R. Lalitha, Singam Veera Shankarreddy, V Dharani, K. Dhanalakshmi · 5 authors
In the era of digital transformation, trust and security are paramount in ensuring the integrity of various online transactions and processes, including voting systems. Traditional centralized trust management and voting systems have faced challenges related to security, transparency, and accountability. To address these issues, this research introduces a novel Blockchain-Enabled Decentralized Trust Management and Secure Voting System. It leverages blockchain technology, a decentralized and immutable ledger, to enhance trust, transparency, and security in trust management and voting processes. This system combines several key components, including smart contracts, cryptographic techniques, and distributed ledger technology, to create a robust and tamper-resistant infrastructure. In the trust management aspect, the proposed system employs a decentralized reputation system where individuals and entities can build trust through transparent and verifiable interactions. Smart contracts automate trust-building processes and provide a reliable mechanism for dispute resolution. This system allows for trust to be quantified and established in a trustless environment. The secure voting component introduces a tamper-proof and transparent voting system. Through the use of cryptographic keys and digital signatures, voters can securely cast their votes while ensuring anonymity and integrity. All voting transactions are recorded on the blockchain, providing a permanent and auditable record of the election process. Verification of election results becomes accessible to all stakeholders, enhancing transparency and trust in the electoral process. The Blockchain-Enabled Decentralized Trust Management and Secure Voting System represents a significant step towards enhancing trust and security in critical online processes, such as voting. By combining blockchain technology, cryptography, and decentralized trust mechanisms, it offers a promising solution to address the challenges associated with centralized trust management and voting systems. This research paves the way for a more transparent, secure, and accountable future in the realm of digital trust and elections.
The Federated Unified Secure Exchange (FUSE) algorithm represents a breakthrough in computer science and cybersecurity, building upon three foundational elements: Homomorphic Encryption, Zero-Knowledge Proofs, and Federated Learning. By deeply integrating these three pillars, FUSE not only guarantees data integrity and privacy in decentralized environments but also maintains high performance and remarkable scalability. This paper details the architecture, mathematical principles, and real-world applications of FUSE across Internet of Things (IoT), healthcare, finance, and supply chains. In addition, it compares FUSE with several “classic” algorithms that have dramatically changed the world—such as QuickSort, RSA, Dijkstra, PageRank, FFT, Backpropagation, SHA, Monte Carlo, Gradient Descent, and Blockchain—highlighting how FUSE addresses existing limitations in data protection and distributed computation.
Federated Learning (FL) has emerged as a promising paradigm for secure data sharing in Industrial Internet of Things (IIoT), enabling collaborative model training without direct exchange of raw data. However, recent studies have shown that FL still suffers from privacy vulnerabilities, where adversaries can reconstruct sensitive information by analyzing shared model parameters. Although several privacy-preserving FL (PPFL) schemes have been proposed to address these challenges, they primarily focus on protecting local model privacy, with limited attention to protecting global model confidentiality during aggregation. Additionally, their reliance on centralized aggregation servers introduces risks of single points of failure. To address these challenges, we propose a novel privacy-preserving blockchain-based FL framework (PBFL) that integrates blockchain, homomorphic encryption (HE), and a single masking. Specifically, PBFL employs HE to enable secure model training within the ciphertext domain, ensuring global model confidentiality. The single masking technique allows clients to apply unique random masks to their encrypted local model updates, enabling secure aggregation while preserving local privacy. Additionally, PBFL leverages blockchain for decentralized aggregation and encrypted model storage, effectively mitigating the risks associated with centralized servers. Experimental results demonstrate that PBFL achieves comparable model accuracy to state-of-the-art solutions while providing enhanced privacy protection. Furthermore, even with a client dropout rate of up to 30%, PBFL outperforms other blockchain-based PPFL methods in terms of computational and communication efficiency.
As quantum computing advances, classical public-key cryptosystems face escalating risks, driving global standardization efforts by NIST, ETSI, and ISO. This paper provides a comprehensive review of recent developments in post-quantum cryptography (PQC), focusing on lattice-based, hash-based, and multivariate approaches. We examine their security foundations, implementation challenges, and applicability across domains such as finance, blockchain, and the Internet of Things (IoT). A core contribution is the proposed Hybrid Cryptographic Framework (HCF), which combines classical and quantum-resistant primitives to enable a secure and interoperable migration path. The framework integrates hybrid key exchange, dual-signature schemes, and PQC-compatible Merkle trees, and is supported by a discrete multi-objective optimization model for selecting algorithms under real-world constraints. The paper also discusses deployment challenges, including hardware acceleration, side-channel resistance, and legacy compatibility. Real-world case studies illustrate how leading institutions are piloting PQC integration in practice. Finally, we propose a strategic roadmap for PQC adoption, incorporating adaptive triggers based on quantum capabilities and evolving threat models. By aligning cryptographic design with international standards and practical system requirements, this work offers guidance for building resilient, quantum-safe infrastructures.
Machine learning has achieved remarkable results in numerous fields, demonstrating strong momentum and promising prospects for future development. However, machine learning is facing issues related to data security. User data contains a vast amount of sensitive personal information, and once privacy is breached, users may not only suffer from harassment but also face threats to their lives and property security. As a result, users’ willingness and trust in sharing local raw data are gradually decreasing. In response to this situation, federated learning technology has emerged, which enables efficient training of decentralized data through distributed machine learning methods while protecting users’ data privacy. Traditional federated learning systems suffer from issues such as single points of failure and lack of trust. Blockchain, as a decentralized, traceable, and tamper-resistant distributed ledger technology, provides a new solution for federated learning. It records every update of the global model, verifies and tracks local updates, and is equipped with a fair incentive mechanism. Based on these ideas, this paper proposes a federated learning framework combined with blockchain, aiming to address data security issues in federated learning.
ObjectiveByzantine Fault-Tolerant (BFT) consensus algorithms in blockchain systems were confronted with two fundamental challenges that significantly impacted their practical implementation: inefficient view changes resulting from dishonest primary nodes and excessive communication overhead during consensus processes.MethodsA novel Dynamic Dual-Master Practical Byzantine Fault Tolerance (DM-PBFT) algorithm was developed to simultaneously overcome these limitations through three key technical innovations. The DM-PBFT architecture was constructed with two tightly integrated components: an optimized consensus process and an intelligent view-change mechanism. To address the critical issue of communication overhead, the algorithm incorporated a dual-primary node framework (designated as B1 and B2) combined with Verifiable Secret Sharing (VSS) to streamline the consensus workflow. In this carefully designed process: (1) client A initiated transactions by sending requests to primary node B1; (2) B1 subsequently broadcast these requests along with A's unique identifier to the network; (3) the secondary primary node B2 distributed cryptographic subkeys and corresponding commitment values to all consensus nodes in set B; (4) each node independently verified the received subkeys using the commitment values to ensure cryptographic integrity; (5) upon successful verification, nodes computed request approval digests incorporating their subkeys and transmitted these digests back to B1; (6) after collecting sufficient subkeys, B1 performed efficient key recovery and compared the results with B2's master key; (7) finally, verification values were broadcast network-wide and persistently stored in all nodes' state databases to complete the consensus process.For optimizing view-change efficiency, a sophisticated reputation evaluation system was implemented based on an enhanced PeerTrust model. This system incorporated multidimensional confidence factors that dynamically synthesized both local interaction history and global reputation metrics through comprehensive analysis of inter-node verification relationships. The algorithm strategically organized nodes into three distinct operational tiers (primary layer, consensus layer, and non-consensus layer), each with specialized reputation evaluation methodologies. View transitions were executed through continuous hierarchical scoring during consensus rounds, with the system automatically adjusting node classifications based on their evolving reputation scores.The view-change protocol incorporated three carefully designed failure recovery procedures: (1) When primary node B1 was identified as Byzantine, the top-ranked consensus node was automatically promoted to replace it while maintaining operational continuity through B2's consistent subkey management; (2) For failures of B2, a similar replacement protocol was activated with additional safeguards to ensure complete subkey redistribution; (3) In the rare case of simultaneous failure of both primary nodes, the two highest-ranked consensus nodes were promoted to form a new primary pair, with the system automatically reinitializing the consensus process. Within the consensus layer, Byzantine nodes were systematically identified through continuous monitoring and temporarily marked rather than immediately replaced. Replacement was only triggered when the concentration of marked Byzantine nodes reached precisely one-third of the consensus layer's capacity, at which point a corresponding number of top-performing nodes from the non-consensus layer were promoted. This threshold-based approach strategically minimized unnecessary view changes while maintaining rigorous fault tolerance guarantees.Through rigorous algorithmic analysis, DM-PBFT was formally proven to satisfy all critical BFT properties: (1) Request messages and subkeys were cryptographically secured through the combined use of advanced signature algorithms and commitment schemes; (2) Consistency was guaranteed through mathematical proof showing all honest nodes would agree on the same sequence of requests; (3) Liveness was ensured by demonstrating the system would always progress within bounded time, even during view changes; (4) Termination was mathematically verified through analysis of the reputation-based view-change protocol. Communication complexity analysis established that DM-PBFT achieved O(n) complexity, representing a significant improvement over PBFT's O(n²) scaling. Time complexity analysis, conducted under realistic asynchronous network models, confirmed the algorithm's superior temporal efficiency compared to existing approaches.ResultsAn extensive experimental evaluation was conducted to validate DM-PBFT's performance across multiple critical dimensions. The testing framework systematically compared DM-PBFT against three established benchmarks (PBFT, reputation grouping, and HotStuff) while examining consensus latency, communication overhead, throughput, and resilience to network dynamics under various operational conditions.In controlled latency testing with network sizes scaling to 500 nodes, DM-PBFT demonstrated remarkable performance, achieving consensus latency of just 0.46176 seconds. This represented a 400× improvement over conventional PBFT (186.44476s), a 100× improvement over reputation grouping (46.09653s), and a 56× improvement over HotStuff (25.8702s). Detailed analysis revealed that while all tested algorithms exhibited increased latency with network growth, DM-PBFT's hierarchical architecture maintained superior scalability, with latency increasing at a sub-linear rate compared to the polynomial growth observed in other approaches.Communication overhead measurements showed that DM-PBFT achieved stable performance after reaching network stratification thresholds, consistently maintaining O(n√n) complexity. In practical terms, this translated to a 62.4% reduction in bandwidth consumption compared to traditional PBFT implementations and measurable improvements over HotStuff's message efficiency. The communication overhead grew gradually with network size until reaching stratification points, after which it remained effectively constant regardless of additional node joins.Throughput benchmarking produced equally impressive results, with DM-PBFT sustaining 3,500 transactions per second (TPS) in large-scale configurations compared to PBFT's 1,200 TPS. While matching HotStuff's peak throughput, DM-PBFT demonstrated significantly better scalability, with throughput improvement rates exceeding those of the reputation grouping approach by substantial margins after crossing stratification thresholds.Additional experiments examined DM-PBFT's resilience under challenging network conditions. Node churn tests confirmed that the frequency of node joins/leaves only impacted performance through net changes in total node count, not through the rate of such changes. Network bandwidth fluctuation tests demonstrated the reputation system's effectiveness in automatically compensating for variable connection quality, with the algorithm maintaining stable performance across diverse bandwidth conditions after several consensus rounds of adaptation.The comprehensive experimental results collectively established DM-PBFT's advantages across three fundamental metrics: (1) Exceptional latency characteristics, delivering 400× improvements at scale; (2) Efficient bandwidth utilization, achieving 62.4% reductions compared to conventional approaches; (3) Consistently high throughput maintenance even under adversarial conditions with Byzantine node concentrations up to 30%. The view-change process demonstrated particular efficiency, completing in just 1.2 seconds compared to PBFT's 12.8 seconds - a 90.6% reduction that proved critical for practical deployment scenarios.ConclusionsThe DM-PBFT algorithm represented a significant theoretical and practical advancement in Byzantine fault-tolerant consensus mechanisms through its novel integration of dual-primary node management, verifiable secret sharing, and dynamic reputation-based stratification. The solution demonstrated particular effectiveness for large-scale consortium blockchain implementations, successfully achieving the dual objectives of sub-linear communication growth and rapid fault recovery without compromising security guarantees.The algorithm's architectural innovations, especially its stratified node management framework and intelligent threshold-based view-change protocol, established a new foundation for next-generation consensus protocol design in increasingly complex and adversarial network environments. Future research directions were identified to further enhance the algorithm's practical utility, including: (1) Optimization for real-world deployment scenarios with heterogeneous hardware; (2) Development of cross-shard coordination mechanisms for sharded blockchain architectures; (3) Enhanced security analysis under sophisticated adaptive adversary models; (4) Integration with emerging cryptographic techniques such as zero-knowledge proofs for additional privacy preservation. These advancements promised to extend DM-PBFT's applicability to an even broader range of production blockchain environments while maintaining its fundamental advantages in efficiency, security, and scalability.
Introduction. The significance and relevance of researching the development of the cryptocurrency market are driven by the digitalization of society, its penetration into all areas of life, including the financial sector, the transformation of its regulation, the need to prevent the cryptocurrency market from operating outside the legal framework, and mitigating the negative consequences of cryptocurrency use. The purpose of the article is to clarify the essence of cryptocurrency, identify its determining factors, and explore the key issues in its functioning. Results. Various approaches to interpreting the concept of "cryptocurrency" have been analyzed. The essence of cryptocurrency, its distinctive features, attractiveness, and associated negative processes have been identified. The author's perspective on the nature of cryptocurrency is presented. The patterns and peculiarities of cryptocurrency market regulation in different countries have been characterized, identified the negative consequences of the absence. The importance of financial monitoring in this sector has been emphasized. Conclusions. Despite the advantages and disadvantages of cryptocurrency implementation, it is essential to establish a well-structured regulatory system, clearly define the regulatory authority, determine the issuance process, usage, application areas, and taxation framework; establish the relationship between cryptocurrency and other financial assets; and ensure state oversight of this sector, including the appropriate development and enhancement of financial monitoring mechanisms.
Blockchain Technology Applications and Security
Cryptography and Data Security
Advanced Steganography and Watermarking Techniques
A novel identity management concept known as decentralized identity (or self-sovereign identity) has drawn significant interest and extensive development within both academic and industrial circles. Decentralized oracles empower users to demonstrate the origin of data accessed through TLS from a specific website and verify statements concerning that data. This system operates without the reliance on trusted hardware or requiring modifications on the server side. In DECO (CCS 2020), the first decentralized oracle within TLS 1.2 was introduced. It also investigates the methodology of optionally proving statements about this data in a zero-knowledge setting, preserving the confidentiality of the data itself. Furthermore, a separate study proposed an attribute-based anonymous credential system incorporating a commitment scheme (ASIACRYPT 2020), introducing show proofs to verify a set of attributes in a credential to verifiers without disclosing attribute specifics. Building upon these research findings, this thesis presents a series of optimizations and expansions aimed at advancing self-sovereign identity solutions. First, we present DIDO and DIDO+ (decentralized identification oracle), which expands DECO to TLS 1.3. In DIDO, we address several unresolved challenges, encompassing the incorporation of X25519 key exchange, the creation of a round-optimal three-party key exchange, the structuring of a 2PC system for TLS 1.3 key scheduling, and the optimization of circuit design for 2PC protocols. Furthermore, we expand upon DIDO to create DIDO+, which introduces a protocol named selective disclosure. This protocol facilitates the extraction of particular plaintext substrings from websites, enhancing the functionality and flexibility of the system. Our implementation is verified against real-world websites, and a security proof is furnished to validate its integrity. Next, we introduce DEVS (decentralized verification service), the first generic decentralized verification service built on a decentralized oracle. DEVS enhances the traditional single-verifier oracle by incorporating a multi-verifier mechanism. Our aim is to streamline the protocol for reduced communication and computation costs while enabling reusable proofs and maintaining security. DEVS comprises three key components: a reconstructed decentralized oracle to bolster data trust, secure storage for share retention, and a verification process tailored to the needs of verifiers and authorities. Considering 10 verifiers, the reconstruction reduces the running time (approximately 89.1% or 89.6%) in WAN settings and communication bandwidth (approximately 89.1% or 87.8%) of TLS 1.2 or 1.3, it provides a significant improvement when working with multiple verifiers. Lastly, we introduce a unified attribute-based anonymous credential system where users consistently receive credentials in a standardized format from the issuer. This system allows users to opt for efficient multi-use or single-use show proofs, offering a more user-centric perspective compared to current schemes. Technically, we suggest an interactive method for the credential issuance protocol utilizing two-party computation with additive homomorphic encryption. This approach maintains crucial security features such as impersonation resilience, anonymity, and unlinkability. In addition to the interactive protocol, we develop show proofs tailored for efficient single-use credentials that uphold user anonymity throughout the process.
This systematic literature review compares two leading selective-disclosure primitives for decentralised identity-BBS+ signatures and Selective-Disclosure JSON Web Tokens (SD-JWT)-to clarify their suitability for privacypreserving credentials. Following Kitchenham's protocol, 226 records from 2017-2025 were screened across IEEE, ACM, SpringerLink, ScienceDirect, IETF and W3C repositories, yielding 31 primary studies with empirical data. Quantitative synthesis shows that BBS+ derived proofs remain constant-size at roughly 140 bytes and verify in about 12 ms on consumer hardware, whereas SD-JWT presentations grow with the number of revealed claims but still verify in under 10 ms for typical twoclaim use cases. Qualitative analysis confirms BBS+ provides strong unlinkability, predicate proofs and zero-knowledge disclosure, while SD-JWT offers seamless integration with existing JOSE/OAuth infrastructures yet carries correlation risk due to stable salted digests. Standardisation progress is comparable: the BBS+ cryptosuite reached W3C Candidate Recommendation in April 2025, and SD-JWT is in late-stage IETF review. The review concludes that privacy-critical scenarios such as age-gated services favour BBS+, whereas high-throughput web applications benefit from SD-JWT; consequently, hybrid wallet support for both formats is recommended. Future research should tackle scalable revocation, post-quantum migration and multi-credential aggregation to sustain long-term trust and interoperability.
Atoms and photons, two things so different but yet so alike. The former, the building block of matter, something we learn about in school and imagine it as some tiny marbles encircled by other tinier marbles. The latter, an electromagnetic wave, a light particle or an excitation of the electromagnetic field. Quantum mechanics tells us about the properties of these two entities. And even if it sounds, looks and writes counter-intuitive, it has proven right for over a century now. In this work, I elaborate on how we tested the laws of quantum mechanics and how we used them learn more about the tiny building blocks of nature and the fields they use to talk to each other. The atoms we use, are artificial. Superconducting qubits, small electrical circuits with quantized energy levels behave like electrons that transition between different orbitals in an atom. One of the qubits' advantages, is also a big disadvantage. We design the circuits' energy levels and fabricate them in a cleanroom. This allows for arbitrary spaced energy levels but in contrast to real atoms, prevents two superconducting qubits from being alike. Still, this qubit platform is one of the frontrunners for future quantum computing technology and testing fundamental physics due to their scalability. We interface superconducting qubits, which operate in the GHz regime, with microwave photons. We use 3D aluminum cavities as mediators between qubits and photons. The cavities allow for non-destructive readout of the qubit state, they shield the qubits from noise at the qubit frequency and they give us an easy way to frequency-tune these joint systems. We need to operate superconducting qubits and their cavities at millikelvin temperatures in dilution refrigerators. At higher temperatures, superconductivity suffers and even worse, the environment is filled with thermal noise photons. This poses a fundamental limitation on the scalability of superconducting qubit devices. Also connecting multiple devices in different fridges does not work over room temperature links because the microwave photons used for this purpose will be covered in noise and the quantum information they carry, will be unusable. Infrared photons do not suffer from this noise problem since there are close to zero thermal noise photons at their frequencies at room temperature. We cannot simply interface superconducting devices with optical photons due their frequency mismatch and the destructive effect of optical photons on superconductors. Therefore, we use microwave-to-optics transducers that allow to convert microwave photons into optical ones and vice-versa. The transducers that we use are macroscopic electro-optic transducers using the Pockels effect in a disk-shaped Lithium Niobate whispering gallery mode resonator. By using a strong optical pump, photons from the two frequency domains experience a beam-splitter interaction and get converted from one to the other. We measure the generated optical photons using elaborate optical setups, optical heterodyning and single photon detectors to gain knowledge about the qubit state or the converted microwave photons. Bridging the microwave and the optical world allows us to take advantage of both of their strengths but it also requires deep knowledge about both of their working principles. In this work, we describe two experiments that our group conducted to showcase the opportunities that arise from interfacing superconducting qubits with optical photons but also the pitfalls, one may encounter on the way. In the first experiment, we managed to all-optically read out a superconducting qubit. We show that the assignment fidelity, the probability that a measurement of the qubit state matches the prepared state, is close to equal for all-optical, microwave-to-optics and conventional microwave readout. We show T1 and T2 measurements for all three readout types and give an analysis of the noise caused by the optics. Finally, we show that the infrared light does not affect the qubit performance in a negative way but that the heating it causes does. This is an important insight that we used in the next experiment. The second experiment is the upconversion of itinerant single microwave photons to the optical domain. We show that we can generate single microwave photons from a qubit-cavity system. We upconvert these single photons, measure them with a single photon detector and reconstruct their shape. By conducting a single photon Rabi measurement, we show correlations between the microwave and the optical domain. And by thorough signal-to-noise measurements and noise analysis, we find that we can generate single infrared photons with high signal-to-noise ratio 5.1 and low transducer added noise (<0.012 quanta). We show that this measurement creates a path towards entanglement of a superconducting qubit and an optical photon and what parameters need to be improved to achieve it. Additionally, this experiment is a proof of principle for an on-demand infrared single photon source. More generally, it allows to link microwave quantum technology in general to the optical domain.
Verifiable Delay Functions (VDFs) introduced by Boneh et al. (CRYPTO'18) are functions that require a prescribed number of sequential steps T to evaluate, yet their output can be verified in time much faster than T. Since their introduction, VDFs have gained a lot of attention due to their applications in blockchain protocols, randomness beacons, timestamping and deniability. This thesis explores the theory and applications of VDFs, focusing on enhancing their soundness, efficiency and practicality. The only practical VDFs known to date are based on repeated squaring in hidden order groups. Consider the function VDF(x,T)=x^(2^T). The iterated squaring assumption states that, for a random group element x, the result of VDF cannot be computed significantly faster than performing T sequential squarings if the group order is unknown. To make the result verifiable a prover can compute a proof of exponentiation (PoE) \pi. Given \pi, the output of VDF can be verified in time much less than T. We first present new constructions of statistically sound proofs of exponentiation, which are an important building block in the construction of SNARKs (Succinct Non-Interactive Argument of Knowledge). Statistical soundness means that the proofs remain secure against computationally unbounded adversaries, in particular, it remains secure even when the group order is known. We thereby address limitations in previous PoE protocols which either required (non-standard) hardness assumptions or a lot of parallel repetitions. Our construction significantly reduces the proof size of statistically sound PoEs that allow for a structured exponent, which leads to better efficiency of SNARKs and other applications. Secondly, we introduce improved batching techniques for PoEs, which allow multiple proofs to be aggregated and verified with minimal overhead. These protocols optimize communication and computation complexity in large-scale blockchain environments and enable scalable remote benchmarking of parallel computation resources. We then construct VDFs with enhanced properties such as zero-knowledge and watermarkability. It was shown by Arun, Bonneau and Clark (ASIACRYPT'22) that these features enable new cryptographic primitives called short-lived proofs and signatures. The validity of such proofs and signatures expires after a predefined amount of time T, i.e., they are deniable after time T. Our constructions improve upon the constructions by Arun, Bonneau and Clark in several dimensions (faster forging times, arguably weaker assumptions). Finally, we apply PoEs in the realm of primality testing, providing cryptographically sound proofs of non-primality for large Proth numbers. This work gives a surprising application of VDFs in the area of computational number theory. Together, our contributions advance both the theoretical foundations and the real-world usability of VDFs in general and in particular of PoEs, making them more adaptable and secure for current and emerging cryptographic applications.
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.
随着信息技术的快速发展，数据安全和用户的隐私越发受到重视。本文提出了一种匿名认证密钥交换(Anonymous Key Exchange, AKE)协议，旨在为医疗场景下的医疗数据共享和患者身份隐私提供安全和隐私的保护。该方案通过使用累加器、零知识证明和关联数据加密等技术，实现用户匿名的认证和安全的会话密钥协商，有效防止敌手对于用户和医用物联网设备的攻击，还能抵御诚实且好奇的医疗机构对患者身份的猜测。相较于现有的方案提供了更强的隐私安全保护，并且很好地平衡了性能和安全性，具有重要的理论价值和意义。With the rapid development of information technology, data security and user privacy have been paid more and more attention. This paper proposes an Anonymous authenticated Key Exchange (AKE) protocol to provide security and privacy protection for medical data sharing and patient identity privacy in medical scenarios. By using accumulator, zero-knowledge proof and associated data encryption technology, the scheme realizes anonymous user authentication and secure session key agreement, which effectively prevents adversaries from attacking users and medical IoT devices, and can resist honest and curious medical institutions from guessing the patient’s identity. Compared with the existing schemes, it provides stronger privacy security protection, and a good balance between performance and security, which has important theoretical value and significance.
Multi-signatures have recently garnered considerable attention, particularly within the domain of smart contracts in blockchain ecosystems, as they enhance account security and mitigate single points of failure by requiring the approval of multiple key holders for transaction execution. However, most existing multi-signature schemes heavily rely on traditional Public Key Infrastructure (PKI), which requires a trusted authority and conflicts with the decentralized nature of blockchain technology. Certificateless multi-signature (CLMS) schemes, which eliminate the requirement for a trusted authority, represent promising solutions to address this issue. Nevertheless, existing CLMS schemes encounter challenges that limit their suitability for smart contract applications, including high communication overhead, expensive verification costs, and “loose" security reductions. To address these challenges, we propose two novel two-round certificateless multi-signature schemes. These schemes not only support key aggregation but also optimize the signing process with two-round communication, maintaining fixed computational overhead during verification. Furthermore, the security proofs for the proposed schemes are independent of the Forking lemma, resulting in tighter security reductions and strengthened security assurance. Finally, experimental results demonstrate that the proposed schemes significantly reduce both communication and computational overhead compared to existing CLMS schemes, making them more efficient and practical for blockchain-based smart contract applications.
The correctness of a computation can be efficiently verified in a privacy-preserving manner without re-execution using zero-knowledge succinct non-interactive arguments of knowledge (zkSNARKs). With short transcript sizes and fast verification times, zkSNARKs enable the potential deployment of computationally intensive algorithms—such as machine learning models—on the blockchain, making them efficiently verifiable through short proofs. However, the prover time for matrix computations in these settings often fails to scale efficiently with increasing model complexity and data size. We are the first to systematically address zkSNARKs for general matrix computations with practical prover efficiency. We achieve an $O(N + nM)$ prover time, asymptotically faster than the unverified matrix computation, for computations involving $M$ matrix operations on $n \times n$ matrices with $N$ total non-zero entries. Starting with a single dense matrix multiplication, we propose zkMatrix, a special-purpose zkSNARK for verifying committed $n \times n$ matrix multiplication through their projections onto random vectors. Among zkSNARKs with $O(\log n)$ transcript size and verifier time, zkMatrix is the first to achieve $O(n^2)$ prover time and $O(n^2)$ RAM usage. Batching multiple proofs together reduces the prover time for each additional multiplication to $O(n)$ group operations. Next, we design zkSNARKs for sparse matrix multiplication with $N$ non-zero entries. zkSmart reduces the prover time from $O(n^2)$ to $O(N + n)$, relying on an $O(N + n)$-prover vector-matrix-vector product argument, achieved by improving Bulletproofs. Moreover, \zksmart formulates verifiable computation represented as a matrix circuit of $M$ nodes, each denoting a matrix operation. Sparse matrix multiplication translates the matrix circuit satisfiability (Mat-Circ-SAT) problem into the high-dimensional rank-1 constraint system (HD-R1CS), a matrix-circuit version of the rank-1 constraint system (R1CS), traditionally used for arithmetic circuits. Using zkSmart, we achieve $O(N + nM)$ prover time for general matrix computations. To reduce the cost of committing to intermediate variable matrices in zkSmart, we introduce Evalyn, which generates proofs using a pre-order tree traversal on the abstract syntax tree (AST) of a matrix expression. Evalyn ensures output and input consistency in serial matrix computations by linking randomness for zkSNARKs between parent and child nodes, eliminating the need to commit to the nodes and significantly improving prover efficiency. Our prover for R1CS outperforms state-of-the-art general-purpose zkSNARKs. As a foundational component of our framework, we optimize Bulletproofs to construct the fastest known inner product argument (IPA). Additionally, we propose a zero-knowledge transformation that commits to transcript elements with only logarithmic overhead—while maintaining compatibility with post-quantum secure, non-homomorphic commitment schemes. We apply our framework to zero-knowledge machine learning (zkML), providing zkSNARKs for neural networks. We translate floating-point truncations and non-linear activation functions into linear algebra equations that can be verified by our framework. We utilize our framework to generate efficient proofs for the attention layer in large language models (LLMs). After resolving all these challenges, we have thoroughly addressed the design of efficient zkSNARKs for matrix computations.
Classical symmetric cryptography interprets ciphers and hash functions as bit-based functions, that is, as functions between binary vector spaces. However, in the past decade, research and adoption of cryptographic protocols with advanced privacy-enhancing features, such as Fully Homomorphic Encryption, Multi-Party Computation and Zero-Knowledge proof systems, have accelerated. These privacy-enhancing protocols often have a novel feature: Their mathematical model is native over a prime field, whose characteristic p is significantly larger than 2. Applications of the aforementioned protocols often require a symmetric cipher for data encryption or a hash function for data compression. For efficiency reasons, two novel criteria must be satisfied: - A primitive should be native over a prime field, so it is modeled as function between vector spaces of characteristic p. - It should be possible to evaluate or represent a primitive with a low number of multiplications. Primitives satisfying these criteria are often summarized under the term Arithmetization-Oriented. To satisfy the second criterion, Arithmetization-Oriented primitives are often constructed by iterating relatively low-degree polynomials. However, this choice comes at a cost: Such primitive can also be modeled as a low-degree polynomial system for key or preimage recovery. Analyzing the complexity of polynomial system solving-based attacks poses a major problem in the design of Arithmetization-Oriented primitives. In practice, cryptodesigners often applied generic complexity bounds for cryptanalysis without proving the underlying assumptions. The aim of this manuscript is the development of tools for complexity estimations based on provable properties of a cryptographic polynomial system. We also apply our tools to the analysis of Arithmetization-Oriented Substitution-Permutation Network and Feistel ciphers. For early Arithmetization-Oriented primitives, our analysis reveals a large complexity gap between initial cryptanalysis and the capabilities of state-of-the-art polynomial system solving techniques.