Blockchain Papers

Lampis Alevizos

Abstract To address the challenges of internal security policy compliance and dynamic threat response in organizations, we present a novel framework that integrates artificial intelligence (AI), blockchain, and smart contracts. We propose a system that automates the enforcement of security policies, reducing manual effort and potential human error. Utilizing AI, we can analyse cyber threat intelligence rapidly, identify non-compliances and automatically adjust cyber defence mechanisms. Blockchain technology provides an immutable ledger for transparent logging of compliance actions, while smart contracts ensure uniform application of security measures. The framework’s effectiveness is demonstrated through simulations, showing improvements in compliance enforcement rates and response times compared to traditional methods. Ultimately, our approach provides for a scalable solution for managing complex security policies, reducing costs and enhancing the efficiency while achieving compliance. Finally, we discuss practical implications and propose future research directions to further refine the system and address implementation challenges.

Aditya Chauhan

Human-AI Collaboration: Exploring Synergies and Future DirectionsAditya Chauhan 11 High School Student, Department of Science, GD Goenka Public School, Kashipur, India *Correspondence should be addressed to Aditya Chauhan; aditya.chauhanx2612@gmail.com Copyright © 2024 Made Aditya Chauhan. This is an open-access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. ABSTRACT: Human-in-the-loop means a revolutionary paradigm shift in multiple fields as IT combines human-centric knowledge with Artificial Intelligence’s computation. As a part of this paper, I try to analyse how people work together with Artificial Intelligence, advantages and disadvantages of such cooperation, and their potential development. Based on the literature review of current applications, theories and practical examples of the given research, readers will receive clear and detailed insight of how combining human resources with AI can positively affect the overall performance, creativity, and the nature of decision-making.KEYWORDS: Human-AI Collaboration, Human-in-the-Loop, Artificial Intelligence Integration, Mixed-Initiative Systems, Collaborative Filtering, Ethical Considerations in AI, Explainable AI (XAI). I. INTRODUCTION The use of artificial intelligence is becoming more popular and the interface between people and newly-developed artificial intelligence agents is becoming more and more blurred. The majority of industries have introduced machine learning algorithms, natural language processors, and robotics into their structures, which altered conventional approaches [8]. It is the concept of aligning people with Artificial intelligence systems to exploit the prowess of both systems in order to execute tasks more effectively. While machines are capable of handling large datasets and have the best shot at recognizing patterns and being able to do things over and over again, humans provide contextuality, morality, and decision-making capabilities [19]. It brings about possibilities of improving problem solving and innovativeness in the outcome of the common venture.This is a very important area of discussion since Human-AI collaboration is imperative and may increase the capabilities of humans and the advancements in many fields. For example, in healthcare, artificial intelligence helps doctors and other healthcare workers in analysing patient information and images which results to proper diagnosis [35]. Likewise, in finance, AI learning patterns select the best way to trade as well as implement them through market analysis and accuracy in trading [21]. In creative industries AI will solve a work of art and find a completely different piece to create a new form of art [13]. The mere inclusion of AI in people’s daily tasks is a plus because it improves performance while opening up newer opportunities [34].In spite of these positive possibilities of Human-AI partnership, there lie some issues and concerns that should be met to allow for a proper symbiosis. One huge concern being the possibility of dependency on the systems once implemented and this would cause a decline in the use of human rationale and intelligence [33]. Furthermore, there is a risk that biases are being learned and reinforced in AI systems: They will reproduce and even Expand social injustices in many spheres of life starting from employment to credit, and policing. The question of accountability also arises when it comes to ethical considerations of using AI decision making in existence of life significant sectors such as health and the law enforcement services. It is necessary that when developing AI systems, concepts such as fairness, privacy and security have to be factored from the outset so that the resultant systems can be accepted by humans [12]. Thus, it can be concluded that implementing collaboration of Human and Artificial Intelligence is a good thing to do but nevertheless, one should do so very carefully and make sure that all possible risks are avoided. II. THEORETICAL FRAMEWORK A. Theories of Human-AI InteractionStudies in how humans can work in employment with AI is helpful in establishing relevant models of interaction. As we concluded at the end of video 1-2, Mixed-Initiative Systems present a situation in which human and AI can both each provide solutions to a problem but in different ways; always using data from the AI and knowledge from the humans as well as aesthetic judgement. It helps in decentralizing the decision-making processes so that everyone feels they are part of the process [16]. Collaborative Filtering, which is a type of recommendation system, AI is capable of filtering user behaviour and delivering content that can further improve the user’s experience. The cognitive theories include the “complementarity” of cognitive abilities where computers/AIs perform fast computations and identify patterns whilst human beings contribute with context and ethics. Combined, these models establish a symbiotic relationship that complements ways through which human beings and artificial intelligence improve decision-making [32].B. Cognitive and Behavioral AspectsThis paper reveals that there are various aspects in the cognitive and behavioral frameworks that relate to collaborative human-AI interactions and interaction dynamics. In decision making, AI assists in decision making through analysing data and presenting findings thus influencing human decision making. For example, in the medical industry, the implementation of artificial intelligence incurs the diagnosis which in turn helps the physicians to make informed decisions [15]. Nevertheless, the last word is given to a human practitioner who is capable of addressing context issues and patients’ particularities.People depend on AI systems in carrying out their activities thus making it important that they develop trust in them. Customers should have faith in the recommendations proposed by Artificial Intelligence interface and realize that AI devices cannot be perfect [26]. This means that the concept of trust as a driver of continued business is established through openness, an ability to explain activities, and actions that are sustainable over various time horizons. Besides, users have to change their working methods and instruments that are proposed by AI systems; therefore, users have to learn and modify their behaviour [31].C. Ethical Considerations in Human-AI InteractionGiven that Human-AI relations are an organic part of fundamental social tasks, questions of ethics are critical in the case of the applied use of AI systems. There are various considerations, such that, fairness or Bias in AI decision-making system is a critical concern. The AI systems mostly work on the data and so they are conditioned to work with limited data with presumptions that are embedded in bias hence lead to bias results in cases such as employment, policing, or credit rationing [3]. Accountability and controllability of AI systems is important; users have to know how the systems derive at a certain decision and if they wish to contest the decision, they should be legally allowed to do so [4]. Privacy is one of them, especially given that AI systems are often based on vast amounts of people’s information. To reduce risk exposure, it is crucial to safeguard user data and ensure that the developed AI systems conform to privacy laws. Additionally, the question of accountability arises—when AI systems make mistakes, it’s vital to determine who is responsible: by the developers, the users or directly by the AI system. This simply means that there has to be some sort of moral guideline when it comes to designing and applying AI especially in a way that will help better the lives of as many people as possible [36]. III. APPLICATION OF HUMAN-AI COLLABORATION A. Industry Application Healthcare: The integration of AI in the healthcare systems has proved beneficial due to better and efficient diagnosis and planning on the treatment to be given to the patients “[35]”. For example, the algorithms are used to read through medical images, for example, in identifying tumours in the human body. These tools support the work of radiologists in identify problems and, therefore, increase the probability of correct diagnoses and prompt treatment. Other benefits of utilizing such technologies include the use of specialised analysis for choosing treatments that can match the specific genetic makeup of patients [15].Finance: When it comes to trading, risk management as well as detection of fraud, then AI is an essential tool in financial service industry [21]. Trading is made more efficient because AI algorithms sift through the available market data looking for trends and then act when it is most appropriate. Furthermore, risk management via AI prescribes the likelihood of certain risks and then determines how to avoid them and risk-based fraud detection whereby algorithms analyse transaction patterns to detect fraudulent activities are other applications of AI [25].Manufacturing: AI systems also aid in increased efficiency in manufacturing through aspects such as maintenance predetermination, quality assurance, and improve on processes. Through performance measurement of the equipment and failure prognosis, AI helps to decrease the time when they are not in service and the costs for repair [27]. Condition monitoring tools work with historical data and data from the equipment sensors in order to estimate probable failures and perform preventive actions. AI also contributes to the effectiveness of quality assurance through the examination of the production data for the flaws and enhancement of the manufacturing procedures [41]. B. Creative DomainsArt and Music: Technological advancements, especially the Application of Artificial intelligence in the various fields, has made new forms of creativity possible. AI based art and music aims at creating new forms and styles of art through unique algorithms which are derived from certain given art and music pieces [13]. For instance, the current AI systems such as DALL-E and MuseNet help artists and musicians to create different pieces of work using both conventional and AI-driven methods. Such tools allow artists to broaden the range of options and overstep the limitations within art making [17]. Writing: Grammarly and GPT based tools are AI writing assistants that assist the writers through making corrections in the grammatical errors and blunders and enhancing the style and content of the document. These tools simplify the writing process in a way that delivering effective feedbacks at real time and improving the quality of the content .AI driven writing tools also helped in coming up with ideas and in making the content writing process a lot faster and more effective and in helping in the creation of interesting and well-written content [37].C. Everyday LifePersonal Assistants: Siri and Google Assistant – are AI bots that assist the users to perform the daily activities by reminding them, answering their questions and controlling the smart appliances [29]. These assistants enhance efficiency and ease since many tasks are repetitive, and the information acquired can be retrieved quickly. Personal assistants also synchronise with other Artificial Intelligent systems including smart home gadgets to offer a combined experience [23].Productivity Tools: Including the applications of artificial intelligence at work allows for increased productivity through the reduction of the number of monotonous tasks like scheduling a meeting or sorting an email [9]. These tools let the user to concentrate on increasing organizational effectiveness and strategic planning and avoid trivial matters. For instance, smart email triage tools sort and prioritize emails and therefore lessen time spent on this activity while one can focus on other important tasks [29]. IV. CHALLENGES AND LIMITATIONSA. Technical ChallengesReliability: It can be also claimed that AI systems can generate incorrect results based on various parameters of data quality or further more on the algorithm bias [2]. Therefore, that strengthen and accuracy of the AI structures are paramount fundamentals for the application of such systems in sensitive areas. It is integrated with ongoing monitoring and validation to solve potential problems as well as achieving ideal performance. Also, the self-learning ability of AI systems is always required to be updated to other changes and other data [33].Integration: It must be said that the integration of AI systems is rather challenging when it comes to an organization’s existing business processes and technological environments [8]. There are challenges that organizations face in regards to compatibility and compatibility in the integration of AI systems and current processes. This is often time consuming, capital intensive and can only be achieved by an organization with competent professionals. Hence, the integration of AI should be in a way that synchronizes the implementation of these systems with the firm’s objectives, operational work flow, and the requirements of the end users [10].B. Ethical and Social Challenges Privacy and Security: Most of the AI systems are based on the big data causing privacy and security issues [42]. Data protection and its conformity with data protection regulation must be conducted responsibly to promote the users trust towards AI systems. It is imperative that organizations today have strong measures to protect the data and the use of that data should be made clear and transparent [36].Bias and Fairness: An AI algorithm for example is a machine learning model that can be biased to certain data set when programmed to make a particular prediction [3]. Overcoming these biases and achieving equal treatment in applications of artificial intelligence the process is continuous. There are some measures that need to be employed in order to alleviate bias such as use of various and fair datasets, use of fairness algorithms among others [4].Employment: There is evidence that shows how the use of AI in the performance of tasks may lead to unemployment as these systems replace employees in various industries. It is imperative to reskill and skill up the workforce to achieve changes in task portfolios and mitigate the impacts of automation on employment [6]. Government and non-government sectors should work hand in hand for the development of policies which can help in the changing needs of the employees along with encouragement for continuous learning [1].C. Human FactorsTrust and Acceptance: One of the most significant concerns which are relevant to interaction with AI systems is the deficit of trust between individuals or organizations. Users must be sure that recommendations given by AI are correct and dependable [26]. Transparency, explainability, and stability play their role in building trust and acceptance from the users’ side. Educating the users on the flow of decision making by the artificial intelligence systems and reassuring users on the reliability of the AI can help in improving the trust [29].Training and Adaptation: People should understand how to properly deal with an AI system since most of the time they are specifically programmed and designed to do one task. This is the reason why we need to be aware of how it works, how to read its output and how to incorporate it into the actual work process [31]. This implies that support and training have to be provided continuously in order to realise the full potential of AI collaboration and integration [9]. V. FUTURE DIRECTIONSA. Advancements in AI TechnologyExplainable AI (XAI): explainable artificial intelligence seeks to enhance the level of trust in artificial intelligence-based systems by making their decision-making processes comprehensible to the users. By using Explainable AI, the user is able to understand and accept AI’s recommendations and thus the incorporation makes collaboration easier [18]. The research in XAI is centred on developing models that allow for expounding the rationale behind their patterns Human-in-the-loop systems implies of the AI processes by the human This therefore human decision making and AI decision making whereby decision made is based on certain ethical or operational considerations Human-in-the-loop allows for in the systems and also means that human knowledge is into a system that on artificial intelligence that models of human to AI interaction is very The also help increase work efficiency and the users’ interaction with AI should be better designed with the help of and and Such approaches of AI systems depend on the user’s and requirements for the of the systems. are based on the users’ recommendations It helps to increase the level of user and make sure that the AI systems are the most for the user’s needs and Ethical Ethical The of a of to when AI is one of the why there is a need for ethical In a these should include privacy fairness, and accountability An example is that the and the industries that are in the development of AI systems can work together to up with the ethical for use by systems These approaches of using AI people to in thus making it more AI means about the of different people and AI is as well in increasing the number of among the working on the of AI will ensure that the of and efficiency in interaction between humans and artificial intelligence Healthcare: for for is an applied artificial system capable of helping and Based on medical and patients’ the system effective treatment of various of the integration of with because the system helps to make better diagnosis and treatment based on the of data and users in this all of which should be well to from this Intelligence tool of is the use of artificial intelligence to essential from It the time in activities such as review of and among to perform tasks that do not while the can work on problems The applied use of AI in the processes reveals for improving efficiency and performance in the financial Creative is an Artificial tool to new with the help of users and other It learning in the of images to turn them over into Customers are given the ability to create based on such as users’ and AI derived patterns [13]. is a perfect example of how AI tools can creativity, and towards new forms of possibilities [17]. Human and Artificial Intelligence in a number of industries is for example it be applied to the for finance, and many AI both analysis and experience to generate better results and learning to the organization [8]. there are some issues that needs to be for effective such as ethical issues and the problems with people AI and human on the has with the help of the of the of AI into the ethical of combining a human and an artificial The trends in the AI systems will improve people’s and such as and AI [18]. Through the of such and the people and Artificial Intelligence can the better together in There is need for to incorporate training that will the users to with the AI systems. This from the AI system is capable of being in a to understand the recommendations given by the system to the way of the AI system into the existing working Ethical The of AI solutions that and a specific set of ethical to avoid more and development of AI as well as its will the development in this OF that they have of AI, and of of the which need in the AI A. I I An of the of Human-AI in on and Transparency, AI and This paper aims at the of and the are and A. The and in an of and The of Human in AI on and Transparency, out how Artificial Intelligence will the of B. a of AI system that bias Creative by and from V. and the B. A. with at Mixed-Initiative and of AI A. for to the of and Artificial Intelligence (XAI). V. B. a of which is – of the on The of Artificial Intelligence on the financial A. Intelligent Personal An of the likelihood model on trust in artificial in the of A. AI Systems for on Systems, for A. in and Human B. A. as a for Industry B. in and It an of of A. Personal in the and of A. for on Accountability Transparency, fair A. The of Everyday and I have the and of this Human and of Social of Data and B. Artificial of in and Challenges of on and Social is The for in and in artificial AI Systems for AI, and AI Systems for of in AI and Data Systems, The of The for a Human Future at the of

Ismael Medeiros, Fausto Carvalho, Alexandre Casadei‐Ferreira, Rodrigo Bonifácio · 5 authors

Ethereum is a distributed, peer-to-peer blockchain infrastructure that has attracted billions of dollars. Perhaps due to its success, Ethereum has become a target for various kinds of attacks, motivating researchers to explore different techniques to identify vulnerabilities in EVM bytecode (the language of the Ethereum Virtual Machine)—including formal verification, symbolic execution, and fuzz testing. Although recent studies empirically compare smart contract fuzzers, there is a lack of literature investigating how simpler grey-box fuzzers compare to more advanced ones. To fill this gap, in this paper, we present DogeFuzz, an extensible infrastructure for fuzzing Ethereum smart contracts, currently supporting black-box fuzzing and two grey-box fuzzing strategies: coverage-guided grey-box fuzzing (DogeFuzz-G) and directed grey-box fuzzing (DogeFuzz-DG). We conduct a series of experiments using benchmarks already available in the literature and compare the DogeFuzz strategies with state-of-the-art fuzzers for smart contracts. Surprisingly, although DogeFuzz does not leverage advanced techniques for improving input generation (such as symbolic execution or machine learning), DogeFuzz outperforms sFuzz and ILF, two state-of-the-art fuzzers. Nonetheless, the Smartian fuzzer shows higher code coverage and bug-finding capabilities than DogeFuzz.

Aryan Patel, Kartikeya Chauhan, Saarthak Maini, Mukta Goyal

This paper into the critical issue of vulnerability detection in smart contracts, focusing on identifying vulnerabilities, proposing mitigation strategies, and developing techniques for detecting Ponzi schemes within smart contracts. By understanding and addressing these vulnerabilities, we aim to enhance the security and robustness of blockchain-based applications and ecosystems.

Xiaolei Ren, Qi-Ping Wei

The advent of blockchain technology has revolutionized various sectors by providing transparency, immutability, and automation. Central to this revolution are smart contracts, which facilitate trustless and automated transactions across diverse domains. However, the proliferation of smart contracts has exposed significant security vulnerabilities, necessitating advanced analysis techniques. Data dependency analysis is a critical program analysis method used to enhance the testing and security of smart contracts. This paper introduces Sligpt, an innovative methodology that integrates a large language model (LLM), specifically GPT-4o, with the static analysis tool Slither, to perform data dependency analyses on Solidity smart contracts. Our approach leverages both the advanced code comprehension capabilities of GPT-4o and the advantages of a traditional analysis tool. We empirically evaluate Sligpt using a curated dataset of Ethereum smart contracts. Sligpt achieves significant improvements in precision, recall, and overall analysis depth compared with Slither and GPT-4o, providing a robust solution for data dependency analysis. This paper also discusses the challenges encountered, such as the computational resource requirements and the inherent variability in LLM outputs, while proposing future research directions to further enhance the methodology. Sligpt represents a significant advancement in the field of static analysis on smart contracts, offering a practical framework for integrating LLMs with static analysis tools.

Rakibul Hasan Chowdhury

The integration of Blockchain technology and Artificial Intelligence (AI) is revolutionizing data management and business intelligence. Blockchain, with its decentralized, immutable ledger, ensures data integrity and security, while AI enhances data analysis through advanced algorithms and predictive capabilities. This article explores the synergy between these two transformative technologies, examining how their combined strengths can address modern challenges in data security and business operations. The paper begins with an overview of Blockchain and AI, detailing their foundational principles and recent advancements. It then delves into their applications in enhancing data security, highlighting Blockchain's role in providing encryption and immutability and AI's capabilities in threat detection and response. The discussion extends to their impact on business intelligence, showcasing how Blockchain contributes to transparent and verifiable data, while AI drives advanced analytics and decision-making. Real-world case studies illustrate successful implementations of Blockchain and AI integration, demonstrating their potential to revolutionize various industries. The article also addresses technical challenges, privacy concerns, and regulatory issues associated with these technologies. Finally, it outlines future directions for research and innovation, emphasizing the need for continued exploration of their combined potential. By providing a comprehensive analysis of Blockchain and AI's transformative impact, this article aims to offer valuable insights for researchers, practitioners, and policymakers seeking to leverage these technologies for improved data security and business intelligence.

Deekshitha Kosaraju

Federated Learning (FL) is a technique in the field of machine learning that prioritizes privacy by allowing collaborative model training without revealing data. This article explores the basics of FL and its importance in protecting data privacy in sectors such as healthcare, finance, and industrial engineering. By using data sources FL enables the development of strong and adaptable AI models without centralizing sensitive information. We delve into the methodologies behind FL including secure multiparty computation, differential privacy, and homomorphic encryption. Additionally, we look at the ways FL is used, such as speeding up medical research improving financial security and streamlining industrial processes. The challenges related to FL - like communication diverse data distributions and scalability - are also addressed. Lastly, we discuss trends, in FL that focus on enhancing privacy techniques and complying with regulations. This thorough overview highlights how FL can revolutionize AI advancement while upholding privacy standards. Keywords: Federated Learning, Privacy Preservation, Decentralized Machine Learning, Secure Multiparty Computation, Differential Privacy, Healthcare AI, Industrial Engineering, Data Silos, Collaborative Learning.

Tiago Nunes, Paulo Rupino da Cunha, João Mendes de Abreu, J. Duarte · 5 authors

Amid global health challenges, resilient health systems require continuous innovation and progress. Stakeholders highlight the critical role of digital technologies in accelerating this progress. However, the digital health field faces significant challenges, including the sensitivity of health data, the absence of evidence-based standards, data governance issues, and a lack of evidence on the impact of digital health strategies. Overcoming these challenges is crucial to unlocking the full potential of digital health innovations in enhancing healthcare delivery and outcomes. Prioritizing security and privacy is essential in developing digital health solutions that are transparent, accessible, and effective. Non-fungible tokens (NFTs) have gained widespread attention, including in healthcare, offering innovative solutions and addressing challenges through blockchain technology. This paper addresses the gap in systematic-level studies on NFT applications in healthcare, aiming to comprehensively analyze use cases and associated research challenges. The search included primary studies published between 2014 and November 2023, searching in a balanced set of databases compiling articles from different fields. A review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework and strictly focusing on research articles related to NFT applications in the healthcare sector. The electronic search retrieved 1902 articles, ultimately resulting in 15 articles for data extraction. These articles span applications of NFTs in medical devices, pathology exams, diagnosis, pharmaceuticals, and other healthcare domains, highlighting their potential to eliminate centralized trust sources in health informatics. The review emphasizes the adaptability and versatility of NFT-based solutions, indicating their broader applicability across various healthcare stages and expansion into diverse industries. Given their role in addressing challenges associated with enhancing data integrity, availability, non-repudiation, and authentication, NFTs remain a promising avenue for future research within digital health solutions.

Giacomo Ibba, Sabrina Aufiero, Rumyana Neykova, Silvia Bartolucci · 7 authors

Smart contracts (SCs) significance and popularity increased exponentially with the escalation of decentralised applications (dApps), which revolutionised programming paradigms where network controls rest within a central authority. Since SCs constitute the core of such applications, developing and deploying contracts without vulnerability issues become key to improve dApps robustness to external attacks. This paper introduces a dataset that combines smart contract metrics with vulnerability data identified using Slither, a leading static analysis tool proficient in detecting a wide spectrum of vulnerabilities. Our primary goal is to provide a resource for the community that supports exploratory analysis, such as investigating the relationship between contract metrics and vulnerability occurrences. Further, we discuss the potential of this dataset for the development and validation of predictive models aimed at identifying vulnerabilities, thereby contributing to the enhancement of smart contract security. Through this dataset, we invite researchers and practitioners to study the dynamics of smart contract vulnerabilities, fostering advancements in detection methods and ultimately, fortifying the resilience of smart contracts.

Sundas Munir, Walid Taha

Smart contracts are programs with mutable state. Transactions submitted to these contracts trigger functions that often modify state. Nodes of the Ethereum blockchain schedule such transactions in a nondeterministic order, potentially leading to races between transactions and concurrency issues. When the outcome of a smart contract varies depending on the order in which transactions are processed, we have a Transaction Ordering Dependency (TOD). TOD enables malicious actors to profit from a smart contract, similar to the well-known frontrunning vulnerability. Existing approaches for detecting TOD in Ethereum smart contracts yield a high rate of false positives and false negatives. To help contract developers and testers detect TOD vulnerabilities with enhanced precision, we propose and evaluate an analysis based on information flow in our tool TODChecker1. We evaluate our approach using a benchmark comprising 513 vulnerable transactions involving 235 real-world Ethereum smart contracts susceptible to frontrunning attacks. Our evaluation finds that our approach outperforms existing approaches, including Oyente, Securify, SAILFISH, TODler, and Nyx, in precision, runtime, and in identifying novel TOD vulnerabilities.

Dimitris Ntalaperas

This deliverable presents the second iteration of the AI4Gov Decentralised Data Governance (DDG) model, focusing on the implementation of mechanisms that ensure data provenance, reliability, and GDPR-compliant privacy within a decentralized architecture. It details the finalized system design and supporting prototypes that enable transparent data governance and the execution of decentralized business processes through smart contracts, while introducing a redesigned, citizen-centric approach that facilitates participation in open and collaborative governance processes. The document outlines key architectural improvements, including the adoption of decentralized identity frameworks and the integration of Digital Autonomous Organization (DAO) principles for self-governed units. It further discusses the updated technology stack, system functionalities, and relevant regulatory considerations, providing a validated and extensible framework for trustworthy, participatory, and privacy-preserving data governance in AI-driven public-sector applications.

Zeqin Liao, Yuhong Nan, Henglong Liang, Sicheng Hao · 7 authors

With the increasing popularity of blockchain, different blockchain platforms coexist in the ecosystem (e.g., Ethereum, BNB, EOSIO, etc.), which prompts the high demand for cross-chain communication. Cross-chain bridge is a specific type of decentralized application for asset exchange across different blockchain platforms. Securing the smart contracts of cross-chain bridges is in urgent need, as there are a number of recent security incidents with heavy financial losses caused by vulnerabilities in bridge smart contracts, as we call them Cross-Chain Vulnerabilities (CCVs). However, automatically identifying CCVs in smart contracts poses several unique challenges. Particularly, it is non-trivial to (1) identify application-specific access control constraints needed for cross-bridge asset exchange, and (2) identify inconsistent cross-chain semantics between the two sides of the bridge. In this paper, we propose SmartAxe, a new framework to identify vulnerabilities in cross-chain bridge smart contracts. Particularly, to locate vulnerable functions that have access control incompleteness, SmartAxe models the heterogeneous implementations of access control and finds necessary security checks in smart contracts through probabilistic pattern inference. Besides, SmartAxe constructs cross-chain control-flow graph (xCFG) and data-flow graph (xDFG), which help to find semantic inconsistency during cross-chain data communication. To evaluate SmartAxe, we collect and label a dataset of 88 CCVs from real-attacks cross-chain bridge contracts. Evaluation results show that SmartAxe achieves a precision of <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline"> <mml:mn>84.95</mml:mn> <mml:mo>%</mml:mo> </mml:math> and a recall of <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline"> <mml:mn>89.77</mml:mn> <mml:mo>%</mml:mo> </mml:math> . In addition, SmartAxe successfully identifies 232 new/unknown CCVs from 129 real-world cross-chain bridge applications (i.e., from 1,703 smart contracts). These identified CCVs affect a total amount of digital assets worth 1,885,250 USD.

Saltanat Mohammed Abdullah Shaikh

Background: Research has been done on the vulnerabilities of Ethereum smart contract detection since the emergence of blockchain technologies. Ethereum is one of the most popular platforms for DApps (decentralized applications) and smart contracts but turns more undoubtedly when their number and popularity grow. Methods: The study evaluates different detection methods including static analysis, dynamic code analysis, symbolic execution, and machine learning. Findings: The performance metrics on key areas, e.g. detection time, true positive rate, false positive rate, and scalability are emphasized in this evaluation analysis. These inferences imply that although Static Analysis can provide fast detection and high accuracy, Machine Learning is better at High scalability. The study also identifies trending flaws often encountered such as re-entrancy attacks and lack of input validation and stresses further the necessity of strong security methods. Besides, you may consider the sensitivity analysis in different network load scenarios as it shows the efficiency of detection technique in changing operational settings. Novelty and applications: Overall, the research brings a reliable development to smart contracts in Ethereum's security industries through analyzing and profiling vulnerability types and performance metrics that inform the development of more stable and efficient security activities for distributed applications.

Isabirye Edward Kezron

The rapid advancement and integration of Artificial Intelligence (AI) across critical sectors — including healthcare, finance, defense, and infrastructure — have exposed an often-overlooked risk: vulnerabilities within the AI supply chain. This research examines the security challenges and potential threats affecting AI model development and deployment, focusing on adversarial attacks, data poisoning, model theft, and compromised third-party components. By dissecting the AI supply chain into its core stages — data sourcing, model training, deployment, and maintenance — this study identifies key entry points for malicious actors. The paper proposes a multi-layered security framework combining blockchain-based data provenance, federated learning for decentralized model training, and zero-trust architecture to ensure secure deployment. Additionally, it explores how adversarial training, model watermarking, and real-time anomaly detection can mitigate risks without sacrificing model performance. Case studies of high-profile AI breaches are analyzed to demonstrate the consequences of unsecured pipelines, emphasizing the urgency of securing AI systems.

Federico Costantini, Francesco Crisci, Silvia Venier, Stefano Bistarelli · 5 authors

Tackling disinformation is crucial for the development of the Information Society. To do this, it is necessary to empower journalists in the production of trustworthy information, and to nurture an economic ecosystem centred on the secure circulation of content. In this contribution we present an interdisciplinary approach that aims (1) to find a balance between freedom of expression and other fundamental rights (e.g., privacy and data protection), (2) to develop business models driven by the production of genuine content, and (3) to exploit the potential of distributed ledger systems to provide media certification.

Zhiyang Chen, Ye Liu, Sidi Mohamed Beillahi, Yi Li · 5 authors

Smart contract transactions associated with security attacks often exhibit distinct behavioral patterns compared with historical benign transactions before the attacking events. While many runtime monitoring and guarding mechanisms have been proposed to validate invariants and stop anomalous transactions on the fly, the empirical effectiveness of the invariants used remains largely unexplored. In this paper, we studied 23 prevalent invariants of 8 categories, which are either deployed in high-profile protocols or endorsed by leading auditing firms and security experts. Using these well-established invariants as templates, we developed a tool Trace2Inv which dynamically generates new invariants customized for a given contract based on its historical transaction data. We evaluated Trace2Inv on 42 smart contracts that fell victim to 27 distinct exploits on the Ethereum blockchain. Our findings reveal that the most effective invariant guard alone can successfully block 18 of the 27 identified exploits with minimal gas overhead. Our analysis also shows that most of the invariants remain effective even when the experienced attackers attempt to bypass them. Additionally, we studied the possibility of combining multiple invariant guards, resulting in blocking up to 23 of the 27 benchmark exploits and achieving false positive rates as low as 0.32%. Trace2Inv outperforms current state-of-the-art works on smart contract invariant mining and transaction attack detection in terms of both practicality and accuracy. Though Trace2Inv is not primarily designed for transaction attack detection, it surprisingly found two previously unreported exploit transactions, earlier than any reported exploit transactions against the same victim contracts.

Petar Radanliev

In the ever-evolving realm of cybersecurity, the increasing integration of Metaverse systems with cutting-edge technologies such as Artificial Intelligence (AI), Blockchain, and Cloud Computing presents a host of new opportunities alongside significant challenges. This article employs a methodological approach that combines an extensive literature review with focused case study analyses to examine the changing cybersecurity landscape within these intersecting domains. The emphasis is particularly on the Metaverse, exploring its current state of cybersecurity, potential future developments, and the influential roles of AI, blockchain, and cloud technologies. Our thorough investigation assesses a range of cybersecurity standards and frameworks to determine their effectiveness in managing the risks associated with these emerging technologies. Special focus is directed towards the rapidly evolving digital economy of the Metaverse, investigating how AI and blockchain can enhance its cybersecurity infrastructure whilst acknowledging the complexities introduced by cloud computing. The results highlight significant gaps in existing standards and a clear necessity for regulatory advancements, particularly concerning blockchain’s capability for self-governance and the early-stage development of the Metaverse. The article underscores the need for proactive regulatory involvement, stressing the importance of cybersecurity experts and policymakers adapting and preparing for the swift advancement of these technologies. Ultimately, this study offers a comprehensive overview of the current scenario, foresees future challenges, and suggests strategic directions for integrated cybersecurity within Metaverse systems utilising AI, blockchain, and cloud computing.

Zexu Wang, Jiachi Chen, Yanlin Wang, Yu Zhang · 6 authors

Reentrancy vulnerability as one of the most notorious vulnerabilities, has been a prominent topic in smart contract security research. Research shows that existing vulnerability detection presents a range of challenges, especially as smart contracts continue to increase in complexity. Existing tools perform poorly in terms of efficiency and successful detection rates for vulnerabilities in complex contracts. To effectively detect reentrancy vulnerabilities in contracts with complex logic, we propose a tool named SliSE. SliSE’s detection process consists of two stages: Warning Search and Symbolic Execution Verification . In Stage I, SliSE utilizes program slicing to analyze the Inter-contract Program Dependency Graph (I-PDG) of the contract, and collects suspicious vulnerability information as warnings. In Stage II, symbolic execution is employed to verify the reachability of these warnings, thereby enhancing vulnerability detection accuracy. SliSE obtained the best performance compared with eight state-of-the-art detection tools. It achieved an F1 score of 78.65%, surpassing the highest score recorded by an existing tool of 9.26%. Additionally, it attained a recall rate exceeding 90% for detection of contracts on Ethereum. Overall, SliSE provides a robust and efficient method for detection of Reentrancy vulnerabilities for complex contracts.

Oyegoke Oyebode

The growing reliance on artificial intelligence in decision-making processes has intensified debates over bias, fairness, and accountability in algorithmic governance. While deep learning models deliver unprecedented predictive performance, their “black box” nature has undermined transparency and public trust, particularly in high-stakes applications such as finance, healthcare, and digital public services. Explainable AI (XAI) has emerged to address this gap by making model reasoning interpretable, yet explainability alone cannot guarantee fairness without verifiable systems of identity and accountability. This study proposes a framework that integrates explainable deep learning with decentralized identity (DID) systems to combat bias, enhance trust, and ensure equitable governance outcomes. In this framework, explainable deep learning models provide human-understandable insights into algorithmic decisions, enabling stakeholders to evaluate reasoning processes. Meanwhile, decentralized identity systems built on blockchain technologies ensure that individuals retain control over their digital identities, reducing risks of centralized manipulation and exclusion. By linking interpretable models with verifiable identity protocols, algorithmic governance can achieve both transparency and fairness while protecting privacy. The integration enables bias detection and correction at both the model and system levels: interpretable models flag discriminatory features, while decentralized identity guarantees equitable access across diverse populations. Applications in digital voting, welfare distribution, and credit scoring illustrate how the framework strengthens accountability and prevents systemic marginalization. Ultimately, combining explainable deep learning with decentralized identity provides a path toward trustworthy and fair algorithmic governance, where decisions are not only accurate but also transparent, inclusive, and ethically aligned with societal values.

Igor Calzada

This article investigates the intricate dynamics of data monopolies, referred to as “data-opolies”, and their implications for democratic erosion. Data-opolies, typically embodied by large technology corporations, accumulate extensive datasets, affording them significant influence. The sustainability of such data practices is critically examined within the context of decentralized Web3 technologies amidst Artificial Intelligence (AI) disruption. Additionally, the article explores emancipatory datafication strategies to counterbalance the dominance of data-opolies. It presents an in-depth analysis of two emergent phenomena within the decentralized Web3 emerging landscape: People-Centered Smart Cities and Datafied Network States. The article investigates a paradigm shift in data governance and advocates for joint efforts to establish equitable data ecosystems, with an emphasis on prioritizing data sovereignty and achieving digital self-governance. It elucidates the remarkable roles of (i) blockchain, (ii) decentralized autonomous organizations (DAOs), and (iii) data cooperatives in empowering citizens to have control over their personal data. In conclusion, the article introduces a forward-looking examination of Web3 decentralized technologies, outlining a timely path toward a more transparent, inclusive, and emancipatory data-driven democracy. This approach challenges the prevailing dominance of data-opolies and offers a framework for regenerating datafied democracies through decentralized and emerging Web3 technologies.

Randhir Kumar, Ahamed Aljuhani, Danish Javeed, Prabhat Kumar · 6 authors

Data-driven modeling using Artificial Intelligence (AI) is envisioned as a key enabling technology for Zero Touch Network (ZTN) management. Specifically, AI has shown huge potential for automating and modeling the threat detection mechanism of complicated wireless systems. The current data-driven AI systems, however, lack transparency and accountability in their decisions, and assuring the reliability and trustworthiness of the data collected from participating entities is an important obstacle to threat detection and decision-making. To this end, we integrate smart contracts with eXplainable AI (XAI) to design a robust cybersecurity framework for ZTN. The proposed framework uses a blockchain and smart contract-enabled access control and authentication mechanism to ensure trust among the participating entities. Additionally, with the collected data, we designed Digital Twins (DTs) for simulating the attack detection operation in the ZTN environment. Specifically, to provide a platform for analysis and the development of an Intrusion Detection System (IDS), the DTs are equipped with a variety of process-aware attack scenarios. A Self Attention-based Long Short Term Memory (SALSTM) network is used to evaluate the attack detection capabilities of the proposed framework. Furthermore, the explainability of the proposed AI-based IDS is achieved using the SHapley Additive exPlanations (SHAP) tool. The experimental results using N-BaIoT and a self-generated DTs dataset confirm the superiority of the proposed framework over some baseline and state-of-the-art techniques. • A new robust cybersecurity framework for ZTN is proposed by integrating smart contracts with eXplainable AI. • To ensure secure communication, a novel blockchain-enabled key establishment and access control mechanism is proposed that authenticates the participating entities in ZTN. The temper-proof property of blockchain ensures high integrity of the data enrichment and builds trust between the participating entities of blockchain network. A smart contract enabled Proof-of-Authority (PoA) consensus mechanism is used to verify and validate the transactions or data. • The authenticated data from smart contract and blockchain-enabled authentication scheme is used to design DT for simulating the attack detection operation in ZTN environment. In particular, the DT is set up with a range of process aware attack scenarios to provide a platform for study and the creation of Intrusion Detection System (IDS). To assess the proposed framework’s ability to identify attacks, a Self Attention-based Long Short Term Memory (SALSTM) network is deployed. Additionally, utilizing the SHapley Additive exPlanations (SHAP) tool, the proposed AI-based IDS is made explainable. • Through experiments using an actual DT simulated dataset and state-of-the-art intrusion dataset (N-BaIoT) is used to evaluate the proposed framework. The outcomes are compared with some baseline and state-of-the-art techniques to show the effectiveness of the proposed cybersecurity framework.

Chhavi Yadav, Amrita Roy Chowdhury, Dan Boneh, Kamalika Chaudhuri

Machine learning models are increasingly used in societal applications, yet legal and privacy concerns demand that they very often be kept confidential. Consequently, there is a growing distrust about the fairness properties of these models in the minds of consumers, who are often at the receiving end of model predictions. To this end, we propose \name -- a system that uses Zero-Knowledge Proofs (a cryptographic primitive) to publicly verify the fairness of a model, while maintaining confidentiality. We also propose a fairness certification algorithm for fully-connected neural networks which is befitting to ZKPs and is used in this system. We implement \name in Gnark and demonstrate empirically that our system is practically feasible. Code is available at https://github.com/infinite-pursuits/FairProof.

Bianca-Mihaela Ganescu, Jonathan Passerat‐Palmbach

Generative AI, exemplified by models like transformers, has opened up new possibilities in various domains but also raised concerns about fairness, transparency and reliability, especially in fields like medicine and law. This paper emphasizes the urgency of ensuring fairness and quality in these domains through generative AI. It explores using cryptographic techniques, particularly Zero-Knowledge Proofs (ZKPs), to address concerns regarding performance fairness and accuracy while protecting model privacy. Applying ZKPs to Machine Learning models, known as ZKML (Zero-Knowledge Machine Learning), enables independent validation of AI-generated content without revealing sensitive model information, promoting transparency and trust. ZKML enhances AI fairness by providing cryptographic audit trails for model predictions and ensuring uniform performance across users. We introduce snarkGPT, a practical ZKML implementation for transformers, to empower users to verify output accuracy and quality while preserving model privacy. We present a series of empirical results studying snarkGPT's scalability and performance to assess the feasibility and challenges of adopting a ZKML-powered approach to capture quality and performance fairness problems in generative AI models.

Simona Ramos, Joshua Ellul

Abstract The article aims to investigate the potential of blockchain technology in mitigating certain cybersecurity risks associated with artificial intelligence (AI) systems. Aligned with ongoing regulatory deliberations within the European Union (EU) and the escalating demand for more resilient cybersecurity measures within the realm of AI, our analysis focuses on specific requirements outlined in the proposed AI Act. We argue that by leveraging blockchain technology, AI systems can align with some of the requirements in the AI Act, specifically relating to data governance, record-keeping, transparency and access control. The study shows how blockchain can successfully address certain attack vectors related to AI systems, such as data poisoning in trained AI models and data sets. Likewise, the article explores how specific parameters can be incorporated to restrict access to critical AI systems, with private keys enforcing these conditions through tamper-proof infrastructure. Additionally, the article analyses how blockchain can facilitate independent audits and verification of AI system behaviour. Overall, this article sheds light on the potential of blockchain technology in fortifying high-risk AI systems against cyber risks, contributing to the advancement of secure and trustworthy AI deployments. By providing an interdisciplinary perspective of cybersecurity in the AI domain, we aim to bridge the gap that exists between legal and technical research, supporting policy makers in their regulatory decisions concerning AI cyber risk management.