Svetlana Boudko

Federated learning enables collaborative computation across multiple decentralized devices, minimizing data transfer overhead while enhancing privacy by keeping data local. However, it remains susceptible to inference attacks and potential data leakage. To strengthen privacy guarantees, especially for sensitive domains, advanced privacy-preserving techniques such as homomorphic encryption are recommended. This work proposes a privacy-preserving federated learning framework that integrates threshold homomorphic encryption into the federated learning pipeline to enable secure aggregation and protect intermediate computations. We employ threshold homomorphic encryption, a cryptographic technique well-suited for multiuser environments such as federated learning. We utilize the Cheon-Kim-Kim-Song (CKKS) scheme, as implemented in the OpenFHE library. Our approach extends the standard Federated Averaging (FedAvg) algorithm by homomorphically encrypting model updates and performing aggregation directly on encrypted data. To assess the trade-offs between efficiency and security, we evaluate the performance of the proposed method against a baseline. The design prioritizes practical constraints, including computational efficiency, making it suitable for deployment in privacy-sensitive domains such as healthcare and finance. To ensure compatibility with continuous integration and deployment (CI/CD) pipelines, all components of the solution are containerized using Docker.

Historisk befolkningsregister (HBR) er et autoritetsregister over alle personene i norsk personhistorisk kildemateriale tilbake til 1801 og gjør det mulig å identifisere og gjenfinne personene i disse kildene. Hver person får en unik ID som er viktig for dokumentasjon og for å finne mer informasjon om vedkommende. HBR utvider perioden med detaljert kunnskap om hver person i den norske befolkningen fra 60 år i dagens folkeregister til 224 år, fra to til syv generasjoner. I denne artikkelen beskriver vi oppbyggingen av HBR med lenkingsstrategier samt personvern og hvordan det avviker fra andre befolkningsoversikter. Videre blir det drøftet hvordan man kan oppnå best mulig kvalitet i registeret, samt representativitet og hvor høy lenkingsgrad det er mulig å oppnå. Kunnskap om hver enkelt persons livsløp, bosted og familie gir en ny innsikt i befolkningen og åpner for nye forskningsmetoder og mer presise beskrivelser og analyser. Dette vises ved å gi et datagrunnlag for migrasjon 1910–1920 og en analyse av fødselsdatoer i folketellinger. Forskere vil kunne trekke ut data om de problemstillinger de interesserer seg for, som grunnlag for sin egen forskning.

Federated Learning is a machine learning approach where a model is trained across multiple decentralized edge devices. Since the data are not uploaded to a server, this approach is particularly useful for data protection and efficient computation. Further, it can be combined with privacy-preserving technologies, e.g., homomorphic encryption for enhanced data protection. Considering all these elements, a practical solution will require an efficient multikey homomorphic encryption as well as an effective integration of federated model aggregation and multikey generation processes. The paper studies the related work for homomorphic encryption in the context of federated learning and outlines the rationale behind practical design of secure federated learning.

Methods for untraceable and anonymous communication, such as anonymous routing networks and dining cryptographers networks, are in general very complex and suffer from high performance overhead of a minimum order of encryptions for N participants. In this paper, we propose an original approach to untraceable communication that avoids some of the significant shortcomings of existing methods. Using non-interactive privacy-preserving aggregation as an underlying building block we achieve attractive features, including unsurpassed low computational and transmission overhead of only 3 encryptions per participant in only a single round.

The proliferation of healthcare data has expanded opportunities for collaborative research, yet stringent privacy regulations hinder pooling sensitive patient records. We propose a multiparty homomorphic encryption-based framework for privacy-preserving federated Kaplan--Meier survival analysis, offering native floating-point support, a theoretical model, and explicit reconstruction-attack mitigation. Compared to prior work, our framework ensures encrypted federated survival estimates closely match centralized outcomes, supported by formal utility-loss bounds that demonstrate convergence as aggregation and decryption noise diminish. Extensive experiments on the NCCTG Lung Cancer and synthetic Breast Cancer datasets confirm low mean absolute error (MAE) and root mean squared error (RMSE), indicating negligible deviations between encrypted and non-encrypted survival curves. Log-rank and numerical accuracy tests reveal no significant difference between federated encrypted and non-encrypted analyses, preserving statistical validity. A reconstruction-attack evaluation shows smaller federations (2--3 providers) with overlapping data between the institutions are vulnerable, a challenge mitigated by multiparty encryption. Larger federations (5--50 sites) degrade reconstruction accuracy further, with encryption improving confidentiality. Despite an 8--19× computational overhead, threshold-based homomorphic encryption is feasible for moderate-scale deployments, balancing security and runtime. By providing robust privacy guarantees alongside high-fidelity survival estimates, our framework advances the state-of-the art in secure multi-institutional survival analysis.

Distributed analytics, such as federated learning, involve collaborative computation across multiple decentralized devices. This approach not only reduces data transfer costs but also offers some degree of protection for privacy-sensitive information. To achieve a higher level of privacy protection, it is recommended to use more advanced privacy-preserving technologies, such as homomorphic encryption. However, the use of holomorphic encryption schemes results in high computational costs. In this study, we evaluate the performance characteristics of threshold fully homomorphic encryption, a technique that can be effectively applied in multi-user environments and distributed analytics scenarios. We present results from the performance evaluation of the Cheon-Kim-Kim-Song scheme.

Healthcare 4.0 is a new paradigm for providing healthcare services in highly distributed and complex settings. The distributed and heterogeneous nature of home-based medical devices, and their need to exchange data with external sources make Healthcare 4.0 solutions susceptible to cyberattacks and require decentralized solutions to protect sensitive local data. This work presents a collaborative approach to security incident detection for distributed healthcare utilizing federated learning. At this stage, the federated learning process has been facilitated and evaluated using simulation, training, and testing.

The healthcare market demands flexible, and secure solutions for personal health data sharing. For this purpose, we developed a distributed multi-agent system that reinforces a negotiable right management environment. We propose a bilateral bargaining mechanism that facilitates computational negotiation between patients and healthcare stakeholders. The mechanism is evaluated using both open and sealed bids. We show that agents reach an agreement within a feasible number of rounds, even if their bids are kept confidential.

The necessity for anonymous and untraceable communication exists in numerous ICT applications and location-based services that require user location data for their operations. Disclosing the locations and positions where people are situated, in some cases over long time intervals or even on a continuous basis, is by many considered invasive and infringement of personal privacy. Privacy-preserving methods are a promising means to protect user privacy in many applications. Untraceability is central privacy property in which the trace (or link) between an object as it moves from one location to another cannot be certainly established. For a person using a GPS-based application, untraceability implies that links between a person and that person’s changing locations cannot be certainly established. We discuss in the report the state of the art of untraceable anonymous communication methods. Further, we assess the research and market opportunities, and present ideas for future projects and proposals.

Multi-microgrids (MMGs) provide economic and environmental benefits to society by improving operational flexibility, stability and reliability of a smart grid. MMGs have greater complexity than conventional power networks due to the use of multiple infrastructures, communication protocols, controllers, and intelligent electronic devices. The distributed and heterogeneous connectivity technologies of the MMGs and their need to exchange information with external sources as well as the vulnerabilities in the communication networks and software-based components, make MMGs susceptible to cyberattacks. In this work, we present a conceptual framework for collaborative adaptive cybersecurity that is able to proactively detect security incidents. The framework utilizes federated learning for collaborative training of shared prediction models in a decentralized manner. The methodology used in this research is mainly analytical. This involves analysis of how the principles of a collaborative adaptive cybersecurity can be applied to the MMG environments resulting in the development of theoretical models which can then be validated in practice by prototyping and using real time simulation.

The blockchain and Peer-To-Peer Payment solutions become adopted by financial institutions. While these changes bring significant service benefits they also increase the risks and vulnerabilities of the financial services. In this paper, we investigate, develop, and evaluate machine learning (ML) algorithms for predicting attacks on blockchain nodes and a Peer to Peer payment system. We have evaluated a set of machine learning algorithms that include classifica-tion ML algorithms from the scikit-learn library. We demonstrate that the pro-posed solution is able to predict cyber-physical attacks close to 100% accuracy. We have implemented a service prototype as a proof of concept. The prediction is done based on the collected data of the blockchain and peer-to-peer payment nodes. For the evaluation of the algorithms, a set of highly reputable classifica-tion metrics has been selected and applied.

The report describes the research and development of adaptive security addressing the protection of "IoT-based smart grids" against evolutionary threats and attacks through the prediction and advanced behavioural analysis of big data from IoT Smart Grids by automating prevention, detection, and recovery from the failures of security and privacy protections at run-time and by re-configuring control parameters and security goals.

This document describes the work performed in the project GB-DIGITALISERING 2020 Covid19, which is funded by the Research Council of Norway. The project identified technological barriers and obstacles for digitalisation in today’s society and provided ideas on how to mitigate these. The use of AR/VR/MR technologies, social robotics, and wide use of data from sensors and devices in the IoT in diverse application areas are important technologies to make migration from analogue to digital solutions possible. The project results will be used to support the industry and the public sector to mitigate barriers related to properties of usability, universal design, privacy, security, personalization of services, user engagement, and others.

We applied evolutionary game theory to extend a resource constrained security game model for confidentiality attacks and defenses in an Advanced Metering Infrastructure (AMI), which is a component of IoT-enabled Smart Grids. The AMI is modeled as a tree structure where each node aggregates the information of its children before encrypting it and passing it on to its parent. As a part of the model, we developed a discretization scheme for solving the replicator equations. The aim of this work was to explore the space of possible behaviors of attackers and to develop a framework where the AMI nodes adaptively select the most profitable strategies. Using this model, we simulated the evolution of a population of attackers and defenders on various cases resembling the real life implementation of AMI. We discuss in depth how to enhance security in AMI using evolutionary game theory either by a priori analysis or as a tool to run dynamic and adaptive infrastructure defense.

The report summarizes the three journal papers which address security in android smartphone, confidentiality in IoT-enabled smart grids, and fault-tolerant privacy-preserving in AMS, respectively.

Historisk befolkningsregister, HBR, er et register over den norske befolkningen fra 1801 frem til Det sentrale folkeregisteret fra 1964. Det lages ved å lenke sammen personforekomster av samme person i folketellinger og kirkebøker og koble sammen nære familiemedlemmer. Vi viser hvordan algoritmer brukes til å finne lenker og koblinger. Vi har funnet familierelasjoner i alle de nominative folketellingene, også der familiestilling ikke er beskrevet i folketellingen. Alle algoritmene baserer seg på likhet i navn, fødselsår, fødested, yrke, bosted og tilsvarende egenskaper hos familiemedlemmer. I tillegg sjekker vi at alle hendelsene gir et naturlig livsløp. Algoritmene må tilpasses de ulike kildene for å holde feilraten under en teoretisk målsetning på 1 % – noen fasit fins imidlertid ikke. Vi må for eksempel ha strengere kriterier for likhet i store kommuner som Oslo, enn i mindre kommuner og strengere i 1865-folketellingen enn i den mer nøyaktige 1891-folketellingen. Vi har funnet samme person i etterfølgende folketellinger på nasjonalt nivå ved å bruke sjeldne etternavn. Vi har også søkt etter lignende personforekomster i samme kommune og fylke i Digitalarkivet for alle personer i en av de nasjonale folketellingene. Det er laget maskinelle lenker ved Registeringssentral for historiske data, UiT og Norsk Regnesentral. Ved nettstedet histreg.no er det også mulig å lage manuelle lenker. Vi viser utviklingen av bidragsytere, lenker og koblinger i databasen i den fireårige perioden til registeret.

This chapter presents the FINSEC adaptive and intelligent data collection and analytics system for securing critical financial infrastructure. It enhances the intelligent, resilient, automated, efficient, secure, and timely manner the collection and analysis of security-related data for securing cyber-physical financial infrastructure and services. Making security data collection and analysis intelligent and capable of quickly spotting, learning from, and addressing zero-day threats is essential to economizing of resources and accessing the right information at the right time. This is achieved through the configuration of configurable collection probes and the adaptation of different collection strategies. The chapter further addresses how, inter alia, (i) the nature and quality of collected data affects the efficiency and accuracy of methods of attack detection and defense, (ii) the detection capability can be improved by correlating wide-ranging data sources and predictive analytics, (iii) the rate of the data collection at the various monitoring probes is tuned by managing the appropriate levels and types of intelligence and adaptability of security monitoring, (iv) the optimization of bandwidth and storage of security information can be achieved by rendering adaptiveness and intelligence and by integrating smart security probes and a set of adaptive strategies and rules, and (v) the increased automation is achieved through a feedback loop of collection, detection, and prevention that allows the early detection and prevention of security compromises and consistently makes security analysis more effective.

Connecting people, processes, devices and data, the Internet of Things brings new security challenges and may significantly increase the vulnerability of healthcare services. This paper investigates advanced adaptive security to anticipate and respond to dynamic and adaptive attacks on healthcare critical infrastructures. We propose the Adaptive Cybersecurity Framework that supports dynamic adaptation to cyber threats. Further, we simulate and evaluate the framework using evolutionary game theory, and outline the further steps for our future work.

The healthcare market demands advanced, flexible, and secure solutions for personal health data sharing. In our paper, we present preliminary work that proposes a distributed infrastructure of negotiating agents for the healthcare domain. This infrastructure will support healthcare stakeholders to share and access patient health data in a secure way, thus providing benefits for patients and their treatment. Distributed ledger technologies and smart contracts can be considered as a basis for negotiations between distributed agents that carry health-related data. We present an overview of related work and outline the research methodology.

This article describes the development of the Norwegian Historical Population Register, which is the first open national register. In the period 1735–1964, 9.7 million people lived in Norway, and for them 37.5 million events (such as birth, death, or migration) have been recorded in sources. We link together as many events as possible for the same persons and families, but only include links that have a high probability of being correct. The linking is performed by automatic methods and crowdsourcing. A national population register is important for migration research. It allows us to reconstruct (stepwise) internal migration in Norway, frequently followed by international migration from Norway, as well as return migration to Norway. Many non-Norwegian sources also specify place of birth by country, and this makes it possible to identify individuals in Norwegian sources.

This research note is Deliverable D4.5 of the Eurostars project EHRPathways E9080! and constitutes the final report of WP4 of this project. It describes the design and development of the workflow support for Electronic Health Record Systems that are used in a nursing context, as well as the outcome of WP4 of the project. This document also contains the documents for the Deliverables D4.3 and D4.4 of the EHRPathways project.

The healthcare market demands more advanced and flexible solutions to provide better care for patients. Moving medical services closer to patients improves the quality of healthcare but brings, at the same time, new challenges. It requires a well-functioning, decentralized infrastructure and better workflow support for health personnel. Therefore, we need to provide consistency for this type of services with patient data accessed and processed by several healthcare stakeholders. In the paper, we present how Petri net-based workflow technology can be used to design flexible treatment pathways, and to coordinate the treatment process in a distributed way among several healthcare workers. We have implemented a system that supports consistency for distributed healthcare and integrates the treatment workflows into a healthcare workflow system. We use a clinical health care example to illustrate the workflow.

This paper describes the building of a Norwegian historical population register that is under construction. It is based on linking together person entries from a large number of sources by linking algorithms and crowd sourcing. There are 9.7 million residents of Norway in the period 1735 - 1964 and 37.5 million events in most relevant sources. We want to link together as many events as possible for the same persons and families but only include links that have a high probability to be correct. As the first open national population register, we believe that it will be particular important for migration studies. It is much easier to follow both national and international migration on a national level. Quite many of the international migration sources are national, and place of birth is specified by nationality for persons born abroad.

This document is the final report of the GB project GB-BOHEME-2014/2015. We present a brief list of the project achievements followed by a characterisation of the tree focus areas: open content lifecycle management, user experiences & gamification, and eHealth & welfare.We also present some thoughts on how to organise a follow-up project. Section 2 of this note presents an overview of the developed concepts and their applications areas.

Mobile devices are typically equipped with multiple access network interfaces, supporting the coexistence of heterogeneous wireless access networks. The selection of an optimal set of multiple serving mobile networks for multicast streams is NP-hard and is, therefore, a challenging problem. We propose a simple heuristic approach that provides configuration of multicast groups for a given network topology and network conditions.We consider that a forward error correction technique is applied to deal with packet loss of the wireless communication.

The thesis formulates effective strategies for delivery of streaming content for mobile users in resource-limited networks. We consider the entire end-to-end delivery path from the source to the destination. In the work, we take into account 1) the presence of a large number of mobile devices operating inside an area with overlapping coverage of several mobile networks; and 2) the ability of the mobile devices to exploit several network technologies and to be connected to different access points simultaneously. We aim to improve the performance of the whole system by jointly considering users’ requirements and network limitations and by identifying and implementing adaptation mechanisms for different parts of the delivery path. Taking into account the complexity of the whole system, the thesis aims to design mechanisms operating in a decentralized manner with partial knowledge of the system. For this purpose, we study the ADIMUS architecture which addresses the problem of distributing streaming content for mobile users. The ADIMUS architecture clearly separates the system into two distinct parts: the backbone part of the network and the wireless part. We study resource adaptation mechanisms for both of these parts that include 1) overlay routing for the backbone and 2) multi-access approaches for heterogeneous wireless network. For overlay routing in the backbone, we consider multipath streaming techniques. For the wireless part, we improve the utilization of resources by intelligently redistributing mobile devices among the available wireless connections. To evaluate the adaptation mechanisms for both of these parts, we build mathematical models for the overlay network and for the heterogeneous wireless network. These mathematical models are used as upper bounds for operation of the adaptation solutions. We also define the lower bounds to these solutions based on earlier proposed solutions from the literature. For the overlay network, we develop and evaluate a distributed rate-allocation algorithm for delivery of video in a Video-on-Demand (VoD) system built upon multipath delivery. The algorithm’s operation is based on information collected by overlay nodes. The algorithm is evaluated in the OMNet++ simulation environment and analyzed against the aforementioned lower and upper bounds. We show that the algorithm performs close to the optimal solution in terms of proportionally fair bandwidth allocation between receivers of the overlay part of the architecture. For the heterogeneous wireless network, we consider the network selection to use for multiuser environments with possible multicast configurations that allows the network to perform load balancing, improve the users’ overall QoS, and increase the throughput of the networks. The novelty of the proposed solutions is that the network selection is done in a decentralized manner with only limited information available to decision makers. The solutions are evaluated through multiple simulations. We show that the solutions provide a substantial improvement in performance compared to their lower bounds. We also investigate a multiple connection approach for multicast groups with mobile terminals connecting to several networks simultaneously and receiving data through multiple paths. We show that the total usage of resources is reduced compared to a single path approach.

Purpose – Coexistence of various wireless access networks and the ability of mobile terminals to switch between them make an optimal selection of serving networks for multicast groups a challenging problem. Since optimal network selection requires large dimensions of data to be collected from several network locations and sent between several network components, the scalability can easily become a bottleneck in large-scale systems. Therefore, reducing data exchange within heterogeneous wireless networks is important. The paper aims to discuss these issues. Design/methodology/approach – The authors study the decision-making process and the data that need to be sent between different network components. To analyze the operation of the wireless heterogeneous network, the authors built a mathematical model of the network. The objective is defined as a minimization of multicast streams in the system. To evaluate the heuristic solutions, the authors define the upper and lower bounds to their operation. Findings – The proposed heuristic solutions substantially reduce the usage of bandwidth in mobile networks and exchange of information between the network components. Originality/value – The authors proposed the approach that allows network selection in a decentralized manner with only limited information shared among the decision makers. The authors studied how different sets of information available to decision makers influenced the performance of the system. The work also investigates the usage of multiple paths for multicast in heterogeneous mobile environments.

The increasing demand for real-time multimedia streaming from mobile users makes important deployment of network selection in wireless networks. Coexistence of various wireless access networks and ability of mobile terminals to switch between them make an optimal selection of serving mobile networks for groups of mobile clients a challenging problem. Since scalability can easily become a bottleneck in large-scale networks, we study the decision-making process and selection of the data that needs to be exchanged between different network components. In this paper, we present two decentralized solutions to this problem that we compare and evaluate in the OMNet++ simulation environment.

Coexistence of various wireless access networks and the ability of mobile terminals to switch between them make an optimal selection of serving networks for multicast groups a challenging problem. Since optimal network selection requires large dimensions of data to be collected from several network locations and sent between several network components, the scalability can easily become a bottleneck in large scale systems. Therefore, reducing data exchange within heterogeneous wireless networks is important. We study the decision-making process and the data that needs to be sent between different network components. We present two decentralized solutions to this problem that operate with reduced sets of information. We define the upper and lower bounds to these solutions and evaluate them in the OMNet++ simulation environment. Both solutions provide a substantial improvement in performance compared to the lower bound.

This document is the final report of the GB project GB-SMS-2012/2013. We present a brief list of the project achievements followed by a characterisation of the focus areas of open content lifecycle management, ICT challenges within health and welfare, user experience research, and open source & open data. We give an overview of the developed concepts and their applications areas before presenting some thoughts on how to organise a follow-up project.

We consider a network selection problem for a group of mobile clients that operate in a heterogeneous wireless access network environment and are equipped with multiple access network interfaces. The involved networks cooperate in order to improve their own and the mobile clients’ performance. We formulate the problem as a team decision problem. In this formulation, several decision variables are involved, and these decisions are made by several decision makers with access to different information but contributing to a common goal. The novelty of the proposed approach is that the network selection is done in a decentralized manner with only limited information available to decision makers. We present two decentralized algorithms to this problem that we compare and evaluate in the OMNet++ simulation environment.

Mobile devices are typically equipped with multiple access network interfaces, supporting the coexistence of heterogeneous wireless access networks. The selection of an optimal set of serving mobile networks for multicast streams is a challenging problem. We consider a network selection problem for multicast groups of mobile clients that operate in a heterogeneous wireless access network environment. We present a solution to this problem with an optimal allocation of mobile users to multicast groups when multiple mobile networks are available for operation. This solution is suited for small scale networks and can be used as reference for complex networks.

Today mobile devices are typically equipped with multiple access network interfaces. Another important issue is a coexistence of heterogeneous wireless access networks. The selection of optimal serving mobile networks for multicast streams is a challenging problem.We consider a network selection problem for multicast groups of mobile clients that operate in a heterogeneous wireless access network environment. We identify several decision makers solving this problem and present our view on what kind of information is needed to be exchanged between these decision markers.

We develop and evaluate a rate-allocation multipath algorithm for delivery of video in a Video-on-Demand (VoD) system built upon an overlay network. To evaluate how well the rate-allocation algorithms approach the optimum, we have earlier built a benchmarking system that provides the optimal solution for assigning available bandwidth to delivery paths. This solution is also proportionally-fair. We implement a distributed algorithm for multipath rate allocation that operates based on information collected by overlay nodes. The algorithm is evaluated in the OMNet++ simulation environment and analyzed against the aforementioned benchmarking.

This document is the final scientific report of the ADIMUS project which was funded by the NORDUnet3 programme. ADIMUS addresses the problem of enhancing the quality of multimedia streams at run-time based on the perceived quality. We present an architecture that addresses the different requirements from end-to-end for a mobile terminal. The architecture comprises an overlay network in the long-distance part, while the multi-access network employs cross-layer technology. Both parts interact, and use adaptation techniques shown in this report. We also present quality estimation techniques for both audio and video streams. This report is based on the scientific papers published during the project, and sets these papers into the context of the ADIMUS project.