Publications

Introducing h-leading-ones as a mixed-category benchmark problem for evolutionary algorithms

Abstract

In the wake of generative artificial intelligence and the exponential growth in the volume of data generated, the associated increase in data complexity in the sense of the quantity of different datatypes present in a single system poses a challenge to evolutionary algorithms. To allow for the development and testing of new algorithms adapted to this new data landscape, test problems are necessary as a way to both evaluate and compare algorithms per-formances. However, while recent advances extended known test problems such as the r-Leading-Ones marking the transition from binary to multi-valued variables, having different data-types coexisting in the search space is still an open question. We propose the h-Leading-Ones as an extension of the r-Leading-Ones to evaluate the ability of an algorithm to solve problems on a search space composed of multi-valued and real-valued data types. Its design with dependency between the different data-types and its continuity with the r-Leading-Ones provides a convenient new environment for benchmark and runtime analysis for mixed-category searchspaces.

Continue reading

Investigation of metabelian platform groups for protocols based on (simultaneous) conjugacy search problem

Abstract

here are many group-based cryptosystems in which the security is related to the conjugacy search problem or the simultaneous conjugacy search problem in their underlying platform groups. In this article, we show that some metabelian groups do not provide strong security for these cryptosystems and so they cannot be chosen as platform groups..

Continue reading

Security analysis of ZKPoK based on MQ problem in the multi-instance setting

Abstract

Bidoux and Gaborit introduced a new general technique to improve zero-knowledge (ZK) proof-of-knowledge (PoK) schemes for a large set of well-known post-quantum hard computational problems such as the syndrome decoding, the permuted kernel, the rank syndrome decoding, and the multivariate quadratic (MQ) problems. In particular, the authors’ idea in the study of Bidoux and Gaborit was to use the structure of these problems in the multi-instance setting to minimize the communication complexity of the resulting ZK PoK schemes. The security of the new schemes is then related to new hard problems. In this article, we focus on the new multivariate-based ZKPoK and the corresponding new underlying problem: the so-called DiffMQ.

Continue reading

Spectral graph analysis of bipartite graphs for advanced attack detection

Abstract

Spectral graph theory offers powerful tools for understanding graph properties through spectral signatures. This work leverages the inherent link between graph topology and spectral characteristics to enhance anomaly detection in network traffic, particularly in medical IoT networks. We introduce SPECTRA, a spectral graph analysis technique designed to detect anomalies in dynamic and complex network structures. This method incorporates five spectral metrics, including the newly proposed BiFlowness metric derived from Singular Value Decomposition (SVD), which captures the f low dynamics within bipartite graph topologies. By combining these spectral metrics, SPECTRA provides a comprehensive model for detecting and analyzing advanced cyberattack patterns, such as multistep intrusions, in critical systems. Focusing on hybrid topologies that integrate star and bipartite structures, this technique applies spectral analysis to evolving networks, enabling the detection of attacks (port scanning, fingerprinting) over time. Performed experiments validate the effectiveness of SPECTRA across IoT datasets, demonstrating its superiority in identifying attack behaviors. The proposed approach aligns with the critical demands of medical IoT environments by providing a good threat detection procedure to enhance security in sensitive networks.

Continue reading

An enhanced formalism for resource management policies specification and fast evaluation in pervasive systems

Abstract

Pervasive systems demand flexible, efficient resource management policies to handle heterogeneous infrastructures and varying application needs. This paper introduces an extended formalism that overcomes limitations in previous approaches by distinctly separating static properties from dynamic context elements, allowing more precise policy definitions. Mandatory and optional policies are explicitly categorized, enabling fail-fast decisions when critical conditions fail, while also supporting opportunistic executions. These design choices reduce evaluation costs—often down to O(1) in the best case—and permit large-scale environments to benefit from parallel evaluations. Practical simulations demonstrate superior performance in collaborative, multi-organization scenarios, highlighting improved adaptability, reduced overhead, and effective integration of organizational knowledge within the resource management process.

Continue reading

How effective are OS-level virtualization tools for managing containers?

Abstract

As reliance on OS-level virtualization tools grows, understanding their efficiency in container management tasks is essential for optimizing performance. This study presents a comprehensive performance analysis of Docker, Podman, and LXD across key container management tasks: loading, starting, stopping, and removing containers and images. Our results indicate Docker’s consistent superiority in speed, achieving the fastest execution times across tasks but at the cost of higher CPU usage. Podman demonstrates balanced resource efficiency, though generally slower than Docker in image loading. LXD, while slower in starting containers, exhibits lower CPU usage in parallel operations, making it suitable for scenarios where resource efficiency is prioritized over speed. These findings underscore the impact of tool choice on containerized environment performance, highlighting the importance of selecting a tool based on specific deployment requirements.

Continue reading

Pathological prior-guided multiple instance learning for mitigating catastrophic forgetting in breast cancer whole slide image classification

Abstract

In histopathology, intelligent diagnosis of Whole Slide Images (WSIs) is essential for automating and objectifying diagnoses, reducing the workload of pathologists. However, diagnostic models often face the challenge of forgetting previously learned data during incremental training on datasets from different sources. To address this issue, we propose a new framework PaGMIL to mitigate catastrophic forgetting in breast cancer WSI classification. Our framework introduces two key components into the common MIL model architecture. First, it leverages microscopic pathological prior to select more accurate and diverse representative patches for MIL. Secondly, it trains separate classification heads for each task and uses macroscopic pathological prior knowledge, treating the thumbnail as a prompt guide (PG) to select the appropriate classification head. We evaluate the continual learning performance of PaGMIL across several public breast cancer datasets. PaGMIL achieves a better balance between the performance of the current task and the retention of previous tasks, outperforming other continual learning methods. Our code will be open-sourced upon acceptance.

Continue reading

Performance evaluation of IoT-enabled edge computing infrastructure for mHealth services

Abstract

The rapid technological advancements and growing demand for innovative healthcare solutions highlight the critical role of integrating the Internet of Things (IoT) with mobile health (mHealth) services. This study evaluates the performance of an IoT platform within an mHealth context, focusing on the MQTT protocol’s effectiveness for healthcare data communication. Using an Orange Pi Win Plus board as the IoT platform, we simulated real-world mHealth conditions with varying workloads to assess platform resilience and scalability. Representative test scenarios were developed to simulate normal, increasing, and extreme load conditions, measuring key metrics such as CPU usage, memory consumption, throughput, and latency. Data were collected and analyzed using custom scripts to evaluate the platform’s response across different Quality of Service (QoS) levels. Results indicated that the platform could effectively manage standard and moderately high demands, while performance under extreme loads highlighted areas for optimization. This study concludes that the MQTT-based IoT platform demonstrated reliable performance in the mHealth environment, providing a basis for future optimizations and scalability improvements.

Continue reading

Performance evaluation of serverless computing infrastructure: Insights from open-source frameworks

Abstract

Serverless computing has gained widespread adoption due to its simplified management and lightweight design, particularly when paired with container orchestration systems like Kubernetes. By enabling developers to focus on application logic without managing underlying infrastructure, serverless computing offers advantages such as runtime-based billing in millisecond units, reducing operational costs and appealing to enterprises. This study evaluates resource utilization across 24 combinations of Ubuntu and Debian operating systems with Docker and Podman container platforms under varied workloads. Results indicate that Ubuntu with Docker achieves superior efficiency in CPU and RAM usage compared to other configurations. This experimental analysis provides practical insights into hardware resource management for serverless deployments and highlights opportunities for improving infrastructure in diverse scenarios.

Continue reading

Tumor detection, segmentation and classification challenge on automated 3D breast ultrasound: The TDSC-ABUS challenge

Abstract

Continue reading