Institute of Information Science

We propose a minimal physical model for the Nakamoto distributed consensus protocol based on non-equilibrium statistical mechanics. We treat the ledger as a one-dimensional lattice system where the consensus state is determined by the minimization of a thermodynamic cost function, analogous to the free energy in spin systems. In this framework, the ”Double Spend” problem is identified as a local symmetry breaking of the time-ordering parameter. We demonstrate that Proof-of-Work (PoW) acts as a dissipative external field that drives the system from a disordered ”liquid” phase (unconfirmed transactions) to an ordered ”crystalline” phase (immutable history). By defining an effective temperature derived from network latency and hashrate, we analyze the probabilistic finality of the ledger not as an event horizon, but as a correlationlength decay characteristic of massive field theories. Finally, we interpret chain forks as topological defects (domain walls) and show that the ”Halving” event acts as a sudden quench, subjecting the network to critical slowing down consistent with the Kibble-Zurek mechanism.

Due to the realm of the blooming generative models, like GANs, VAE and diffusion-based environment, the trustworthiness of authentication in visual media has been vastly challenged. AI-generated images appearing quite photorealistic go beyond human perceptual boundaries often overlying concerns for disinformation, digital tampering, forensic evidence authentication, and or the security of biometric purposes. This review organizes throughout the chronology of deep learning advancement to segregate real from synthetic images by focusing not much on its generalization snags, extraction of fake image or artifacts from sustainability in the aspect of datasets. Also this review provides systematic analysis on publicly available datasets and those essential research directions in the future being necessary to become a robust detection mechanism for fake images.

This book examines the remarkably preserved Soča/Isonzo Front battlefield, exploring how its material heritage has shaped World War I (WW1) remembrance across changing political regimes along the Italian-Slovenian border. It introduces the reader to how conflict landscapes function as independent agents of remembrance, transcending political changes and national narratives. The book offers a multidisciplinary approach that combines anthropology, archaeology and memory studies to trace the ‘social life’ of this lesser-known WW1 battlefield. By examining material heritage through multiple lenses, it reveals how physical remnants continue to shape collective memory and cultural understanding of the conflict, providing fresh perspectives on war commemoration beyond traditional historical accounts. This book is written for scholars and students in the fields of memory studies, military history, heritage preservation and modern conflict archaeology. It will appeal to researchers interested in WW1, particularly those focusing on the South-Eastern European theatre and the Italian Front. Libraries supporting programs in anthropology, European history, cultural heritage studies and war studies will find this an essential addition to their collections. The book is also suitable for heritage professionals working with battlefield preservation and commemoration.

Veliki jezikovni modeli omogočajo, da običajnim aplikacijam dodamo funkcionalnosti umetne inteligence. Pametni asistenti z uporabo orodij omogočajo, da veliki jezikovni model preko povratnega klica pokliče poslovno metodo v aplikaciji. V javanskem okolju je najbolj priljubljena knjižnica Langchain4j. S pomočjo Langchain4j lahko uporabimo različne implementacije velikih jezikovnih modelov (OpenAI, Ollama, JLama ...), ne da bi nam bilo treba spreminjati programsko kodo. V prispevku bomo podrobneje predstavili knjižnico Langchain4j in Quarkusovo razširitev Quarkus Langchain4j, ki uporabo knjižnice Langchain4j še bolj poenostavi, saj omogoča deklarativno uporabo knjižnice večinoma le z uporabo anotacij. Tako lahko hitro razvijamo nove funkcionalnosti. Podobni pristop uporablja Spring AI v aplikacijah SpringBoot. Pametnega asistenta COBISS, ki smo ga predstavili lani, smo nadgradili z uporabo knjižnice Langchain4j. Tako je precej ponavljajoče se kode odpadlo. Na kratko bomo predstavili tudi ogrodje Quarkus, ki omogoča zelo hiter zagon aplikacij in je konkurenčno ogrodjema Spring in Jakarta EE.

On April 7 2025, Zeeya Merali wrote an article for the Foundational Questions Institute (https://qspace.fqxi.org/news/165296/gerard-t-hooft-and-13508-lhc-physicists-are-among-this-years-us$3million-breakthrough-prize-winners) which announced a special Breakthrough Prize in fundamental physics was awarded to Nobel laureate Gerard 't Hooft, one of the principal architects of the Standard Model of particle physics. Alongside this, 13,508 physicists at CERN were honored for their collaborative efforts at the Large Hadron Collider, where they’ve rigorously tested the Standard Model. The holographic principle, which states that information within a 3-D region of space can be encoded on its 2-D boundary, also originated with 't Hooft – although today it is not one of his favorite insights. People run away doing mysterious things with it, but they don't really build theories that I can understand," says 't Hooft. He also says - “when someone of the older generation tells you something cannot be done, don't believe them. You have to think harder than those old guys did, and that's always how breakthroughs get done.” I’d like to do my best to “think harder” and write a few brief comments on points mentioned in the article. They’re taken from my article in the science journal “IPI Letters” called “A Different Perspective on Cosmological and Quantum Phenomena That Involves the Temporal Multiverse and the Static Universe” (with some extra thoughts included). (1) Also, a few paragraphs have been added which are not from that article (or any other). They speak of the contributions of James Clerk Maxwell’s equations, the Aharonov-Bohm effect, Paul Dirac, and Akira Tonomura to this article’s conclusion that the universe is the magnetic monopole physicists seek.

This paper explores the nature of threats posed by Distributed Denial of Service (DDoS) attacks on large networks, such as the Internet, emphasizing the need for effective detection and response mechanisms. These mechanisms must be implemented not only at the network edge but also within its core. The paper introduces methods to detect DDoS attacks by analyzing entropy and frequency-sorted distributions of specific packet attributes. Anomalies in these attributes' characteristics serve as indicators of potential DDoS attacks. The proposed methods are evaluated for detection accuracy and performance using live traffic traces collected from diverse network environments, including core Internet nodes and edge networks. Results demonstrate the effectiveness of these methods against current DDoS attacks and provide insights into improving detection capabilities for more sophisticated, stealthier threats. Additionally, the paper describes a detection-response prototype and discusses how the detection system can be extended to support effective response decision-making.

Abstract Containers in High Performance Computing (HPC) provide a flexible and efficient alternative to traditional modular software. They allow researchers and engineers to package applications and their dependencies in a single file, ensuring portability, reproducibility and consistent performance across different computing environments. With the emergence of AI, industrial users and SMEs in HPC, containers have become even more convenient to use, as they can also meet the security requirements for isolating user workloads. As the HPC community embraces container technology, it is ushering in a new era of more efficient, repeatable and scalable research computing. However, adoption of containers in HPC has been relatively slow, largely due to misconceptions about how they work and concerns about potential performance degradation, energy inefficiency and security risks. This article looks at the security considerations when using containers and uses practical examples to show that, with the right configuration and tools, containers can be as secure as any other system application.

This presentation walks through our work on creating a non-nanosecond datetime for Pandas and the development of Python toolboxes to query databases and analyze datasets using Pandas objects for interoperability.

Online manipulation is a pressing concern for democracies, but the actions and strategies of coordinated inauthentic accounts, which have been used to interfere in elections, are not well understood. We analyze a five million-tweet multilingual dataset related to the 2017 French presidential election, when a major information campaign led by Russia called "#MacronLeaks" took place. We utilize heuristics to identify coordinated inauthentic accounts and detect attitudes, concerns and emotions within their tweets, collectively known as socio-linguistic characteristics. We find that coordinated accounts retweet other coordinated accounts far more than expected by chance, while being exceptionally active just before the second round of voting. Concurrently, socio-linguistic characteristics reveal that coordinated accounts share tweets promoting a candidate at three times the rate of non-coordinated accounts. Coordinated account tactics also varied in time to reflect news events and rounds of voting. Our analysis highlights the utility of socio-linguistic characteristics to inform researchers about tactics of coordinated accounts and how these may feed into online social manipulation.

The advent of powerful transformer-based discriminative language models and, more recently, generative GPT-family models, has led to notable advancements in natural language processing (NLP), particularly in commonsense reasoning tasks. One such task is commonsense reasoning, where performance is usually evaluated through multiple-choice question-answering benchmarks. Till date, many such benchmarks have been proposed and `leaderboards' tracking state-of-the-art performance on those benchmarks suggest that transformer-based models are approaching human-like performance. However, due to documented problems such as hallucination and bias, the research focus is shifting from merely quantifying accuracy on the task to an in-depth, context-sensitive probing of LLMs' generalization and robustness. To gain deeper insight into diagnosing these models' performance in commonsense reasoning scenarios, this thesis addresses three main studies: the generalization ability of transformer-based language models on commonsense reasoning, the trend in confidence distribution of these language models confronted with ambiguous inference tasks, and a proposed risk-centric evaluation framework for both discriminative and generative language models.