentials, and professional experience, both by collecting and sharing information with employers. We propose a blockchain framework comprising three components: (1) Self-sovereign identity (SSI) enabling cryptographically secured, portable control over credentials via decentralised storage; (2) Immutable reputation registries leveraging consensus mechanisms to secure tamper-proof work histories; and (3) Privacy-preserving signalling using zero-knowledge proofs (ZKPs) to let workers selectively disclose reputation metrics without revealing sensitive details. We combine Signalling Theory (ST) and the Unified Theory of Acceptance and Use of Technology (UTAUT) to empirically assess real workers’ intentions to use this type of decentralised reputation system. Our framework enhances transparency, worker autonomy, and privacy in the digital economy.
Existing regulatory frameworks for decentralized anonymous payments help combat illicit activities such as money laundering and terrorism financing. However, With the development of the European Union's General Data Protection Regulation (GDPR) and data sovereignty, Existing frameworks struggle to balance privacy with regulatory compliance, often compromising user autonomy and data sovereignty. To address this, we propose the Linkable Distributed Regulatory Tag (LDRT) scheme, which enables traceability without altering transaction structure by leveraging Paillier homomorphic encryption and secret sharing. Building upon this, we introduce Decentralized Anonymous Payment with Data Sovereignty(DAPDS), the first regulatory framework for decentralized anonymous payments that complies with data sovereignty. DAPDS supports both UTXO and account-based models and employs a distributed regulator group with incentives to prevent dishonest behavior. We define and analyze the security properties of both LDRT and DAPDS, proving anonymity, linkability, traceability, collusion resistance, data sovereignty, distributed regulation, and fairness of incentive mechanism. Our work provides a valuable reference for future regulatory framework paradigms for anonymous privacy-preserving traceable blockchain. Experimental results show that DAPDS only incurs an additional 34ms of time cost over ETH in transacation phase and an average of only 2.66s per tracing request in tracing phase.
Love Allen Chijioke Ahakonye, Cosmas Ifeanyi Nwakanma, Jae Min Lee, Dong‐Seong Kim
The growing complexity of distributed industrial IoT systems heightens cybersecurity risks, exposing the limitations of centralized ML-based intrusion detection. Federated Learning (FL) enables decentralized, privacy-preserving model training but remains susceptible to adversarial threats and system-level failures. This study introduces PureChain, a decentralized ledger using a proof-of-authority and association (PoA 2 ) consensus mechanism to enhance FL-based IDS security. The study offers insight into the mathematical model of the PureChain-enhanced FL, which integrates blockchain-inspired consensus protocols for collaborative intrusion detection across organizations, ensuring data privacy while providing tamper-proof logs and automated responses through smart contracts. It incorporates dynamic fault tolerance, poisoning resistance, and privacy preservation with FL, enhancing security and performance in decentralized systems. Experimentation with varying client subsets demonstrates its adaptability with a TPS range of 312 . 5 − 1178 . 3 and a low latency range of 0 . 0008484 − 0 . 0032 . The framework ensures comprehensive security, reliability, and privacy, providing a scalable solution for decentralized, secure systems.
Advances in differential equations and control theory are reshaping how secure, efficient medical data-exchange systems are designed. In parallel, blockchain offers decentralized trust, cryptographic integrity, and auditable access control for healthcare networks. Yet the choice of storage and transmission architecture strongly affects scalability, latency, privacy, and cost. This work investigates how mathematical modeling via differential equations and modern control processes can be coupled with blockchain to strengthen security and interoperability across distributed healthcare systems. We comparatively examine three deployment models: (1) on-chain storage, (2) off-chain, cloud-backed storage with blockchain access control, and (3) local institutional storage integrated with federated learning. On-chain designs maximize transparency and tamper-resistance but incur substantial computation and storage overhead. Off-chain approaches improve scalability while retaining verifiable control through the ledger. Local storage with federated learning safeguards patient privacy by keeping raw data within institutions and sharing only encrypted updates or proofs on chain. Persistent challenges include storage bloat, network delays, heterogeneous regulations, and evolving attack surfaces. To address these issues, we outline optimization strategies grounded in system dynamics stability analysis, resource allocation, and control-oriented tuning to balance throughput, privacy, and reliability. The study synthesizes theoretical insights with implementation considerations, offering a unified perspective on building resilient, performant, and privacy-preserving medical data-exchange frameworks that leverage blockchain under mathematically principled control.
The rapid progress of large-scale models, including foundational and generative, brings to the forefront the tension between data-driven innovation and core privacy concerns. Such contracts as the GDPR and the undue privacy threats of data aggregation make centralized training approaches less desirable. To analyze the data’s distributed characteristics and their application to FLO, we investigate the role of federation analytics in a plausible paradigm that shunts data. In this paper, we present a new federated learning (FL) framework enhanced with cutting-edge privacy technologies (PET) such as Differential privacy for user-level formal guarantees of confidentiality, and strengthened secure Multi-Party Computation (SMPC), which guards the model updates. This paper studies more recent approaches to resolving the principal challenges of FL: statistical heterogeneity, communication bottlenecks, and vulnerability to adversarial attacks. We greatly appreciate what this new method portends, especially for training large language models (LLMs) and the more delicate areas of healthcare and finance. By evaluating certain existing limitations, such as the complexities of federated fine- tuning and model fairness, it is clear that an architecture with exemplary performance in FL serves as a model for scalable, secure, and privacy cop.
Privacy-preserving technologies have introduced a paradigm shift that allows for realizable secure computing in real-world systems. The significant barrier to the practical adoption of these primitives is the computational and communication overhead that is incurred when applied at scale. In this paper, we present an overview of our efforts to bridge the gap between this overhead and practicality for privacy-preserving learning systems using multi-party computation (MPC), zero-knowledge proofs (ZKPs), and fully homomorphic encryption (FHE). Through meticulous hardware/software/algorithm co-design, we show progress towards enabling LLM-scale applications in privacy-preserving settings. We demonstrate the efficacy of our solutions in several contexts, including DNN IP ownership, ethical LLM usage enforcement, and transformer inference.
, a novel, multi-layered framework designed to overcome these critical limitations in the Medical IoT domain. Med-Q Ledger integrates a permissioned Hyperledger Fabric for transactional integrity with a scalable Holochain Distributed Hash Table for high-volume telemetry, achieving horizontal scalability and sub-second commit times. To fortify long-term data security, the framework incorporates post-quantum cryptography (PQC), specifically CRYSTALS-Di lithium signatures and Kyber Key Encapsulation Mechanisms. Real-time, privacy-preserving intelligence is delivered through an edge-based federated learning (FL) model, utilizing lightweight autoencoders for anomaly detection on encrypted gradients. We validate Med-Q Ledger's efficacy through a critical application: the prediction of intestinal complications like necrotizing enterocolitis (NEC) in preterm infants, a condition frequently necessitating emergency colostomy. By processing physiological data from maternal wearable sensors and infant intestinal images, our integrated Random Forest model demonstrates superior performance in predicting colostomy necessity. Experimental evaluations reveal a throughput of approximately 3400 transactions per second (TPS) with ~180 ms end-to-end latency, a >95% anomaly detection rate with <2% false positives, and an 11% computational overhead for PQC on resource-constrained devices. Furthermore, our results show a 0.90 F1-score for colostomy prediction, a 25% reduction in emergency surgeries, and 31% lower energy consumption compared to MQTT baselines. Med-Q Ledger sets a new benchmark for secure, high-performance, and privacy-preserving IoMT analytics, offering a robust blueprint for next-generation healthcare deployments.