Mahmoud Aljawarneh

Single-Source Shortest Path problem is one of the most basic algorithmic problems, the canonical solution to which is the Dijkstra algorithm. The one with a Fibonacci heap, with a worst-case time complexity of O(E+V log V), is practically regarded as the theoretical maximum. But between this theoretical equivalent and practical performance, there is an enormous gap especially in real-world large and highly sparse networks. The paper is a critique on the feasibility of complex heap structures in SSSP tasks. We perform a strict empirical study of an alternative algorithm strategy that completely avoids the costly Decrease-Key operation, in exchange maintaining a priority queue of edges itself, instead of a priority queue of vertices, sometime known as ”Lazy Dijkstra” algorithm. Acting on large-scale generated and real-world road network (up to 3.8 million edges), we show that this simplified edge-prioritized algorithm is by far the fastest logical implementation of Dijkstra algorithm with both Binary and Fibonacci heaps.Our performance demonstrates that there are up to 27 percent differences in the execution time between our results and the Fibonacci version. Moreover, we determine the crossover point of the algorithms (average graph degree ≈ 50) at which this benefit is not so significant. We explain this inversion of performance by modern locality of CPU caches and the amortization of the overhead of memory-management of the pointer-intensive data structures. We have found it to be better in practice to go back to a simpler, hardware-friendly algorithmic core, when dealing with the ubiquitous category of sparse networks.

The convergence of Industrial Internet of Things (IIoT) and hybrid clouds creates data-rich ecosystems but introduces severe security vulnerabilities, especially from insider threats who can bypass traditional defenses. This challenge is amplified in distributed IIoT settings where compromised components act as insiders. This paper proposes a novel framework synergizing Federated Learning (FL) for privacy-preserving, decentralized anomaly detection with a Zero Trust (ZT) architecture for robust security. Using the HAI dataset, our simulations show FL models achieve high discriminative power (AUC > 0.95) but struggle with precision, leading to volatile F1-scores. We critically analyze the framework's resilience to attacks, finding that while ZT provides a secure operational foundation, more intelligent ZT mechanisms are needed to actively filter content-based model attacks, a key direction for future work.

Pervasive services use non-compatible protocols (SDPs, SOAP, REST, UPnP), which are difficult to interoperate with. In this paper, the new broker architecture, ReSHAPE, solving this service heterogeneity, is introduced. ReSHAPE is the only system to combine centralized knowledge of service and native protocol interaction functionality, and coordinated by the broker. This helps reduce the overhead of translation used in intermediate techniques as well as reducing the client-side overhead of dealing with real multi-protocols. ReSHAPE will allow an efficient direct communication using different types of services by leveraging a protocol-agnostic gateway and specialized adapters. Experimental measurements with Web Services, UPnP and RMI adapters show that there is little operational overhead in local and distributed applications. ReSHAPE proposes a scalable and extendable framework to provide a significant improvement in the service interoperability and user experience in complex smart environments. The future direction is to add more protocol adapter support and add more optimization and security capabilities.

In the era of Edge Computing and the growing presence of the Internet of Things (IoT), the traditional network perimeter has been broken, making static access control models, such as Role-Based Access Control (RBAC), inadequate. Edge deployments are very diverse, resource-constrained and have very dynamic environments, which requires adaptation to security in real-time. In order to tackle these problems, the paper introduces a modular, context-aware adaptive security framework, ContexGuard, specially tailored to dynamic edge environments. ContexGuard provides a plug-in architecture that separates context acquisition service logic from policy decision service logic, allowing for dynamic plug-in of new sensors, policies, and AI-based risk models at runtime, without disrupting services. Moreover, a lightweight resilient orchestration between distributed edge nodes is enabled by the framework, which is based on a native Message Queuing Telemetry Transport (MQTT) coordination mechanism. Realistic synthetic workloads and multi-node simulations are used to evaluate ContexGuard. The experimental results show fine-grained and context-aware authorization where the latency overhead per request is roughly 8.0ms and is linear up to 500 requests per second. The results show the effectiveness of ContexGuard to close the theory-to-practice gap between Zero Trust principles and resource-efficient, edge deployments.

Retrieval-Augmented Generation (RAG) has emerged as a transformative approach for enhanc-ing large language models (LLMs) by integrating external knowledge sources to address limitations in parametric knowledge and factual accuracy. The survey offers a detailed comparative study of RAG methods in dense retrieval, sparse retrieval, and hybrid approaches and their performance on knowledge-intensive tasks. Through systematic evaluation of 85 recent studies (2020-2024), we present empirical comparisons showing that hybrid RAG systems achieve 15-35% performance im-provements over baseline LLMs on question answering tasks, while dense retrieval methods demon-strate superior semantic understanding capabilities. We find that although RAG systems promise much, their current approach is limited by computational efficiency, retrieval precision, and stan-dardization of evaluation. We propose an integrated assessment system and outline five essential future avenues, including multi-modal RAG combination and adaptive retrieval schemes.

The paradigm of decentralized cities, reliant on edge computing, redefines urban functionality but introduces acute cybersecurity challenges within a borderless threat environment. Traditional perimeter-based security is ill-suited for dispersed edge assets and sophisticated adversaries. This chapter critically examines these challenges, dissects threat vectors, and proposes adaptive defense mechanisms. Through simulation-based evaluation over multiple replications (N=5), we demonstrate the efficacy of a combined defense strategy (Basic Firewall and AI-driven IDS) against defined attack scenarios. Averaged results show substantial improvements in threat detection (e.g., F1-Score increased from 0.0000 to 0.8876 ± 0.0150) and mitigation compared to a baseline, alongside a quantified analysis of performance overhead and consistently immediate Time-to-Detect (TTD). The work provides foundational insights, robust quantitative validation, and strategic foresight, emphasizing proactive postures essential for safeguarding future decentralized smart cities.

The digital transformation of cities into smart urban ecosystems creates a vast and complex cybersecurity attack surface. This paper provides a foundational survey of this threat landscape, systematically deconstructing the smart city's anatomy from a security perspective. We analyze inherent vulnerabilities within key technologies like IoT, 5G/6G, AI, and cloud/edge computing. The survey presents a multi-dimensional taxonomy of threats, classifying them by attacker intent, target asset, and vector, while examining domain-specific vulnerabilities in transportation, energy, and healthcare. A forward-looking analysis of emerging threats is also provided. The objective is to establish a comprehensive, structured understanding of the threats themselves, offering an essential baseline for researchers, practitioners, and policymakers dedicated to securing the cities of tomorrow.