Artificial Intelligence (AI) models are increasingly integrated into high-stakes domains such as finance, healthcare, autonomous systems, and legal decision-making. As their influence expands, concerns about accountability, fairness, transparency, and regulatory compliance have become central to both researchers and practitioners. One of the key challenges is auditing AI models in a manner that is tamper-proof, verifiable, and compliant with evolving regulatory frameworks. Traditional auditing mechanisms rely heavily on centralized logs and organizational trust, which creates vulnerabilities in terms of manipulation, incomplete records, and opacity in data flows. Blockchain technologyâowing to its immutable, decentralized, and transparent natureâoffers a powerful paradigm for establishing data provenance in AI auditing. By ensuring traceability of datasets, model updates, training logs, and inference outcomes, blockchain can provide regulators, stakeholders, and organizations with reliable audit trails. This paper presents a comprehensive exploration of blockchain-powered data provenance for AI model audits. It analyzes the limitations of current audit systems, evaluates how distributed ledger systems can strengthen accountability, and proposes an integrated framework that combines blockchain with cryptographic verification, zero-knowledge proofs, and federated logging to ensure verifiability without exposing sensitive data. The study synthesizes contributions from literature, presents a methodology for deploying blockchain-based provenance systems in AI pipelines, and evaluates potential results in terms of efficiency, compliance traceability, and security. Simulation experiments suggest that blockchain-enabled audits improve transparency, reduce fraudulent activities in AI operations, and enhance compliance readiness by more than 50% compared to traditional audit approaches.
NuLink provides privacy-preserving technology for decentralized applications via APIs. Users can securely store its valuable data, trade with others and so on. To ensure the privacy and security of service provided by NuLink, (zero-knowledge) proof systems are necessary. Zero-knowledge proof systems allow the prover to make the verifier believe that a certain conclusion is correct without providing any useful information to the verifier. In NuLink, we are going to use (zero-knowledge) proof system in the following three methods: 1. Users store their data through NuLink in a decentralized manner. To ensure that the storage clients are indeed storing the data, we employ proof of storage systems. In this system, users prepare certain challenges that can only be correctly answered by those who are actually storing the data. 2. Users have the option to outsource computations to NuLink. To verify the correctness of the computation results provided by the compute node, we require the node to provide a proof of correctness via SNARK systems. When sensitive parameters are used as inputs for computation, we utilize zk-SNARKs to prevent any potential leakage of these parameters. 3. Users may choose to trade their data through NuLink. To confirm that the buyer has sufficient digital funds and the seller possesses the desired data, both parties can provide a proof via zk-SNARKs. This builds confidence and prevents cheating during transactions. Using zero-knowledge proof systems, we can ensure that all nodes in NuLink behaves honestly and avoid cheating in the whole system.
Zero-knowledge proofs (zk-Proofs) are communication protocols by which a prover can demonstrate to a verifier that it possesses a solution to a given public problem without revealing the content of the solution. Arbitrary computations can be transformed into an interactive zk-Proof so anyone is convinced that it was executed correctly without knowing what was executed on, having huge implications for digital currency. Despite this, interactive proofs are not suited for blockchain applications but novel protocols such as zk-SNARKs have made zero-knowledge ledgers like Zcash possible. This project builds upon Wolfram's ZeroKnowledgeProofs paclet and implements a zk-SNARK compiler based on Pinocchio protocol.
Lu Zhou, Abebe Diro, Akanksha Saini, Shahriar Kaisar · 5 authors
Identity sharing systems, regardless of their architectural models, share common vulnerabilities. These systems compel users to divulge personal information and furnish proof of identity for accessing services, leaving them susceptible to data breaches that can culminate in identity theft and jeopardize online data security. While blockchain technology offers a potential remedy, delivering enhanced security, immutability, and traceability, it simultaneously raises pertinent concerns surrounding privacy and transparency. The integration of zero-knowledge proof (ZKP) technology has emerged as a promising solution, particularly in enhancing privacy within the transparent blockchain ecosystem. Our paper conducts an exhaustive survey of the existing literature, with a particular focus on the assimilation of ZKP technology into blockchain for the secure sharing of user identities. We undertake a critical evaluation of the advancements achieved in this domain, pinpoint the formidable challenges that must be confronted, and uncover nascent opportunities for further exploration. Our contribution transcends the realms of mere summarization and analysis; we go a step further by offering recommendations drawn from real-world case studies and delineating future research directions.
Dear colleagues, we have a Cressington 108 carbon coater that uses carbon rods. After shaping the rods, I use sandpaper to make a flat surface and I wipe the powder away from the surface with a cloth. We use the maximal distance, approximately 10 cm, 4 volts and 10 seconds (we repeat 6-8 times with 1 minute pause in-between). Sometimes, there are sparks coming from the shaped rod and I don't know what I can do to avoid them. Quite frankly, this doesn't seem to affect the quality of the coating but still I wonder (1) why they appear (2) how they can affect my sample and (3) how to avoid them. Any thoughts? Stephane Nizet [email protected] If you have a shutter and the sparks are coming in the beginning or at a specific time: use it. Maybe a longer burn time and a shutter will do the trick. Stefan Diller [email protected] I suspect that it has something to do with the structure of the rod and the bonds between particles. I have come to expect them and to live with them. As you say, they don't seem to affect the bulk of the coating. I do try to limit the number of sparks. I control the current/voltage to allow a few per second. If there are none, it seems that the coating rate is very low. If the rate gets too high, the rod often fails too quickly. Warren Straszheim [email protected] You can try rinsing the rod with EtOH and then air gas blow-drying before use. Also, you can try a light burn (rod fire-red) for a few seconds before increasing the current. I don't believe the sparks will ruin your coating. Mike Delannoy [email protected] The sparks are basically glowing chunks of carbon coming from the arc. My experience is that the higher the current the more chunks are present. Generally, the chunks don't significantly affect the final product. So, like Mike Delannoy, I would set the current so occasional âsparksâ are present as the optimum for deposition. You don't want a spray of sparks like a fireworks sparkler. Henk Colijn [email protected] I usually find it is because there is loose carbon on the tip. We use a folded paper towel. When the end of the flat carbon is scraped across an emory paper to remove embers from the last coating, it is always wiped on the paper towel. If we use a rod we have sharpened to a point, again we wipe off any loose carbon on the paper towel. Elaine Humphrey [email protected] You may need to shape the carbon rod with a carbon rod shaper for sharpening the 6.5 mm carbon rods into 3.5/3 mm cylindrical tips and 1.75/1.50 mm (3 diameters). It is important that the carbon rod has a reduced diameter at the tip. It is best to increase the voltage to 0.5 V. Abdelyamine Naitbouda [email protected] Dear EM Experts, some of my crystallographer colleagues and I wonder about the fundamentally different appearance of density histograms in EM Coulomb potential maps versus X-ray electron density maps. Here is the question: What I've noticed is that they are on different scales. That is understandable, as e/A3 is different than V. But there are papers showing that these values are somewhat proportional to each other for lower resolutions. But the other thing that I have noticed is that electron density maps have close to normally distributed value distributions, whereas cryoEM maps have a sharp spike and a very long tail. As a result, an electron density blob in an X-ray map looks nice somewhere around 3 sigma, whereas for cryoEM it's sometimes 10 sigma, 17 sigma, 20 sigma, all over the place. I'm thinking of using thresholds based on percentiles rather than sigmas, but my main question is: shouldn't the values on cryoEM maps also be approximately normally distributed? What is the cause of this non-normality? Sharpening? The raw experimental data themselves? Here is a potential partial answer: In cryo-EM, there is no absolute scaling, which means that density values can vary significantly. This variability can explain why different sigma values are consistently encountered. I can confirm that the density value distribution behaves as described, and I have also observed this. However, I cannot provide a definitive answer as to why this occurs. My best guess is that it may be related to B factor weighting in motion correction, but I cannot provide a conclusive explanation. What are we missing here? Ben Rupp [email protected] According to PDB/EMDB validation reports, this spike in the voxel value histogram is caused by masking. One validation report I have says: âA spike in this graph at zero usually indicates that the volume has been maskedâ. You can probably also find this note in validation reports of released PDB/EMDB entries. Opening a map from 3D refinement and one of the two half-maps from the same job seems to confirm this. The mapâs histogram shows this spike at zero, but the half-mapâs histogram doesnât. Half-maps are never filtered nor masked, whereas the main map is masked (in cryoSPARC this is done by default, unless one turns off automatic masking and doesnât provide any mask). The fact that there is no absolute scale for contour level in cryoEM maps is indeed annoying (I would like to compare maps without worrying that maybe I chose inadequate contour levels). My understanding is that it is caused at least in part by the fact that the size of the box enclosing the particle is arbitrary. Different amounts of low-value voxels between different maps give them different voxel value histograms, therefore choosing a contour level in terms of a certain number of standard deviations above the mean produces different results with different maps. Electron density maps from crystallography donât have this variability because the box always spans a full unit cell, without this variable padding around the region of high density. At least this is how I understand Tom Goddardâs explanation in this discussion from last month on the Chimerax-users list. Maybe there are other reasons adding to this. I hope this helps. Guillaume Gaullier [email protected] Unfortunately, a fundamental mistake has been made in the much-cited paper by Erickson and Klug (1971) [https://doi.org/10.1098/rstb.1971.0040] more than 50 years ago. They assumed that the EM amplitude contrast of the (stained) biological object is proportional to the phase contrast of the same object over all spatial frequencies. If that were indeed the case, one single transfer function would suffice to describe how the linear imaging device would generate an output image. In reality, the amplitude contrast and phase contrast are two separate properties of the complex transmission function of the object, and these are associated with different physical properties [HA Ferwerda and MG van Heel (1978) https://doi.org/10.1007/978-1-4757-0665-9_47]. The problem is that the proportionality error has crept into almost all popular CTF determination programs where, say, 10% or 15% amplitude contrast is suggested ab initio. Any percentage of amplitude contrast erroneously causes the average density of the cryo-EM 3D reconstruction to deviate from zero. Any phase-contrast image must yield a zero average as it should be for any phase contrast image where what is measured is the difference in phase between any point in the back focal plane of the system with respect to the phase at the origin! That means that the phase at the origin must be zero. (Zero being the average density over the image around which the phase information is a to the transfer function will the of the CTF and the values in programs longer the That results no longer to each other and will also any with density maps in X-ray more Heel [email protected] I would be so with that paper from is always a certain level of experimental and as as I understand it at that time and level it as amplitude and phase be As you point this is the case, and I may a we years into the CTF for an filtered Here a CTF to and the in the experimental and CTF an contrast contrast by the is I for the quality of the but it the best we to with that very at that I may a of to all too that uses this at present. The explanation is for the of very data a CTF to than a few But the problem with the density and the phase at Erickson and Klug (1971) and MG van Heel [email protected] I with about the of using a distribution of the amplitude But a single paper as you say, the problem is rather the of an by the have been of like to amplitude my paper on using and a paper with and where we to use on the cryo-EM in The thing the in the amplitude contrast so we something must have to the all and as we know part and part My 1 [email protected] is adding I guess I will some One to the amplitude contrast from the is from the amplitude contrast by using an as The to an more and I will to I with that a for amplitude contrast is to the We to this in of using and more of this contrast would almost that an as to information is so for that a to the for cryoEM I believe one the phase for the CTF is it is the as it may to for amplitude contrast in It would be we some to the terms in how we the of the experimental image and the of the This at all on the However, as is often the case, the biological paper paper in the paper Electron of an Electron and of In this shows how terms in the give to amplitude I what means by the amplitude contrast and phase contrast are two separate properties of the complex transmission function of the object, and these are associated with different physical is no the amplitude contrast the is by the same as the phase As with it is that is it is the discussion on the of amplitude contrast is I it gets to the of one be to use standard deviations of voxel values to compare different I it the to compare two cryoEM maps at a standard we have CTF and other in the image reconstruction so we have a What do you expect you the voxel The value in each voxel will the average value of the and you will have an approximately distribution of voxel What about the with we have on average one the distribution then be it be a and as the distribution of the potential by the imaging would be So, variable rate is one that may of maps of the same for there is no I can why the distribution should be the same between different or for [email protected] point about the density is the are we a physical unit then we should be to difference maps [email protected] I you are a different As you a difference I to it is between two maps of with a different or In that case, this should be for variable image and data quality you try a distributed with This will the average in each from one map to the average in each of to a difference be to to the lower quality map as it the on [email protected] Dear I never any distribution of the amplitude in my My paper on the and the part of the complex transmission function of the object in and to as as it to electron Any is in electron My paper is based on a with the size of the in the back focal and on an function which is an more van Heel [email protected] I a years and very of as the for the and at the The we are with is about to on a to the and a system for the to any and all from other for what system do you what do you like about your current what do you like least about your current what do you your system do but it what other are you of any or about also to from have can therefore and you have to to a different system and how that to for and tips to a you donât to the list. to from a of [email protected] We use across (I 3 What do you like about your current it my data to current to do It for and are in the The or one or more to the to also in a across [email protected] gets the job done very because we have other the bulk of and is a that for all the If we as a to do this it would be a from I donât have to generate reports because we can to do this. One of my to generate reports, is that the data are in an to last and are into a single and as a and them is because the is for too and The data for a are Also, we can So, a has a with a of the are for must to a of or make it the to a because they donât use the then the function doesnât they must use the the in the and this at 3 is a to to the system to using for time by how do you But I donât know there is that is this may be the of and sometimes this of control is to the and about the fails is that can for 1 but for a and the is to have the have but what this for a system that has but because we donât it and the do it are very at If we or it [email protected] As a of one of the I would that what is important is that the you is to a of other as and to as data and with to a However, it is to from a system to as each one has as [email protected] Here at we use the system for It is a by and made and is by It can be into the and It also reports over and more reports in for and you probably want to for it into your there is no and we don't use it for In the we use there is no or data I never to with the or do with the but we are with as it [email protected] We have been using for a few The is and but at the and then you can as and they have It has been across 50 all at very and and so uses it. It is in the system and we find that some and are We use to and there are for The is that also on in a like for image has been a for [email protected] I have and them to be to the However, the by the 3 sample and do you the for I them with air for that they were very but I I have the [email protected] any that are you must a of seconds to make that the be over so you need to it. If you are about the of a you can it by one at in a and the are papers this. [email protected] and other light and turns into It must be and [email protected] It is for to be As you and have a can However, this may the to is them to uses a we on the in a or other with the output close to the for You can the time for the and by adding In my is long adding a light the in with may know you have any [email protected] We a to by the of a of as also find a for the of and for these more [email protected] I in how I in the and more how I I understand that I would need to back to (I in and but what of I need to in the and of I have any with so an can I your time and you in for your [email protected] is always If you want to a in a biological what like and are of is a can be the but what a are should be in a is also very that which need to be in a or is to but again I would a for a or two to this is something you would like to are that have on be to from the but these would be very to One is to a where is a part of the this and then to a where you are [email protected] If you are to as a and a few then this is than in a but you need to where you I the a very job a of I to make of a and it for to to with and the with to and for but in a very One of my something in by in a in are you can find and can to this [email protected] from the there is always an to in has a for this. this in or in [email protected] is to the a Electron and a you need to at least years of The in has the where I It is there that you will all you need to know to the you will biological or or and do the The of also has for the a to an with the of the time will be for and other If you have some and the then you don't need to do the you don't want you can the two or more years to electron you will need to with the in as may have I I the in I to use the and higher I from and that with biological and My job at the to an then I have in in and biological I also do EM for or You can always try to and by That too for some on the But to the the is the to as the is [email protected] or electron is into a which also electron two will you to and some will you how to and for light as have need to have experience to be as a will you with about and how to the When often for [email protected] If in there are some for you to may you some or at least and the of the of They have and at least to and of are in thinking of back to for a or other I and as What of are you light or biological or or you you [email protected] We are to and then a image in the have a for that have some I'm that a system would be more but we do have that so would to compare We are for and for to for image for and some image from a This is for and a or [email protected] is They will your is for the you want to use and make to for in I with them. They with a [email protected] I also an from over a and a They a and were in I an but they seem to by I don't know they would be to for but a and may the value from a [email protected] I my from with very [email protected] I have been as to current We because we are and other in a so we are of with we were in the we have been a for by and have been very with We have a and data which is years and still It has of as as and a In the years we have one of and at the the and to higher all at the same time (we were for 10 the in to also at (I have that of the and other from them and to that you need and (I or which be for with always as to what to my on a is for in of the and you on for of the But then is in of the so you do the by to a system or number of We have and have that the difference with and is very on between and for [email protected] If you have somewhere to have been a with are I believe that a few years been more than any that I have of is for what you but a single be this and you the of a and [email protected] We are [email protected] you to all have or The information has been very [email protected] light has a of very mm mm 50 that are I can find at around 50 mm on each As I to them I would to find a no more than around [email protected] I would I donât know the but the we are with are nice and we have with in the last of have been [email protected] I have to with for and observed colleagues using them with light with and my of are to The are or and can the is often is the This is very and is where of the are coming This is why you at the of a flat or it seems but they are into a they are it's the and [email protected] We and I of the from You them When some in quality is we have made by with that we [email protected] have a for from to on two of on and my best to remove it have so Any are [email protected] have to from We have to use it in the and as long as you donât use too you should be I very come with a different that I this is a being [email protected] I use a with It is a and I wipe it off the You must be with but it with of Mike [email protected] is it be off the using a or the end of a We use the for [email protected] This is why we use or for at I of the with I to off the but on the is The from the of the don't (we have a of experience with so why would off other We these is but we don't a difference with the biological it's a have the (I do have these on my of to do I have [email protected] I the you use on as the coating are but I to use a for the would and the back to and they it the in some and then the and time [email protected] all of the that I would normally The from is very for different about how to off of and other My other is to use that are for We these from in This to years by from and they have a few of from You the and then use the end to the I do all of this a [email protected] We the same with time the using the same to the on the It to it but However, it a my colleagues have for image quality and on a surface the they and of on a and wiped it off the be wiped off for longer than were the [email protected] I don't have experience with but have of to with and the other I to all the air on my because they all been in be very with using too but some is probably to remove the I try to to or between the them on the a and then wipe off in with an or a of paper that folded into a some but too your each as you don't want any to and the each a or using the as a of this can a but I have to all so [email protected] I that you have with coming from or can be on the and come from It is also to remove them with other than I always with before using to and to in [email protected] to use to remove the from This [email protected] note of on The is part of the that by the When you use an that is by the of the you want to before using it in we use I to and According to the and are the same and But to be a in in the at would any more with the it probably would have I don't use I you in using a different you will be But it's nice to an before it a In I would for in with the to the [email protected] I with on using other may be very different in different the is usually for a specific is the I would the of the with that for as different have different in and across the and try a of as a difference in in may [email protected] I the for at an It I I try it but it and I it by at the in [email protected] is from what I've and Unfortunately, they never away so I it You I'm the make you don't the the that would the of the If you in the have to the [email protected] I've and the is to be it into any before it If it into a or you be to it to on flat be to give it of time to as it to be set to be off the If you do allow it to it will into you try to remove it. are also for it a paper is for this to a of allow it to a then the paper on then a of This around the so is in you should be to on the paper and the of will come with it. [email protected] you all for the of on the end of a the trick. on the I'm also the we as it seems that this is a problem with That will be for [email protected]
Open access
Diamond and Carbon-based Materials Research
activated carbon and charcoal
Aerosol Filtration and Electrostatic Precipitation
Due to exponential demand in IoT based healthcare, the demand for robust mechanisms to ensure data privacy, security, and scalability with the increasing dependence on cloud-based healthcare systems is immensely felt. Current approaches to dealing with health-care data in cloud settings lack the potency to tackle challenges emanating from the distribution of non-IID data, dynamic access control requirements, and secure cross-chain data analysis. These methods could not provide a holistic solution to adapt with the heterogeneous nature of healthcare data while maintaining advanced privacy and security levels over the distributed networks. In this way, the present work proposes to offer a secure and scalable protocol that is based on the blockchain for healthcare cloud data samples. It integrates the following four new methodologies: Adaptive Federated Learning for Healthcare Data, Secure Homomorphic Blockchain Encryption, Dynamic Attribute-Based Encryption for Healthcare, and Proof of Healthcare Privacy (PoHP) consensus based cross-chain federated Analytics with Zero Knowledge Protocol (ZKP) for healthcare. AFL-HD would work with optimal model training over the distributed healthcare data and thereby handle the challenges that are non-IID in nature, while reducing the communication overhead by 30-40%. SHBE would ensure a 1.5x improvement in encryption and decryption times and also enable secure computations on encrypted data samples. Thus, DABE-HC enables dynamic access control policy management in blockchains, while ensuring access control precision in excess of 99%, with near-instant policy updating. CCFA-HC supports X-blockchain privacy-preserving analytics, thereby reducing the cross-chain communication overhead by 20-30%. In this protocol, therefore, cloud healthcare data management is made more scalable, secure, and private. It allows tackling challenges in the healthcare domain and gives a holistic solution supporting meaningful and secure, efficient, and collaborative healthcare data processing and analytics across distributed environments. The impact of this work is immense in providing a foundation for the next generation of secure healthcare data systems.
Permissionless blockchain operates as a fully decentralized, transparent, and immutable ledger. Preserving privacy in such systems is a complex challenge, as privacy cannot rely on restricting access or deleting data. Bitcoin, the first application of blockchain technology, was initially praised as an anonymous digital currency, but the transactions on the network have been shown to be relatively easy to trace. This realization has led to the development of advanced privacy-enhancing mechanisms with stronger anonymity guarantees. This thesis offers a comprehensive overview privacy-preserving techniques for permissionless blockchain through a systematic tertiary review of existing surveys. It identifies and categorizes key techniques such as zero-knowledge proofs, ring signatures, homomorphic encryption, secure multi-party computation and decentralized mixing protocols. Their capabilities to mitigate risks of linkability and information leakage, as well as limitations like computational overhead, are examined. Furthermore, unresolved challenges and research interests in the field are analyzed. By consolidating fragmented insights into a coherent and accessible resource, this work aims to support the privacy-aware development and adoption of blockchain applications. The findings highlight a fundamental trade-off between the privacy capabilities, efficiency, and trust assumptions of existing techniques. Privacy in permissionless blockchain often requires computationally complex cryptographic methods, leading to significant delays and increased costs for users. Efficiency can be improved by assuming some level of trust in entities or hardware, but this may conflict with the principle of decentralization. Although many powerful techniques exist, there is no universal solution, and the best results are achieved with combining techniques on a case-by-case basis. Current research on permissionless blockchain privacy focuses on improving efficiency, interoperability and usability of privacy preserving techniques. Additionally, regulatory compliance and accountability are critical concerns, as the technology must comply with privacy regulations while preventing anonymity in illegal activities such as money laundering.
This work proposes an end-to-end architecture for secure data federation and privacy-preserving analytics across multi-tenant cloud environments using homomorphic encryption (HE). We address the core challenge of enabling cross-tenant joins, aggregations, and model scoring without exposing plaintext or weakening tenant isolation. The framework integrates schema-level federation with encrypted data lakes, columnar ciphertext packing for vectorized operations, and an adaptive HE planner that selects between CKKS for approximate analytics and BFV/BGV for exact computations. To bound latency while maintaining correctness, we apply batching, ciphertext relinearization, and rotation scheduling, and offload heavy primitives to accelerator-ready microservices. Policy-aware orchestration enforces per-tenant keys via cloud KMS and supports fine-grained access control and revocation. For sensitive workflows, we compose HE with complementary protections secure enclaves for control-plane logic, differential privacy on result releases, and zero-knowledge proofs to attest query policy compliance achieving defense-in-depth without collapsing the HE trust model. The system exposes SQL-like and DataFrame APIs, a query optimizer that estimates noise budgets and bootstrapping costs, and lineage-rich audit trails for regulatory reporting. We outline deployment patterns on containerized clusters, discuss cost/performance trade-offs under realistic workloads, and provide guidance on tenancy hardening (noisy neighbor resistance, side-channel hygiene). The result is a practical pathway for organizations to collaborate on analytics and machine learning across clouds and jurisdictions while preserving confidentiality, minimizing data movement, and meeting compliance obligations
The rapid expansion of artificial intelligence (AI) in healthcare has revolutionized diagnostic practices, enabling applications such as tumor detection in medical imaging, genomic analysis, and predictive risk modeling for early disease prevention. Despite these advancements, concerns about the opacity, trustworthiness, and auditability of AI systems remain significant barriers to clinical adoption. Medical practitioners, regulators, and patients increasingly demand systems that not only produce accurate results but also provide verifiable guarantees regarding the integrity and accountability of diagnostic processes. Blockchain technology, with its intrinsic features of decentralization, immutability, and consensus-driven validation, offers a promising solution to these concerns. This manuscript investigates the integration of blockchain-powered verifiable AI models for medical diagnosis. We present a comprehensive framework that leverages federated learning for decentralized training, blockchain for immutable storage and consensus validation, and zero-knowledge proofs for cryptographic verification of model outputs. The proposed system ensures transparent audit trails, enhances data integrity, protects patient privacy, and simplifies compliance with regulatory frameworks such as HIPAA and GDPR. Through simulated case studies in medical imaging and predictive diagnostics, we demonstrate that blockchain integration improves diagnostic verifiability, reduces susceptibility to adversarial manipulation, and fosters patient-centric trust. While slight computational latency is introduced, the trade-off is justified by significantly stronger guarantees of transparency, reproducibility, and ethical accountability. This research underscores the transformative role of blockchain in shaping the future of verifiable AI-driven healthcare, providing pathways toward more reliable, transparent, and equitable medical diagnostic ecosystems.
In the realm of digital advertising, trust and transparency are vital for sustainable and effective marketing strategies. However, the industry grapples with challenges such as ad fraud, data privacy violations, and opacity in ad placements. Blockchain technology emerges as a promising solution to decode trust and foster transparency in digital advertising. Blockchain, initially developed for cryptocurrencies like Bitcoin, offers a decentralized and immutable ledger, ensuring data integrity without intermediaries. Applied to digital advertising, blockchain can revolutionize transparency by providing a tamper-proof record of ad transactions. Smart contracts automate agreements, enhancing accountability and ensuring fair compensation. Key findings reveal the potential benefits of blockchain in digital advertising. Firstly, it enhances accountability and transparency by tracing advertising funds and verifying ad authenticity. Smart contracts automate payments, reducing ad fraud and increasing supply chain accountability. Secondly, blockchain mitigates ad fraud by providing transparent ad transaction records and leveraging decentralized identity verification to authenticate users. Thirdly, transparency fosters consumer trust and confidence, empowering informed decision-making and benefiting the entire ecosystem. However, challenges persist, including technical scalability, interoperability, and regulatory compliance. Adherence to data privacy regulations and integration with existing ad tech infrastructure are crucial considerations. Organizational and cultural barriers, along with industry fragmentation, hinder blockchain adoption. Future directions and opportunities lie in tokenization, decentralized identifiers, and zero-knowledge proofs, promising solutions to existing challenges. Collaboration among stakeholders is essential to develop standards and protocols, overcoming barriers to blockchain adoption. Continued research, innovation, and collaboration are necessary to realize the full potential of blockchain in promoting trust and transparency in digital advertising. DOI: https://doi.org/10.52783/eel.v14i2.1539
Nicholas Brandt, Dennis Hofheinz, Michael KlooĂ, Michael Reichle
We construct the first blind signature scheme that achieves all of the following properties simultaneously: â it is tightly secure under a standard (i.e., non-interactive, non-q-type) computational assumption, â it does not require pairings, â it does not rely on generic, non-black-box techniques (like generic NIZK proofs). The third property enables a reasonably efficient solution, and in fact signatures in our scheme comprise 10 group elements and 29 Zp-elements. Our scheme starts from a pairing-based non-blind signature scheme (Abe et al., JoC 2023), and uses recent techniques of Chairattana-Apirom, Tessaro, and Zhu (CRYPTO 2024) to replace the pairings used in this scheme with non-interactive zero-knowledge proofs in the random oracle model. This conversion is not generic or straightforward (also because the mentioned previous works have converted only significantly simpler signature schemes), and we are required to improve upon and innovate existing techniques in several places. As an interesting side note, and unlike previous works, our techniques only require a non-programmable random oracle, and our signature scheme achieves predicate blindness (which means that the user can prove state ments about the signed message during the signing process).
We present a passwordâauthenticated (2, 3)âthreshold group key share (PATS) mechanism. Although PATS resembles threshold secret sharing schemes, it has a different structure. The innovative perspective of the PATS mechanism that makes a difference from the standard secretâsharing schemes is that it involves parties in the generation of the shares. PATS allows parties to communicate securely to establish their shares over insecure channels. Parties (shareholders) construct a secret (key) using shares obtained at the end of the protocol. PATS takes advantage of zeroâknowledge proofs compared to wellâknown threshold key exchange schemes and will tolerate the existence of semiâtrusted parties. We present two variants of PATS, centralized and distributed, and then generalize PATS to ( t , n )âthreshold scheme. PATS supports the distributed operation and optionally facilitates group key verification by a trusted third party, which may also partake in group key sharing. In this paper, we present PATS, which employs finite fields and elliptic curves, along with its security and complexity analyses.
The last decade has changed the trends of using peering networks. One of the areas of use of P2P networks is communication between people. Today, it is very important that communication is as protected as possible, especially if it is carried out between employees of the enterprise, because the number of cyber threats is constantly increasing. Modern approaches to the security of peering networks consist of data encryption, node authentication, detection and prevention of malicious nodes, access restriction, traffic monitoring, etc. However, one of the very first steps is the exchange of identification data itself, and this process must be as secure and secure as possible. The article proposes a method of secure exchange of identification data between peering network nodes, based on the use of NFC technology in combination with proof of zero knowledge. NFC is used for direct data exchange over the radio interface, which, thanks to its short range, makes it impossible to intercept data. To establish a connection, nodes must exchange identifiers, public encryption keys, and network addresses. In order to find out whether a node is not malicious, mutual verification of nodes using zero-knowledge proof is assumed. A unique identifier of the GUID type generated by each of the nodes acts as a secret that is not disclosed. Nodes first exchange public keys that encrypt and exchange identifiers. After decryption with their private keys, the nodes check whether the received value is equal to the initial one. In case of equality of values, the nodes are mutually verified and exchange identification data. The method proposed in the article is aimed at ensuring fault tolerance and confidentiality. It also provides protection against traffic interception attacks and the reliability of the verification process.
Jensen, Markus V. G., H. Kjeldsen, Nielsen, Andreas S., Niklas Bille Olesen · 5 authors
End-to-end-verifiable voting systems can only meet their goals if independent auditors are capable of verifying election outcomes as easily and efficiently as possible. ElectionGuard is a prominent effort in this direction, in which simplifying the verification process guides several design decisions, from the choice of cryptographic group to the building blocks for encryption and zero-knowledge proofs. In this paper, we present the development of optimized ElectionGuard verifiers in the Go programming language, targeting versions 1.1 and 2.0 of the specification, focused on analyzability and efficiency. Our verifiers are built on an architecture emphasizing efficiency that exploits parallelism to achieve a running time up to 10 times faster than related work. We also show that version 2.0 of the specification introduces several changes that improve the verification performance by an overall factor of 2, with the decryption process being around 24 times faster. We expect that our design can be used as a reference for future ElectionGuard verifiers.
In MLaaS, DNN models are kept in a server operated by the service provider and inputs to the DNN models are provided by the clients. Such inputs are used to execute the DNN models and classification results are sent back to the client. In MLaaS, the DNN model owner does not reveal the DNN model parameters to the client. MLaaS there are a few trust problems: (a) The server may not be secure and an attacker may send manipulated classification results to the client. In the case of safety-critical systems using such classification in the decision-making process, an attacker may specifically manipulate the classification result to disrupt the operations of the safety-critical system, (b) The server may intentionally send wrong or random classification results without executing the DNN model to respond to a massive number of classification requests from the clients. In this paper, we investigate the problem of verifying DNN model execution by the service provider in an MLaaS paradigm. A proof of DNN model execution will prove that given an input, the DNN model is executed to generate the classification result by providing sequences of outputs of all functions used in the DNN model. As the service provider in MLaaS does not share the DNN model with the client, we need to verify DNN function outcomes without the knowledge of DNN function parameters. Hence zero-knowledge proof can be used for verifying DNN model execution. In this paper, we use Zero-Knowledge Succinct Non-interactive Arguments of Knowledge (zk-SNARKs) which reduces the size of proof and complexity of proof verification considerably. In particular, we use a quadratic arithmetic program-based zkSNARK for DNN model verification. Our main results in this paper are as follows: (a) We have developed a DNN model execution verification method using a QAP-based zkSNARK. (b) We prove that the verification protocol is correct and privacy-preserving. (c) We analyzed the cost of using such a verification protocol.
The secure sharing and privacy protection of medical data have become pain points for medical data management platforms. Therefore, a secure sharing electronic health record privacy protection method based on blockchain is proposed in the study, aiming to improve data security privacy and ensure absolute ownership of patients' medical data. Attribute encryption and blockchain computing are utilized to construct a data secure sharing model, and zero-knowledge proof and ElGamal encryption algorithms are introduced to further improve the construction of data privacy protection methods. Experimental verification showed that the data secure sharing method proposed in the study has more advantages in terms of production key size and time cost. Compared with other public recognition mechanisms, zero-knowledge proof reduced the average time cost of generating keys by 54.36%. The proposed data privacy protection method had an average increase of 7.73% in protection effectiveness compared to other methods. The results indicate that the data secure sharing and privacy protection methods proposed in the study can improve the overall performance and security of the system while fully ensuring the absolute ownership of patients' data. This method has positive application value in the privacy protection of medical data.
van Trijp, Remi, Beuls, Katrien; id_orcid 0000-0003-4451-4778, Van Eecke, Paul
This paper presents a case study on how to process cooking recipes (and more generally, how-to instructions) in a way that makes it possible for a robot or artificial cooking assistant to support human chefs in the kitchen. Such AI assistants would be of great benefit to society, as they can help to sustain the autonomy of aging adults or people with a physical impairment, or they may reduce the stress in a professional kitchen. We propose a novel approach to computational recipe understanding that mimics the human sense-making process, which is narrative-based. Using an English recipe for almond crescent cookies as illustration, we show how recipes can be modelled as rich narrative structures by integrating various knowledge sources such as language processing, ontologies, and mental simulation. We show how such narrative structures can be used for (a) dealing with the challenges of recipe language, such as zero anaphora, (b) optimizing a robot's planning process, (c) measuring how well an AI system understands its current tasks, and (d) allowing recipe annotations to become language-independent.