The Industrial Internet of Things (IIoT) is a transformative paradigm that integrates smart sensors, advanced analytics, and robust connectivity within industrial processes, enabling real-time data-driven decision-making and enhancing operational efficiency across diverse sectors, including manufacturing, energy, and logistics. IIoT is susceptible to various attack vectors, with Advanced Persistent Threats (APTs) posing a particularly grave concern due to their stealthy, prolonged, and targeted nature. The effectiveness of machine learning-based intrusion detection systems in APT detection has been documented in the literature. However, existing cybersecurity datasets often lack crucial attributes for APT detection in IIoT environments. Incorporating insights from prior research on APT detection using provenance data and intrusion detection within IoT systems, we present the CICAPT-IIoT dataset. The main goal of this paper is to propose a novel APT dataset in the IIoT setting that includes essential information for the APT detection task. In order to achieve this, a testbed for IIoT is developed, and over 20 attack techniques frequently used in APT campaigns are included. The performed attacks create some of the invariant phases of the APT cycle, including Data Collection and Exfiltration, Discovery and Lateral Movement, Defense Evasion, and Persistence. By integrating network logs and provenance logs with detailed attack information, the CICAPT-IIoT dataset presents foundation for developing holistic cybersecurity measures. Additionally, a comprehensive dataset analysis is provided, presenting cybersecurity experts with a strong basis on which to build innovative and efficient security solutions.

Privacy-preserving location-based skyline queries (LBSQ) enable location-based services (LBS) over outsourced spatial data while preserving data privacy. Prior studies primar ily consider Euclidean distance as the spatial-attribute metric. However, this does not capture reachability in three-dimensional (3D) obstacle environments, where obstacles invalidate direct Euclidean distances and necessitate obstacle-aware path computations. In this work, we present PLOSQ3D, a privacy-preserving location-based skyline query framework in 3D obstacle space. We first formalize the problem, specify its privacy requirements, and identify the challenges introduced by obstacle-aware distances. Subsequently, we introduce a novel index, Secure Augmented R-Tree (SAR-Tree) that jointly encodes spatial and non-spatial at tributes, enabling effective pruning under 3D spatial constraints. To enforce privacy, we leverage Function Secret Sharing (FSS) to design three protocols: SLDC for secure dominance computation, SSPC for secure shortest-path computation, and OLDU for oblivious skyline-set update. Building on them, we carefully design the PLOSQ3D scheme. Our security analysis demonstrates strong privacy guarantees against honest-but-curious adversaries. We further conduct extensive experiments on both synthetic and real 3D datasets. Experimental results show that our framework achieves high efficiency in both computation and communication overhead, confirming its practicality for 3D spatial applications that demand spatial-awareness and privacy protection.

Reverse spatial skyline query (RSQ) is a powerful multi-criteria decision tool that has been widely used in recommendation and spatial decision-making applications. However, existing RSQ research either overlooks realistic 3D obstacle-aware reachability constraints or fails to provide privacy protection in outsourced settings. In practice, outsourcing spatial data to the Cloud introduces serious privacy risks, since data objects, query requests, and query results may reveal sensitive information. Meanwhile, incorporating 3D obstacle-aware spatial constraints into RSQ is fundamentally challenging, especially when spatial data must remain hidden. To address these issues, we propose PRSQO3D, the first privacy-preserving reverse skyline query scheme for realistic obstacle-aware 3D environments. Our design supports secure reverse dominance computation and obstacle detection by leveraging a non-interactive homomorphic secret sharing (NHSS) technique, together with the Secure R-tree (SR-Tree) and the Absolute-Difference Coding (ADC) mechanism. This design enables efficient pruning and inner-product-based dominance evaluation while preserving data privacy. We further develop a privacy-preserving obstacle-detection protocol to ensure correctness under reachability constraints without leaking spatial or query information. We provide a complete PRSQO3D framework and formal proofs showing that our scheme achieves simulation-based security against honest-but-curious adversaries. Experiments on California and Florida Open City Model datasets show that PRSQO3D completes full obstacle-aware RSQ over 10,000 objects in about 0.9s, remains executable with 100 non-spatial attributes, and outperforms the closest baselines in component-level comparisons.

Accurate prediction of maritime transit delays is critical for improving port efficiency and ensuring safe and reliable vessel navigation. Although traditional statistical and deep learning approaches demonstrate predictive capabilities, they often lack interpretability and provide limited insight into the causal factors underlying delays. This study proposes a causal machine learning framework for transit delay prediction in the Kattegat Strait, leveraging Automatic Identification System (AIS) data, port call logs, and meteorological information like wind speed and direction. By applying causal inference techniques, including individual treatment effect (ITE) and average treatment effect (ATE) estimation, the framework quantifies the impact of key operational and environmental variables. The results underscore the significant influence of weather conditions (notably wind speed) and port operations like berth time on vessel transit time. The model enables vessel-class-specific analysis, differentiating the effects on cargo ships, passenger ships, HSC, and tankers, and achieves high predictive accuracy (98.19 %) while maintaining interpretability. The proposed framework offers actionable insights to support operational decision-making, including berth scheduling, route planning, and congestion mitigation.

Vehicular crowdsensing enables Connected and Autonomous Vehicles (CAVs) to jointly contribute driving-related data to support applications such as traffic management and accident analysis. Ensuring the reliability of such data requires identifying observations that have been corroborated by multiple vehicles. However, achieving this corroboration typically necessitates comparing each vehicle’s private set of observations, which can inadvertently reveal sensitive trajectory information. To address this privacy challenge, we introduce Edge-Assisted Private Set Intersection (EA-PSI), a scheme that enables secure computation of the intersection among CAV observation sets without disclosing individual data elements. Our design begins with a protocol that leverages polynomial-based set encoding and additive secret sharing, in which each vehicle encodes its observation set as a polynomial and divides it into two shares, delegating one share to a roadside unit (RSU) while retaining the other locally. The RSU then performs intersection computation on the collected shares using randomized encoding techniques, without learning any private observations or intersection results. To support necessary polynomial operations over secret-shared data, we develop a secure multiplication mechanism based on Beaver triples, enabling the RSU and vehicles to jointly compute polynomial products without reconstructing underlying values. In addition, we design a key-distribution protocol that facilitates secure communication among vehicles through the RSU, eliminating the need for direct vehicle-to-vehicle exchange. We analyze the security of EA-PSI under the simulation-based paradigm and formally prove privacy against semi-honest adversaries. Experimental evaluation of computational and communication costs demonstrates the efficiency and practicality of the proposed EA-PSI scheme.

Connected and autonomous vehicles (CAVs) equipped with diverse sensors can enhance environmental perception through cooperative sensing, overcoming individual sensor limitations such as restricted range and occlusion. However, privacy concerns regarding location exposure and data leakage significantly hinder widespread adoption. This article presents a comprehensive privacy-preserving cooperative sensing framework that enables secure data sharing among CAVs without compromising performance. We introduce two key innovations: the vehicular spatial index tree (VSITree), which provides efficient spatial indexing while preventing location leakage through cryptographic encoding, and the vehicular attribute matching protocol (VAMP), which enables oblivious membership testing between encrypted sensing data and queries. Our framework leverages arithmetic secret sharing (ASS) and predicate encryption to protect both sensing providers (SPs) and requesters throughout the data lifecycle. The system is designed to operate through roadside units (RSUs) that facilitate secure matching and aggregation without learning sensitive information. Theoretical analysis and extensive simulations demonstrate the security and efficiency properties of our approach, confirming its resilience against various attack vectors while maintaining real-time performance suitable for safety-critical vehicular applications.

Industrial Internet of Things (IIoT) systems in sectors like manufacturing, energy, and healthcare are increasingly deployed in cloud-assisted operational environments, where network telemetry and security analytics are routinely processed in the cloud. However, these systems remain highly vulnerable to cyber threats from shared threat actors. Among these, low-rate Denial of Service (LDoS) attacks, marked by subtle periodic traffic patterns, are particularly challenging to detect when analyzed in isolation. Cross-organization collaborative detection across cloud platforms can improve identification accuracy, but sharing threat intelligence risks exposing sensitive operational information. To tackle this challenge, we propose Labelled-Threshold Private Set Intersection (LT-PSI), a cryptographic framework that allows two organizational clouds to securely identify common elements whose associated label vectors satisfy a similarity threshold, without revealing any additional data. Our LT-PSI protocol introduces an innovative combination of position encoding, Diffie-Hellman Oblivious Pseudorandom Functions (DH-OPRF), and Bloom filters, effectively transforming threshold-based label similarity matching into efficient and privacy-preserving set membership tests. Particularly, our protocol achieves sublinear online complexity and is well-suited for cloud execution, integrating an adaptive early termination strategy that significantly reduces the number of OPRF invocations. We provide formal security proofs under the semi-honest model and validate the protocol through extensive experiments across diverse similarity thresholds and dataset sizes. Results show that LT-PSI is significantly more efficient than brute-force threshold matching while preserving privacy. The framework naturally supports cloud-to-cloud collaborative security analytics and generalizes to broader cloud and edge threat intelligence scenarios requiring private, threshold-based feature matching.

As the aging population continues to grow, healthcare systems face increasing pressure to support timely and personalized medical services to older adults. To address this need, Aging in Place (AiP) has emerged as a promising approach, enabling seniors to remain in their own homes while still receiving continuous medical support. AiP services increasingly rely on similarity search over sensitive electronic health records (EHRs) to enable personalized interventions. To support such queries with limited local resources, medical institutions outsource EHRs to cloud servers. However, outsourcing raises three key challenges: preserving data and query privacy, hiding access patterns, and allowing flexible similarity measures. To address these challenges, we propose an efficient, privacy-preserving scheme for fine-grained similarity queries tailored to AiP. Our design combines p-stable locality-sensitive hashing (p-stable LSH) with additive secret sharing (ASS) and function secret sharing (FSS) to return all records that satisfy users’ fine-grained similarity requirements while ensuring privacy throughout the process. Our security analysis confirms privacy guarantees under the honest-but-curious model. Extensive experimental evaluations demonstrate computational efficiency and practical scalability for real-world AiP implementations.

Database query processing and optimization are critical components of modern database management systems (DBMS) that efficiently process user queries. In big data application scenarios, the movement of large volumes of data influences performance, power efficiency, and reliability, which are the three essential aspects of a computing system. Large-scale data centers require an exceptionally efficient server and storage infrastructure. The systems currently employed for managing and processing big data are increasingly showing inefficiency, both in terms of energy usage and scalability, primarily due to the constraints imposed by existing CPU architectures. A significant challenge in Database Management Systems (DBMS) is the growing disparity between the speeds of processors and memory access, which results in notable performance bottlenecks.This paper presents a comprehensive survey of reconfigurable acceleration in database systems, offering a structured taxonomy that categorizes existing work based on query types, integration models, and hardware/software co-design strategies. We examine key acceleration techniques across relational operators, indexing, join algorithms, and compression, highlighting their trade-offs in performance, scalability, and adaptability. Furthermore, we identify current limitations in programmability, data movement, and workload variability, and outline open research challenges including dynamic reconfiguration, hybrid architectures, and compiler support. This taxonomy-driven perspective aims to guide both researchers and practitioners in navigating the design space and pushing the boundaries of FPGA-accelerated data processing.

This paper introduces HQT-TI, a novel indexing method designed to improve the efficiency of spatial keyword queries. HQT-TI consists of two main components: a Hilbert QuadTree (HQT) based spatial index and a Trie-Inverted index (TI) combined textual index. HQT integrates the Hilbert curve with a Quadtree, establishing a direct relationship between the two. TI combines a trie and inverted index to minimize the intersection cost associated with long lists, thus improving the speed of keyword queries. The HQT based Spatial Query algorithm (HQT-SQ) reduces overlap checks and limits irrelevant object retrieval by employing query drill-down and depth first search with limited breadth expansion in spatial queries. Meanwhile, the Segment List Intersection based Keyword Query algorithm (SLI-KQ), built on TI, efficiently handles segment list intersections for keyword queries. The combination of HQT-SQ and SLI-KQ results in HS-SK, a highly efficient spatial keyword query algorithm. Extensive experimental results demonstrate that HS-SKQ outperforms SFC-Quad by up to two orders of magnitude, achieving up to a 5.46× speedup over the best existing competitors, making it a promising solution for large-scale spatial keyword query processing.

Aging in Place (AiP) programs enable elderly individuals to live independently and comfortably within their homes and communities by utilizing technological innovations such as smart homes and remote healthcare monitoring. In practical AiP scenarios, data necessary for accurate health predictions are typically vertically partitioned across multiple medical institutions, raising significant privacy concerns during data integration and analysis. To address this challenge, we propose an efficient and privacy-preserving logistic regression (LR) prediction scheme tailored explicitly for vertically partitioned AiP data. Our scheme effectively combines the computational efficiency of Trusted Execution Environments (TEE) under the honest-but-curious model with cryptographic security based on the Matrix Diffie-Hellman (MDDH) assumption. Security analysis confirms that our approach provides privacy protections against honest-but-curious adversaries. Extensive experimental evaluations demonstrate that our proposed scheme achieves computational efficiency, privacy protection, and practical scalability for real-world AiP implementations.

In relational database management systems (DBMS), the join operation is of utmost importance as it merges rows from two different tables into a unified dataset. This procedure is executed by pinpointing a shared column that serves as a link between the tables involved. Consequently, the join operation enables the simultaneous retrieval and accessibility of information from multiple tables. Accelerating joins in DBMS improves query performance, reduces response time, and efficiently handles large datasets. This is important for applications that require real-time data processing and analytics. This paper introduces a novel framework designed for accelerating the join operation using a learned index. This approach, which is called LIJA (Learned Index-based Join Acceleration), can significantly improve the join operations across multiple tables, speeding up the execution time. In our experimental analysis, we thoroughly assess the CPU efficiency of our approach, comparing it with traditional methods such as hash-join, merge-join, nested-loop join, and partition-wise join techniques. Employing LIJA, our research uncovered about a 14.9x enhancement in performance during the build phase, along with an over 1.8x speedup during the probe phase when assessed using SOSD benchmark datasets. However, as data sizes grow, this often leads to longer runtime and higher memory consumption. These challenges motivated us to leverage FPGA technology to accelerate the build phase of our learned join approach. After this implementation, we evaluated and compared its efficiency with CPUs. Our results indicated that with the FPGA-based learned join accelerator, we attained an approximate 4.4x speedup during the build phase, and total speedup of 6.63x over the CPU-based implementation.

The increasing frequency of road traffic accidents worldwide poses significant challenges in economic, societal, and public health areas, resulting in millions of injuries and deaths each year. This research utilizes machine learning methods to forecast the severity of accidents and identify primary contributing factors, using data from Ethiopia. To address the imbalance in these datasets, the Synthetic Minority Oversampling Technique (SMOTE) is applied to ensure balanced data representation. For predicting accident severity, Random Forest and XGBoost is used, with Explainable Artificial Intelligence (XAI) methods-namely Shapley Additive eXplanations (SHAP) for overarching insights and Local Interpretable Model-agnostic Explanations (LIME) for detailed, localized interpretations. This study enhances our understanding of road traffic accidents and helps mitigate their effects through sophisticated analytical approaches, providing valuable insights for policymakers, urban planners, and public health officials worldwide.

Similarity search is the problem of finding in a collection of objects those that are similar to a given query object. It is a fundamental problem in modern applications and the objects considered may be as diverse as locations in space, text documents, images, twitter messages, or trajectories of moving objects. In this paper we are motivated by the latter application. Trajectories are recorded movements of mobile objects such as vehicles, animals, public transportation, or parts of the human body. We propose a novel distance function called DistanceAvg to capture the similarity of such movements. To be practical, it is necessary to provide indexing for this distance measure. Fortunately we do not need to start from scratch. A generic and unifying approach is metric space, which organizes the set of objects solely by a distance (similarity) function with certain natural properties. Our function DistanceAvg is a metric. Although metric indexes have been studied for decades and many such structures are available, they do not offer the best performance with trajectories. In this paper we propose a new design, which outperforms the best existing indexes for kNN queries and is equally good for range queries. It is especially suitable for expensive distance functions as they occur in trajectory similarity search. In many applications, kNN queries are more practical than range queries as it may be difficult to determine an appropriate search radius. Our index provides exact result sets for the given distance function.

Indexes can significantly improve search performance in relational databases. However, if the query workload changes frequently or new data updates occur continuously, it may not be worthwhile to build a conventional index upfront for query processing. Adaptive indexing is a technique in which an index gets built on the fly as a byproduct of query processing. In recent years, research in database indexing has taken a new direction where machine learning models are employed for the purpose of indexing. These indexes, known as learned indexes, can be more efficient compared to traditional indexes such as B+-tree in terms of memory footprints and query performance. However, a learned index has to be constructed upfront and requires training the model in advance, which becomes a challenge in dynamic situations when workload changes frequently. To the best of our knowledge, no learned indexes exist yet for adaptive indexing. We propose a novel learned approach for adaptive indexing. It is built on the fly as queries are submitted and utilizes learned models for indexing data. To enhance query performance, we employ a query workload prediction technique that makes future workload projection based on past workload data. We have evaluated our learned adaptive indexing approach against existing adaptive indexes for various query workloads. Our results show that our approach performs better than others in most cases, offering 1.2x - 5.6x improvement in query performance.

Leveraging travel time data is essential for various applications such as ride sharing, traffic management, and traffic accident detection. Despite significant advancements in travel time forecasting, selecting the most appropriate prediction method remains challenging due to the constantly changing nature of road traffic. Previous research have investigated several machine learning models for predicting travel time; however, the opaque nature of these models often limits the ability to explain their predictions, creating a notable research gap. Recognizing the potential of eXplainable AI (XAI) techniques to enhance decision-making, our study aims to address this gap by incorporating XAI methods. Our work introduces a taxonomy of XAI techniques and applies explainable travel time prediction methods from key XAI categories, specifically ante-hoc techniques (such as QLattice) and post-hoc techniques (including SHAP, LIME, and ELI5). We conduct a thorough evaluation using two real-world datasets, NYC taxi and California PeMS, demonstrating that XAI techniques can improve predictive accuracy and provide insights into the factors affecting travel times, while also showcasing the comparative performance of different XAI-based approaches for travel time prediction.

The pay-what-you-use model of serverless Cloud computing (or serverless, for short) offers significant benefits to the users. This computing paradigm is ideal for short running ephemeral tasks, however, it is not suitable for stateful long running tasks, such as complex data analytics and query processing. We propose FunDa, an on-premises serverless data analytics framework, which extends our previously proposed system for unified data analytics and in situ SQL query processing called DaskDB. Unlike existing serverless solutions, which struggle with stateful and long running data analytics tasks, FunDa overcomes their limitations. Our ongoing research focuses on developing a robust architecture for FunDa, enabling true serverless in on-premises environments, while being able to operate on a public Cloud, such as AWS Cloud. We have evaluated our system on several benchmarks with different scale factors. Our experimental results in both on-premises and AWS Cloud settings demonstrate FunDa’s ability to support automatic scaling, low-latency execution of data analytics workloads, and more flexibility to serverless users.