Blockchain Papers

Liang Guo

Game-based teaching (GBT) has gained widespread adoption in modern education, yet teachers bear heavy burdens in designing GBT activities and interpreting student learning performance, while centralized educational data storage brings prominent security and credibility risks. To tackle the above bottlenecks, this paper proposes SmartTA, an integrated teaching assistant system combining GBT recommendation modules, automated machine learning (AutoML), and blockchain. Specifically, SmartTA supplies customized GBT cases and exam scoring suggestions for teachers, and leverages AutoML to automatically mine student learning behaviors with zero coding requirements. Three groups of experiments are conducted to validate the system: AutoML achieves a maximum prediction accuracy of 93% on six public educational datasets; the Hyperledger Fabric-based blockchain prototype enables data insertion with an average latency of approximately 2.2 seconds and query latency of approximately 150 ms; 20 frontline educational practitioners provide 85% positive user feedback. The experimental results suggest that SmartTA may help reduce teachers’ lesson preparation workload, support improved instructional quality, while enabling tamper-resistant data storage via blockchain. This study realizes the practical fusion of AutoML and blockchain for GBT scenarios, and establishes a novel, secure, data-driven teaching assistance paradigm that is accessible to non-technical educators.

Fuan Xiao, Jiahui Huang, Jia-Xin Huang, Jia-Xin Huang · 7 authors

Abstract Online Education in Higher Education is rapidly evolving through the integration of Large Language Model (LLM)-powered intelligent systems, which enable personalized tutoring, dynamic content generation, and automated assessment. However, the widespread adoption of LLMs in education is hampered due to their inherent limitations, including susceptibility to hallucinations, insufficient domain-specific knowledge validation, and output inconsistency. These deficiencies can lead to misleading or erroneous content, potentially causing significant negative learning outcomes. A core challenge lies in ensuring that such errors are immutably logged and traceable, thereby establishing a mechanism for accountability among the entities deploying these LLM services. To address these challenges, this paper proposes a novel framework that Integrates LLM with consortium blockchain for personalized and verifiable online education. Our design features a synergistic architecture in which LLM based services provide the intelligent educational interface, while a permissioned consortium blockchain serves as a secure and tamper proof ledger. This blockchain infrastructure records critical educational transactions ranging from learning process data and academic credentials to the outputs generated by the LLMs. This integration not only secures academic credentials but also establishes a fully auditable trail, making it possible to trace responsibility for educational deficiencies caused by AI errors. Collectively, this work demonstrates a robust and accountable framework for leveraging LLMs in education, effectively mitigating the risks of AI inaccuracies through the verifiable and immutable nature of consortium blockchain.

Hamza Zarfaoui

Master's

Su.Suganthi, Shanmuga Sundar.G, Felix Rainston.V.G, R. Hariharan · 6 authors

To revolutionize the educational scene, this project proposes a Web3 Token Incentive Mechanism, an Interactive Doubt Clarification Chatbot, and an advanced Intelligent Course Recommendation System. The system uses AI algorithms to analyze technical and non-technical input to recommend courses that suit students' interests and professional objectives. Using voice interaction for answering student queries and adding relevant video content will enhance accessibility and engagement. This AI-powered real-time support will be provided by the chatbot. A token-based incentive system, based on Web3, motivates students after they have completed courses and assessments. Once completed, students will be bridging the gap between education and work. They can use these tokens to avail free internships or get help in finding a job, thus adding utility to their educational process. This integrated strategy involves personalized learning, interactive participation, and incentive rewards to create a strong foundation for schooling focused on careers. AI, Web3 tokens, chatbots, personalized learning, career counseling, educational incentives, interactive meta learning, job assistance, and skill development are some of the keywords.

Pedro Baptista, Bernardo Pacheco, Filipe Apolinário, João Silveira · 6 authors

Having the ability to prove your knowledge is essential for obtaining a job. In the programming field, applicants make claims about the programming languages they master, and it is up to the interviewer to check the veracity of those claims. The goal of this work is to facilitate this process by extracting important information from GitHub, such as the number of bytes programmed in each programming language. The user is then able to ask for a Zero Knowledge Proof which can be downloaded and sent to any entity which places trust on out platform. The proof is verifiable without the entity interacting with our platform and does not leak information about the users' GitHub. The obtained results are promising, even though proofs can take several minutes to generate, they can be verified in many devices, such as laptops and smartphones, which greatly increases the number of users who can use our platform.

D. Chandravathi, M. Swetha, S.K. Khaja Shareef, Apathi Haripriya · 6 authors

Mass adoption of online learning and remote assessment has created significant challenges to maintaining exam integrity, verifying examinees, and ensuring veracity in scores. Existing solutions are frequently not based on consistent and robust validation mechanisms, and it is easy to commit impersonation, cheating and manipulation of the results. This research proposes deployment of a blockchain-based trust architecture to improve security in online examination systems by using smart validation. The primary objective is to create an all-inclusive system that ensures the veracity of exam takers and the validity of exam records when integrating biometric, behavioral, and the use of smart contract automation. We created a dataset of 50 scenarios from exams that included real world elements (face match confidence, keystroke patters, gaze tracking difference, device information and time logs). Deep learning algorithms were used on the dataset for session classification while dynamic validation criteria were applied by smart contracts and a private Ethereum blockchain for validated results maintained. Four main methods were applied to review the proposed model: rule-based logic, random forest, logistic regression and support vector machine (SVM) models. The outcome showed that the proposed technique gave an accuracy of 92%, an F1-score of 0.92 and a ROC-AUC of 0.95 which is lower than the other approaches. By examining confusion matrices and performing statistical tests, it was proved that the suggested model is both robust and generalizable and it has an unbelievably low p-value of 1.95 × 10−20in comparison with the weakest baseline. This research offers a scalable framework for handling e-proctored high stakes assessments, which is secure and auditable to overcome the current limitations and advances reliable digital educational platforms.

Mohammad M Maheri, Alex Davidson, Hamed Haddadi

Machine learning providers commonly distribute global models to edge devices, which subsequently personalize these models using local data. However, issues such as copyright infringements, biases, or regulatory requirements may require the verifiable removal of certain data samples across all edge devices. Ensuring that edge devices correctly execute such unlearning operations is critical to maintaining integrity. In this work, we introduce a verification framework leveraging zero-knowledge proofs, specifically zk-SNARKs, to confirm data unlearning on personalized edge-device models without compromising privacy. We have developed algorithms explicitly designed to facilitate unlearning operations that are compatible with efficient zk-SNARK proof generation, ensuring minimal computational and memory overhead suitable for constrained edge environments. Furthermore, our approach carefully preserves personalized enhancements on edge devices, maintaining model performance post-unlearning. Our results affirm the practicality and effectiveness of this verification framework, demonstrating verifiable unlearning with minimal degradation in personalization-induced performance improvements. Our methodology ensures verifiable, privacy-preserving, and effective machine unlearning across edge devices.

Habeeb Yusuf, Arthur Money, Damon Daylamani-Zad

Abstract The ever-changing global educational landscape, coupled with the advancement of Web3, is seeing rapid changes in the ways pedagogical artificially intelligent conversational agents are being developed and used to advance teaching and learning in higher education. Given the rapidly evolving research landscape, there is a need to establish what the current state of the art is in terms of the pedagogical applications and technological functions of these conversational agents and to identify the key existing research gaps, and future research directions, in the field. A literature survey of the state of the art of pedagogical AI conversational agents in higher education was conducted. The resulting literature sample (n = 92) was analysed using thematic template analysis, the results of which were used to develop a conceptual framework of pedagogical conversational agents in higher education. Furthermore, a survey of the state of the art was then presented as a function of the framework. The conceptual framework proposes that pedagogical AI conversational agents can primarily be considered in terms of their pedagogical applications and their pedagogical purposes , which include pastoral , instructional and cognitive , and are further considered in terms of mode of study and intent . The technological functions of the agents are also considered in terms of embodiment (embodied/disembodied) and functional type and features . This research proposes that there are numerous opportunities for future research, such as, the use of conversational agents for enhancing assessment, reflective practice and to support more effective administration and management practice. In terms of technological functions, future research would benefit from focusing on enhancing the level of personalisation and media richness of interaction that can be achieved by AI conversational agents.

Yi Zhang, Ziying Fan

This research paper explores the intersection of zero-knowledge proofs (ZKPs) and machine learning (ML), presenting a comprehensive overview of recent advancements, applications, and challenges in this fast growing area. The jointers of ZKPs and ML techniques shall go a meter further to fuse privacy, security, and integrity in a number of solutions, which include forming of groups for data sharing and safe machine learning. Through the investigation of the well-respected sites in that area and also the thorough description of formulas and their experimental outcome, this paper looks for the clarification of the current state of affairs and the possible future directions of ZKPs in the AI world. By inserting the verification mechanism of ZKPs into machine learning ecosystem, it allows devising novel solutions for the problems of privacy and confidentiality that have for long been not solved. With this approach, the concatenation of parties collectively performs the process of dealing with private inputs without revealing any of these data and this, in return, opens the possibilities of secure multi-party computation. Furthermore, ZKPs protect data sharing as it gives people the opportunity to construct confidential data and share them to model training without compromising any one’s private details. Being a part of the dynamic conversations, which focus on the game-changing capacity of transparent zero-knowledge proofs (ZKPs), this paper brings the role of ZKPs in preserving the confidentiality and integrity of artificial intelligence (AI) applications into the centre of attention. As scientists still fight to improve protocols and circumvent computational complications, ZKPs are likely to establishment as critical tools in the effort to increase ML systems in the digital sphere.

Wassim Alsindi

No abstract is available for this record.

Yannick Bourrier, Francis Jambon, Catherine Garbay, Vanda Luengo

No abstract is available for this record.

Peggy Eastman

FigureWASHINGTON, DC—The ambitious continuous learning database project of the American Society of Clinical Oncology known as CancerLinQ (OT, 8/25/12) has demonstrated its feasibility for the first time, according to speakers at a news briefing at the National Press Club here. The new prototype, demonstrated for briefing attendees on a hypothetical post-surgical patient with hormone-responsive breast cancer, included anonymous data from 100,000 breast cancer patients treated at U.S. cancer care sites. CancerLinQ is not the only cancer continuous learning database—Georgetown University has pioneered a similar project (see box). The prototype CancerLinQ, which makes available to oncologists via computer massive amounts of data to inform clinical decision-making and improve the quality of cancer care, has now demonstrated through a real-time testing process that it can work in actual practice, said ASCO President Sandra M. Swain, MD, Medical Director of the Washington Cancer Institute at MedStar Washington Hospital Center. Swain noted that the majority of oncologists, about 60 percent, are currently using electronic health records (EHRs), a necessity for CancerLinQ. Swain explained that when ASCO embarked on this multi-stage project about a year and a half ago—which she described as “very bold” and “scary”—it was with the continuous learning vision of the Institute of Medicine (IOM) in mind. “Our work is really grounded in the work of the IOM over the last few years,” she said.Figure: ASCO President-Elect CIFFORD HUDIS, MD, noted that one key benefit of the new prototype is that it can accept data from different electronic health records: “The system is independent of the EHR that the physician is using. We will work with anyone; we hope all vendors will end up with transformable data.”The vision, as set forth in a number of IOM reports, seeks to help clinicians both learn from and contribute to diagnostic and treatment data through a health information technology (HIT) computerized database containing electronic health records (EHRs). Now, she said, “the physicians are just clamoring to give us the data,” because they realize its importance in making informed clinical decisions. She said use of the large data set should help to counter the fragmentation in cancer care that makes it very difficult to draw insights from the collective clinical experience with cancer patients. “It means having the whole medical community available for an opinion. It confirms that every cancer patient can be an information donor.” The database makes available a vast amount of valuable patient data that cannot now be mined because it is hidden away—since only about three percent of adult cancer patients participate in clinical trials. “The worst situation is not having information,” Swain continued. “Every time I see a patient, there are one or two things that make that patient different. This helps us to get more answers.” She said it isn't just oncology that will benefit, but that the data gathered will likely be relevant to other diseases as well. ‘Proof-of-Principle Prototype’ “This is a proof-of-principle prototype,” said ASCO President-Elect Clifford A. Hudis, MD, Chief of the Breast Cancer Medicine Service and Attending Physician at Memorial Sloan-Kettering Cancer Center and Professor of Medicine at Weill Medical College. “It's a real-time, push-of-the-button load of the patient data upfront.” Hudis said much work on the prototype remains, and that over the next year “we're going to write white papers on what we've learned.” He noted that right now ASCO's Quality Oncology Practice Initiative (QOPI), is paper-based—an initiative that could become much more streamlined and efficient if CancerLinQ is eventually widely adopted. One key benefit of the new prototype is that it can accept data from different EHRs, he said. “The system is independent of the EHR that the physician is using. We will work with anyone; we hope all vendors will end up with transformable data.”Figure: ASCO President SANDRA M. SWAIN, MD, said use of the large data set should help to counter the fragmentation in cancer care that makes it very difficult to draw insights from the collective clinical experience with cancer patients.In the hypothetical breast cancer case demonstrated, the patient is put on an aromatase inhibitor but develops arthralgia. The CancerLinQ database prototype tells her physician to consider using tamoxifen as an alternative, and provides supporting data for that treatment choice. “For 25 years I've been doing one-on-one medicine,” said another speaker, W. Charles Penley, MD, a partner with Tennessee Oncology, PLLC, Board Chair of the Conquer Cancer Foundation, and a member of the Dean's Advisory Board of the College of Arts and Sciences at the University of Tennessee. “Patients have been telling me, ‘Doctor, I want you to learn from my case to help other patients.’ This [CancerLinQ] is that taken to the modern information age.” Penley, who is one of about 25 clinicians in the network testing the ASCO database prototype and whose practice contributed breast cancer patient data to it, added, “This tool really can be a game changer in that regard.” What it means for cancer patients, he said, is that they can have confidence that they are receiving the highest quality care no matter where they are located. The database prototype, which he called “a remarkable step forward,” offers “an opportunity to query not just a few experts known to us, but the collective experience of treating clinicians—thus adding “second opinions times multiples.” Lessons from Pediatric Oncology Lynn M. Etheredge, who leads the Rapid Learning Project at George Washington University, said lessons from pediatric oncology can be valuable for CancerLinQ as it moves forward. Pediatric oncologists built a system to capture data from every patient as if he or she were on a clinical trial and then learn from that experience, noted Etheredge, who worked for the White House Office of Management and Budget in the Carter and Reagan Administrations, and who proposed the concept of the “rapid learning health system” in a special issue of Health Affairs in 2007 (26: w107-w118). “Pediatric oncologists realized early on that there were genetic differences,” he said. “Hopefully we will have the same success in treating adult patients.” Asked by OT if he could have envisioned his concept of a rapid learning health system coming to this database prototype point, Etheredge said, “I'm an optimist,” but noted that “This is astonishing.” He said that in the past physician groups have largely been reactive—responding to “things done to them,” and he praised ASCO for being proactive, innovative, and forward-thinking. “What we are saying now is that we have put the stake in the ground; we have demonstrated everything we wanted to demonstrate,” Joshua Mann, ASCO's Associate Director for Oncology Technology Solutions, Quality and Guidelines, said in an interview. “Now we're ready to engage the broader audience.” For the full CancerLinQ system, “we plan to siphon off data feeds from anyone,” including small oncology practices, not just large cancer centers. The message is: “Send us whatever you have however you can.” He noted that “machine-learning algorithms” convert data into a standardized format, thus allowing practices using different EHRs to participate in the continuous learning database. Lombardi's G-DOC Integrates New Knowledge with Practice At Georgetown University's Lombardi Comprehensive Cancer Center, Director Louis M. Weiner, MD, has pioneered a continuous learning system similar to CancerLinQ called Georgetown Database of Cancer, known as G-DOC. This system uses both local data and publicly available data sets to put the concept of personalized medicine into practice, Weiner explained. Commenting on ASCO's CancerLinQ prototype proof-of-principle, Weiner—a member of the Board of Scientific Advisors of the National Cancer Institute—said, “CancerLinQ is very ambitious. Currently, cancer specialists have access to only limited data to help them make critical life-altering decisions for their patients. In particular, it is very difficult to knowledgeably personalize therapies based upon a person's particular circumstances that are dictated by their genetics, comorbidities, and molecular properties of the cancers that afflict them. G-DOC has been designed as a first step towards that goal.”FigureWeiner noted that while there are patient confidentiality issues that need to be overcome in drawing on large databases to make treatment decisions, the concept is sound. It is clear that “it would be logical and desirable to link multiple datasets and then to create physician- and patient-friendly user interfaces that allow for shared decision-making that is based on a nuanced understanding of who to treat, what to use, and when to use it.”