Sparse-view tomographic reconstruction aims to recover 3D volumes from limited projection views, but often suffers from incomplete structures and volumetric artifacts. Gaussian splatting has recently emerged as an efficient representation for continuous volumetric modeling, reducing memory cost compared to voxel grids and training time compared to implicit methods. However, existing Gaussian splatting methods for CT reconstruction struggle with needle-like artifacts in sparse-view settings. To address this, we introduce two key contributions. First, we propose a structure-aware initialization strategy that uses gradient and density magnitude from preliminary reconstructions to intelligently place Gaussian primitives in high-contrast regions. Second, we adapt the well-established Beer-Lambert law from CT physics to stabilize Gaussian splatting optimization, transforming the exponential attenuation relationship into a linear domain that mitigates vanishing gradients, and stabilizes optimization. Together, these innovations lead to sharper and more stable reconstructions, achieving average improvements of 2.32 % in PSNR and 2.41 % in SSIM while using 6.47 % fewer primitives across three standard CT datasets.

Graph Neural Networks (GNNs) power applications such as content recommendation, knowledge graph reasoning, and social network analysis, where modeling both structure and features is essential.Knowledge Distillation (KD) enables transferring knowledge from large GNNs to compact models for efficient deployment, yet most approaches rely on a pre-trained teacher.We propose a mutual learning framework in which shallow GNNs collaboratively distill knowledge by iteratively exchanging predictions during training.The framework integrates adaptive logit weighting to balance peer influence and entropy enhancement to promote exploration and prevent early convergence.Experiments on multiple benchmark datasets show that our approach improves GNN performance and that the learned knowledge can be effectively transferred to lightweight graph-less models, offering a scalable alternative for graph learning.

Autonomous vehicles rely on deep neural networks (DNNs) for traffic sign recognition, lane centering, and vehicle detection, yet these models are vulnerable to attacks that induce misclassification and threaten safety. Existing defenses (e.g., adversarial training) often fail to generalize and degrade clean accuracy. We introduce Vehicle Vision-Language Models (V2LMs), fine-tuned vision-language models specialized for autonomous vehicle perception, and show that they are inherently more robust to unseen attacks without adversarial training, maintaining substantially higher adversarial accuracy than conventional DNNs. We study two deployments: Solo (task-specific V2LMs) and Tandem (a single V2LM for all three tasks). Under attacks, DNNs drop 33-74%, whereas V2LMs decline by under 8% on average. Tandem achieves comparable robustness to Solo while being more memory-efficient. We also explore integrating V2LMs in parallel with existing perception stacks to enhance resilience. Our results suggest V2LMs are a promising path toward secure, robust AV perception.

Parasitic infections remain a pressing global health challenge, particularly in low-resource settings where diagnosis still depends on labor-intensive manual inspection of blood smears and the availability of expert domain knowledge. While deep learning models have shown strong performance in automating parasite detection, their clinical usefulness is constrained by limited interpretability. Existing explainability methods are largely restricted to visual heatmaps or attention maps, which highlight regions of interest but fail to capture the morphological traits that clinicians rely on for diagnosis. In this work, we present MorphXAI, an explainable framework that unifies parasite detection with fine-grained morphological analysis. MorphXAI integrates morphological supervision directly into the prediction pipeline, enabling the model to localize parasites while simultaneously characterizing clinically relevant attributes such as shape, curvature, visible dot count, flagellum presence, and developmental stage. To support this task, we curate a clinician-annotated dataset of three parasite species (Leishmania, Trypanosoma brucei, and Trypanosoma cruzi) with detailed morphological labels, establishing a new benchmark for interpretable parasite analysis. Experimental results show that MorphXAI not only improves detection performance over the baseline but also provides structured, biologically meaningful explanations.

Graph Neural Networks (GNNs) are deep learning models designed to address the complexities of graph-structured, non-Euclidean data. Due to their complexity, knowledge distillation (KD) is often employed to transfer knowledge from a GNN to a simpler, more efficient student model, such as a Multi-Layer Perceptron (MLP), enabling deployment in large-scale industrial applications. However, KD can inadvertently leak sensitive information from the teacher to the student, posing significant privacy risks. We present the first membership inference attacks targeting GNNs in KD pipeline, showing that student MLPs can reveal whether a node appeared in the teacher’s training data. Our attacks operate in a black-box setting, requiring access only to the student outputs, and remain effective in cross-dataset scenarios. Experimental evaluations across four GNN models and eight datasets show the effectiveness of our approach, achieving up to 0.9014 precision under low FPR of 1% in cross-dataset settings. These results expose significant vulnerabilities in GNN-based KD frameworks, emphasizing the need for strong security measures during the KD process involving GNNs.

The capability to detect radar signals autonomously, without reliance on radar transmitters, is pivotal for the advancement of contemporary shared-spectrum wireless networks like the Citizens Broadband Radio Service (CBRS). Recent trends underscore the integration of AI-driven methodologies to address this challenge effectively. In this paper, we present a novel supervised deep learning framework for radar detection, denoted as Spec-SCAN. We design Spec-SCAN to efficiently identify low-power radar signals amidst interference within a condensed timeframe and over a narrower frequency spectrum compared to existing benchmarks. Our approach employs a YOLO-based training strategy tailored for the detection of radar signals and prevalent interference patterns within the CBRS band. We perform rigorous experiments encompassing scenarios involving LTE, 5G, and DSSS signals as interfering signals to evaluate Spec-SCAN. Our findings indicate that Spec-SCAN attains a radar detection recall of 99% for type 1 radar signals, even at Signal-to-Interference-Noise Ratios (SINR) as low as 15 dB, while scanning a 100MHz spectrum within a 15ms timeframe-demonstrating superior performance compared to alternative methodologies. Spec-SCAN framework also offers comparable performance while scanning 50MHz spectrum for 15 milliseconds.

The Controller Area Network (CAN) has been the de facto in-vehicle network protocol since the 1980s, despite lacking essential security principles like authenticity, confidentiality, integrity, and availability. CAN is especially vulnerable to Denial-of-Service (DoS) attacks, threatening the availability of safety-critical functions. Existing countermeasures have seen limited adoption due to challenges in real-time detection, prevention, and high overhead on Electronic Control Units (ECUs). To address these issues, we propose MichiCAN, a distributed, backward-compatible, real-time defense against DoS and spoofing attacks. MichiCAN leverages integrated/on-chip CAN controllers in modern MCUs, enabling bit-level access to CAN messages. This allows MichiCAN to detect DoS attacks during the arbitration phase and neutralize them by bussing off the attacker ECU swiftly. Experiments on a CAN bus prototype and a real vehicle demonstrate MichiCAN’s effectiveness in enhancing automotive network security.

Vision-language models (VLMs) demonstrate impressive zero-shot and few-shot learning capabilities, making them essential for several downstream tasks. However, fine-tuning these models at scale remains challenging, particularly in federated environments where data is decentralized and non-iid across clients. Existing parameter-efficient tuning methods like LoRA (Low-Rank Adaptation) reduce computational overhead but struggle with heterogeneous client data, leading to suboptimal generalization. To address these challenges, we propose FedVLM, a federated LoRA fine-tuning framework that enables decentralized adaptation of VLMs while preserving model privacy and reducing reliance on centralized training. To further tackle data heterogeneity, we introduce personalized LoRA (pLoRA) which dynamically adapts LoRA parameters to each client’s unique data distribution, significantly improving local adaptation while maintaining global model aggregation. Experiments on the RLAIF-V dataset show that pLoRA improves client-specific performance by 24.5% over standard LoRA, demonstrating superior adaptation in non-iid settings. FedVLM provides a scalable and efficient solution for fine-tuning VLMs in federated settings, advancing personalized adaptation in distributed learning scenarios.

The Internet of Things (IoT) landscape has grown exponentially in recent years, making robust and efficient intrusion detection systems (IDS) even more critical. While Large Language Models (LLMs) have gained significant traction, their effectiveness in network intrusion detection remains largely unexplored. This paper proposes an LLM-based framework for enhanced threat detection and analysis in IoT networks. We explore using advanced LLMs like OpenAI’s Generative Pre-trained Transformer (GPT) model, focusing on techniques such as fine-tuning and embedding similarity. Using real-world intrusion datasets, we evaluate the proposed LLM’s performance in detecting common network attacks and compare it with ensemble-based IDS solutions. We assess the efficiency of the LLM in binary class and multiclass classification task using standard metrics, such as accuracy, recall, precision, and F1 scores. While the fine-tuning approach does not produce comparable results to the current baseline ensemble-based IDS models, the embedding approach, however, yields comparable results. This analysis represents a starting point for exploring the utilization of advanced large language models for intrusion detection within an IoT ecosystem.

Industrial control system (ICS) protocols face the threat of adversaries launching cyber-physical attacks against protocol endpoints. Vulnerability discovery approaches such as fuzzing can be effective at reducing the risk of such threats. In this paper, we present MCFICS, a coverage-guided greybox fuzzing framework that uses (1) active automata learning for stochastic reactive systems to infer the state machine of a stateful ICS protocol server implementation, and (2) guided fuzzing to explore the state space using this learned state machine. During fuzzing, new input sequences that increase code coverage are used to improve the state space exploration of the ICS protocol implementations. We implemented and tested MCFICS with six example server implementations spanning three widely used ICS protocol implementations. Experimental results show that MCFICS achieves higher branch coverage than the AFLNwe, AFLNet and StateAFL fuzzers by an average (mean of means) of 15.82%, 1.99%, and 37.52%, respectively, with an overall average of 18.44% increased branch coverage. Furthermore, using MCFICS we discovered a new bug in a protocol implementation that we have reported to its upstream maintainer.

Named Data Networking (NDN) revolutionized IP-based communication by introducing a content-centric model, based on name-based communication. This paradigm shift offers benefits, including optimized network traffic through in-network caching, improved data security, and resilient communication for Internet of Things (IoT) applications. While these benefits are significant, the deterministic data delivery necessary for time-sensitive IoT applications cannot be guaranteed using the NDN’s best-effort routing mechanism. This paper addresses this challenge by proposing deterministic NDN of things (D-NDNoT), a protocol-level integration of a schedulability algorithm into NDN, making it deadline-aware and addressing the specific requirements of time-sensitive IoT applications. We present a time-sensitive NDN protocol incorporating a critical deadline-first scheduler to prioritize traffic. By integrating deadline awareness, quality of service metrics, and network characteristics, the algorithm ensures the delivery of time-sensitive data takes precedence over non-time-sensitive content. To validate the effectiveness of the proposed protocol, we evaluate using simulation experiments in OMNET++ and consider metrics such as end-to-end latency, delay, and deadline. The results demonstrate that the deadline-aware deterministic NDN protocol effectively meets the communication needs of time-sensitive IoT applications, ensuring the timely delivery of critical data.

The integration of the Internet of Things (IoT) devices and solutions into passenger vehicles has transformed cars into a complex system with intelligence and a platform for extending information technology possibilities. These devices communicate through in-vehicle networks that use the controller area network (CAN) as a de facto standard for the safety-critical functionality of the vehicles. One creative exploit against CAN is the bus-off attack, which uses the fault tolerance capabilities of the CAN bus to coerce a victim electronic control unit (ECU) into the bus-off state from which it is not allowed to access the bus. As a result, the victim ECU is unable to send or receive messages. The WeepingCAN attack is a stealthy variation of the bus-off attack that reduces its observability and therefore the effectiveness of detection-based mitigation. In this paper, we introduce three software-based improvements that greatly increase both the efficiency and effectiveness of the WeepingCAN attack. First, we introduce a novel zero-phase approach for synchronizing the attack. Second, we discover an alternative approach to disable retransmissions, which is a key capability of WeepingCAN, that allows the attack to be conducted from more ECUs than before. Third, we identify a transitive attack strategy that enables an attacker to target many more ECUs than originally possible. We evaluate our improvements experimentally using a CAN benchmark and find that the zero-phase synchronization improves the attack success rate from 75% to over 90% and the transitive attack strategy enables all the ECUs in the benchmark to be attacked.

The Internet of Vehicles (IoV) is envisioned to improve road safety, reduce traffic congestion, and minimize pollution. However, the connectedness of IoV entities increases the risk of cyber attacks, which can have serious consequences. Traditional intrusion detection systems (IDS) transfer large amounts of raw data to central servers, leading to potential privacy concerns. Also, training IDS on resource-constrained IoV devices generally can result in slower training times and poor service quality. To address these issues, we propose a split learning-based privacy-preserving IDS that deploys IDS on edge devices without sharing sensitive raw data. In addition, we propose a regret minimization-based adaptive offloading technique that reduces the training time on resource-constrained devices. Our approach effectively detects anomalous behavior while preserving data privacy and reducing training time, making it a practical solution for IoV. Experimental results show the effectiveness of our approach and its potential to enhance the security of the IoV network.

The Internet of Things presents a transformative era of device connectivity while creating a new paradigm shift in the process. This, however, has been met with some major pitfalls, such as an increase in device insecurity, characterized by Main-in- The Middle (MiTM) attacks. In this paper, we propose IoT-MGSec, a novel solution to mitigating MiTM attacks using graph-based learning. Our approach employs graph modeling and embedding techniques to learn node and edge features such that we can generate a robust classifier to detect MiTM attacks with high accuracy. We validate the effectiveness of our approach by comparing its performance to baseline models, and the results indicate that our approach outperforms the baseline models. The findings suggest that this approach offers a more robust solution to detecting and mitigating Man-in-the-middle attacks, and it holds potential for integration into real-time intrusion detection systems, further enhancing its capacity to secure devices within the loT landscape.

In the Internet of Vehicle ecosystem, multi-access edge computing (MEC) enables mobile nodes to improve their communication and computation capabilities by executing transactions in near real-time. However, the limited energy and computation capabilities of MEC servers limit the efficiency of task computation. Moreover, the use of static edge servers in dense vehicular networks may lead to an influx of service requests that negatively impact the quality of service (QoS) of the edge network. To enhance the QoS and optimize network resources, minimizing offloading computation costs in terms of reduced latency and energy consumption is crucial. In this paper, we propose a cooperative offloading scheme for vehicular nodes, using vehicles as mobile edge servers, which minimizes energy consumption and network delay. In addition, an optimization problem is presented, which is formulated as a Markov Decision Process (MDP). The solution proposed is a deep reinforcement-based Twin Delayed Deep Deterministic policy gradient (TD3), ensuring an optimal balance between task computation time delay and the energy consumption of the system.

Named Data Networking (NDN) has evolved as a networking model that can facilitate Internet of Things (IoT) applications by providing a name-based communication model, innetwork caching, and inherent support for data-centric security. However, despite the benefits, the best-effort NDN cannot offer the deterministic data delivery required by safety-critical IoT applications. This paper proposes a novel deadline-aware NDN protocol that utilizes a critical deadline first scheduler to prioritize traffic based on the approaching deadline. Evaluation results show that the proposed deadline-aware NDN can meet the communication needs of time-sensitive IoT applications.

Ensuring the data integrity of messages transmitted over the Controller Area Network (CAN) bus and other vehicular networks is achieved through the implementation of cryptographic authentication protocols. However, these protocols raise concerns about a significant increase in response time due to the restrictions on CAN frame size and bandwidth. This paper presents a comprehensive analysis of the impact on response time of CAN and CAN Flexible Data-rate (CAN FD) messages with the implementation of cryptographic message authentication codes (MACs) and the periodic transmission of these codes. Our evaluation is based on a randomized schedulability experiment to provide insights into the overhead incurred by adding authentication to the frame payloads.

Controller Area networks (CAN) provide a standard means of communicating across vehicular electronic units without a centralized computing unit or complex dedicated wiring. Despite the benefits offered by in-vehicle networks, CAN networks have been susceptible to network attacks such as replay, fuzzing, and denial of service attacks. In addition, the proliferation of internet-connected vehicles motivates the need to build a robust and secure vehicular network system. Deep learning-based language models such as Bidirectional Encoder Representations from Transformers (BERT) models have proven to produce remarkable results for natural language tasks. BERT models provide a deep understanding of the underlying semantics in textual data. In this paper, we propose CANBERT, a language-based intrusion detection model for CAN bus. We leverage the power of transformer models to provide the detection of malicious attacks on the CAN network. We provide a thorough analysis of our approach using a CAN dataset produced in a realistic driving scenario which consists of a combination of normal data and malicious data from various types of attack scenarios such as Denial of Service (DoS), fuzzy, and impersonation attacks. Our approach is able to detect all of the attacks with high precision and accuracy. In addition, we compare our method with other baseline models and state-of-the-art deep learning intrusion detection approach for in-vehicle networks.

The Internet of Vehicles (IoV) is an active area for innovation and an essential tool in achieving smart cities through the integration of vehicles with the Internet of Things (IoT). IoV is a distributed network that aids in handling the data generated by vehicular sensors and vehicle-to-everything communication (V2X), thus enabling novel applications such as autonomous driving and platooning while increasing safety and energy efficiency. In IoV, the sensors and the interdependent devices relay critical information for the efficient implementation of real-time applications in the ecosystem. Despite all these advancements, a vital challenge is establishing smooth communication among interconnected devices, concretely, interoperability in the IoV—a deceptively simple notion that is not yet fully addressed to achieve a fully integrated ecosystem. This is mainly because the networked domains, such as home, grid, and health care, are developed in silos, operating independently with diverse processes and protocols. Hence, seamless exchange of information is yet to be achieved across the ecosystem, hindering the maximization of the full promise of IoV. In this paper, we provide an in-depth analysis of the present state of interoperability and comprehensively survey the challenges in IoV. We present a taxonomy of interoperability approaches, review solutions that prior work have proposed, and provide insights on how to address the current challenges. Finally, we identify open problems that persist and future directions for research.

The secure functioning of automotive systems is vital to the safety of their passengers and other roadway users. One of the critical functions for safety is the controller area network (CAN), which interconnects the safety-critical electronic control units (ECUs) in the majority of ground vehicles. Unfortunately CAN is known to be vulnerable to several attacks. One such attack is the bus-off attack, which can be used to cause a victim ECU to disconnect itself from the CAN bus and, subsequently, for an attacker to masquerade as that ECU. A limitation of the bus-off attack is that it requires the attacker to achieve tight synchronization between the transmission of the victim and the attacker’s injected message. In this paper, we introduce a schedule-based attack framework for the CAN bus-off attack that uses the real-time schedule of the CAN bus to predict more attack opportunities than previously known. We describe a ranking method for an attacker to select and optimize its attack injections with respect to criteria such as attack success rate, bus perturbation, or attack latency. The results show that vulnerabilities of the CAN bus can be enhanced by schedulebased attacks.

The Internet of Things (IoT) has become commonplace in our lives. From smart refrigerators to thermostats, the heterogeneous connectivity it presents has allowed for the automation of tasks and ease of use. However, it has also led to security challenges introducing new attack vectors that are atypical to traditional computing systems. This also complicates tracing a point of system fault in case of an anomalous event or intrusion. In this paper, we propose a visualization framework that can be used to aid the detection of anomalous system events in an IoT ecosystem. This can assist in uncovering valuable insights from data interactions. The proposed framework provides a visual representation of system events in an IoT device. This can be beneficial for use in digital forensic analysis and uncovering system fault or intrusion. Also, it can be used by system administrators or consumers for situational awareness. We discuss our implementation details using a smart home system as a use case and provide future research directions.

Security of vehicular networks has often been an afterthought since they are designed traditionally to be a closed system. An attack could lead to catastrophic effect which may include loss of human life or severe injury to the driver and passengers of the vehicle. In this paper, we propose a novel algorithm to extract the real-time model of the controller area network (CAN) and develop a specification-based intrusion detection system (IDS) using anomaly-based supervised learning with the real-time model as input. We evaluate IDS performance with real CAN logs collected from a sedan car.