This study addresses a critical challenge in medical imaging diagnostics by proposing a unified, lightweight deep learning model capable of diagnosing multiple diseases across diverse imaging modalities, including chest X-rays, MRIs, skin images, and endoscopic images, within a single efficient framework. Each modality presents unique feature characteristics, introducing complexities in the diagnostic process. To enhance image quality, we apply Contrast Limited Adaptive Histogram Equalization and utilize the Vision Transformer for improved feature extraction and diagnostic performance. To tackle remaining challenges, we introduce the ChirpMBPru-Net model, designed to analyze multiple image modalities in medical imaging while minimizing computational demands. This model employs the efficient MobileNet architecture as its backbone and systematically applies pruning to remove redundant layers. Moreover, a dense module for multi-scale feature extraction and the Chirplet transformation are employed in the pruned model, capturing both frequency and spatial patterns at varying scales. Additionally, the ChirpMBPru-Net model demonstrates its versatility by adapting to domain shifts in engineering fields, such as defect detection in industrial applications (e.g., scholar defect detection), where it can classify multiple categories of the same object or defect type. The model achieves an impressive accuracy of 97% across 16 disease categories and proves effective in handling real-world domain shifts, demonstrating its potential for both medical and engineering applications.

Conventional standalone approaches for diagnosing individual diseases often fail to achieve robust generalization because they are severely impacted by overfitting. This results in poor adaptability to diverse image representations and an inability to balance performance with computational efficiency. In this study, we propose KDLight, a lightweight, novel CNN model designed for efficient medical image classification across diverse modalities, including MRI, X-ray, radiography, skin images, and histopathology. We employ Knowledge Distillation (KD), where insights from an efficient teacher model (MobileNet) guide the learning process of the KDLight student model. The KDLight model minimizes the number of parameters while enhancing feature learning across diverse medical image representations. Experimental results show that KDLight achieves 95.55% classification accuracy with only 2.96 seconds and a compact 7.5 MB disk size, significantly reducing parameter size, accelerating inference, and lowering computational costs compared to traditional pre-trained models. Additionally, KDLight ability to efficiently learn diverse image representations can be extended to other domains, such as crack classification (e.g., road, window, and building cracks), enabling high-performance detection across different surface defect categories.

Unstructured (or, semi-structured) natural language is mostly used to capture the requirement specifications both for legacy software systems and for modern day software systems. The adoption of a formal approach to the specification of the requirements, using goal models, enables rigorous and formal inspections while analyzing the requirements for satisfiability, consistency, completeness, conflicts and ambiguities. However, such a formal approach is often considered burdening for the analysts’ activity as it requires additional skills, and is therefore, discarded a priori. This works aims to bridge the gap between natural language requirement specifications and efficient goal model analysis techniques. We propose a framework that uses extensive natural language processing techniques to transform a set of unstructured natural language requirement specifications to the corresponding goal model. We combine techniques such as parts-of-speech tagging, dependency parsing, contextual and synonymy vector generation with the FiBER transformer model. An extensive unbiased crowd-sourced evaluation of the proposed framework has been performed, showing an acceptability rate (total and partial combined) of 95%. Time and space analyses of our framework also demonstrate the scalability of the proposed solution.

Abstract In autonomous and robotic systems, the functional requirements (FRs) and non‐functional requirements (NFRs) are gathered from multiple stakeholders. The different stakeholder requirements are associated with different components of the robotic system and with the contexts in which the system may operate. This aggregation of requirements from different sources (multiple stakeholders) often results in inconsistent or conflicting sets of requirements. Conflicts among NFRs for robotic systems heavily depend on features of actual execution contexts. It is essential to analyze the inconsistencies and conflicts among the requirements in the early planning phase to design the robotic systems in a systematic manner. In this work, we design and experimentally evaluate a framework, called SCARS, providing: (a) a domain‐specific language extending the ROS2 Domain Specific Language (DSL) concepts by considering the different environmental contexts in which the system has to operate, (b) support to analyze their impact on NFRs, and (c) the computation of the optimal degree of NFR satisfaction that can be achieved within different system configurations. The effectiveness of SCARS has been validated on the iRobot Create3 robot using Gazebo simulation.

Named entity recognition (NER) is a natural language processing task that has been used in Requirements Engineering for the identification of entities such as actors, actions, operators, resources, events, GUI elements, hardware, APIs, and others. NER might be particularly useful for extracting key information from Software Requirements Specification documents, which provide a blueprint for software development. However, a common challenge in this domain is the lack of annotated data. In this article, we propose and analyze two zero-shot approaches for NER in the requirements engineering domain. These are found to be particularly effective in situations where labeled data is scarce or non-existent. The first approach is a template-based zero-shot learning mechanism that uses the prompt engineering approach and achieves 93% accuracy according to our experimental results. The second solution takes an orthogonal approach by transforming the entity recognition problem into a question-answering task which results in 98% accuracy. Both zero-shot NER approaches introduced in this work perform better than the existing state-of-the-art solutions in the requirements engineering domain.

Abstract In the design of autonomous systems, it is important to consider the preferences of the interested parties to improve the user experience. These preferences are often associated with the contexts in which each system is likely to operate. The operational behavior of a system must also meet various non-functional requirements (NFRs), which can present different levels of conflict depending on the operational context. This work aims to model correlations between the individual contexts and the consequent conflicts between NFRs. The proposed approach is based on analyzing the system event logs, tracing them back to the leaf elements at the specification level and providing a contextual explanation of the system’s behavior. The traced contexts and NFR conflicts are then mined to produce Context-Context and Context-NFR conflict sequential rules. The proposed Contextual Explainability (ConE) framework uses BERT-based pre-trained language models and sequential rule mining libraries for deriving the above correlations. Extensive evaluations are performed to compare the existing state-of-the-art approaches. The best-fit solutions are chosen to integrate within the ConE framework. Based on experiments, an accuracy of 80%, a precision of 90%, a recall of 97%, and an F1-score of 88% are recorded for the ConE framework on the sequential rules that were mined.

The Crowdsourced IoT Service (CIS) market is inherently different from other service markets, e.g., web services and cloud. The CIS market is dominated by transient services as both consumers and providers are dynamic in space and time. Consumer requests are usually long-term and demand continuity in service provision. We propose a novel egalitarian transient service composition framework from the CIS market perspective. We apply a Dynamic Bayesian Network to model the dynamic service provision behavior of the providers. The proposed framework transforms the composition of transient services into a multi-objective temporal optimization, i.e., providing continuous services to the maximum number of consumers, and minimizing the consumers’ cost of service usages over a long-term period. We incorporate a Pareto-based genetic algorithm to enable the fair distribution of services among the consumers. Experimental results prove the efficiency of the proposed approach in terms of continuous availability of service as well as fair distribution among consumers.

An emerging feature of mobile application software is the need to quickly produce new versions to solve problems that emerged in previous versions. This helps adapt to changing user needs and preferences. In a continuous software development process, the user reviews collected by the apps themselves can play a crucial role to detect which components need to be reworked. This paper proposes a novel framework that enables software companies to drive their technology value stream based on the feedback (or reviews) provided by the end-users of an application. The proposed end-to-end framework exploits different Natural Language Processing (NLP) tasks to best understand the needs and goals of the end users. We also provide a thorough and in-depth analysis of the framework, the performance of each of the modules, and the overall contribution in driving the technology value stream. An analysis of reviews with sixteen popular Android Play Store applications from various genres over a long period of time provides encouraging evidence of the effectiveness of the proposed approach.

Requirements keep changing and getting updated in any incremental software development - including DevOps. The notion of regression validation ensures compliance to a given set of rules (or properties) even when either the requirement set changes or the associated rules themselves change due to new business policies and regulations. In this tool paper, we propose the RV-SLC tool which has the SLC framework at its core. The tool uses a formal data model that demonstrates how regression validation of safety and liveness constraints can be achieved across multiple iterations. The tool also has an analytics dashboard that allows the developers to monitor and visualize how the requirements and the associated safety and liveness constraints have been modified and updated across these iterations.

Requirement prioritization is an inherently important step in the DevOps framework. Unfortunately, the prioritization process often disregards the non-functional requirements and the possible conflicts among them. This implies that unresolved dependencies and conflicts would be identified at integration time only, which may lead to major refactoring issues. We introduce CARO a new tool that generates an ordering among the requirements based on conflicts and dependencies among the requirements. The tool provides a quantitative risk evaluation framework along with risk mitigation strategies based on conflicts and dependencies among the requirements.

We propose a tool prototype that allows the evolution of goal models in changing business environments. Like jUCMNav, the tool supports annotated goal models where goals, tasks, and resources may be annotated with specific contexts. Softgoal context annotations and their interpretations are quite complex and left out for the time being. The tool performs a qualitative evaluation for identifying conflicts between annotations. It also performs reconciliation of each conflict and presents the user with conflict-free goal model alternatives.

Goal model like i* are inherently sequence agnostic whereas standard model checkers like SPIN or NuSMV accept extended finite state models which represent an ordering of state transitions that can occur within a system. These two notions are mutually conflicting and, thus, we cannot check temporal properties (like compliance rules), described using temporal languages (like Linear Temporal Logic (LTL), or Computational Tree Logic (CTL)), on goal models by feeding them as input to a standard model checker. The i*ToNuSMV tool aims to bridge this gap such that the compliance of i* models towards temporal constraints can be verified in the early requirements phase of software development itself.

ABSTRACT Wireless mesh networks (WMNs) are proliferating as one of the key technologies of the next‐generation networks. Security is one of the prime concerns towards actual implementation of any network technology for commercial applications. Network security has intrinsically two approaches—prevention based and detection based. Implementing firewalls or intrusion prevention techniques is often not an attractive solution for energy‐constrained network nodes such as mobile ad hoc network (MANET) nodes or mesh clients in WMNs. However, in the era of pervasive and ubiquitous computing, commercial transactions are performed on the move and over portable devices such as cell phones and laptops. These devices have energy constraints, and hence, one cannot afford to adopt security measures with high computational overhead. This influences a shift in paradigm from active intrusion prevention to passive intrusion detection. In this paper, a new cluster‐oriented reward‐based intrusion detection system (CORIDS) has been proposed for WMNs. The performance of CORIDS has been evaluated using the Qualnet network simulator. Simulation results also establish superiority of CORIDS over Misbehavior Detection Algorithm, another recent trust‐based IDS for WMN, both in terms of higher detection efficiency and lower false positives. Copyright © 2013 John Wiley & Sons, Ltd.