The rapid advancement of Large Language Models (LLMs) has driven significant progress in Natural Language Interface to Database (NLIDB). However, the widespread adoption of LLMs has raised critical privacy and security concerns. During interactions, LLMs may unintentionally expose confidential database contents or be manipulated by attackers to exfiltrate data through seemingly benign queries. While current efforts typically rely on rule-based heuristics or LLM agents to mitigate this leakage risk, these methods still struggle with complex inference-based attacks, suffer from high false positive rates, and often compromise the reliability of SQL queries. To address these challenges, we propose SafeNLIDB, a novel privacy-security alignment framework for LLM-based NLIDB. The framework features an automated pipeline that generates hybrid chain-of-thought interaction data from scratch, seamlessly combining explicit security reasoning with SQL generation. Additionally, we introduce reasoning warm-up and alternating preference optimization to overcome the multi-preference oscillations of Direct Preference Optimization (DPO), enabling LLMs to produce security-aware SQL through fine-grained reasoning without the need for human-annotated preference data. Extensive experiments demonstrate that our method outperforms both larger-scale LLMs and ideal-setting baselines, achieving significant security improvements while preserving high utility.

Despite some promising results in federated learning using game-theoretical methods, most existing studies mainly employ a one-level game in either a cooperative or competitive environment, failing to capture the complex dynamics among participants in practice. To address this issue, we propose DualGFL, a novel federated learning framework with a dual-level game in cooperative-competitive environments. DualGFL includes a lower-level hedonic game where clients form coalitions and an upper-level multi-attribute auction game where coalitions bid for training participation. At the lower-level DualGFL, we introduce a new auction-aware utility function and propose a Pareto-optimal partitioning algorithm to find a Pareto-optimal partition based on clients' preference profiles. At the upper-level DualGFL, we formulate a multi-attribute auction game with resource constraints and derive equilibrium bids to maximize coalitions' winning probabilities and profits. A greedy algorithm is proposed to maximize the utility of the central server. Extensive experiments on real-world datasets demonstrate DualGFL's effectiveness in improving both server utility and client utility.

To address the limitations of traditional computational fluid dynamics (CFD) simulations, such as high computational cost, long processing times, and limited scalability, this study identifies the inefficiencies of existing data‐driven prediction methods, which often lack spatial–temporal coordination mechanisms and fail to capture fine‐grained dynamic features of UAV airflow fields. We propose a novel deep learning model, VAN‐ConvLSTM, for rapid and accurate prediction of UAV downwash airflow. Unlike conventional ConvLSTM‐based frameworks, which struggle with modeling long‐range dependencies and detailed spatial variations, our model introduces a visual attention unit (VAN) to enhance spatiotemporal sensitivity. The model architecture combines a convolutional encoder for spatial feature extraction, a VAN module for attention‐guided temporal modeling, and a ConvLSTM decoder for sequence generation. This synergistic design improves both the accuracy and interpretability of airflow prediction. Experimental results show that the VAN‐ConvLSTM model achieves an SSIM score of 0.96, demonstrating high consistency with CFD simulations. Compared to baseline methods, our model reduces error while improving stability and spatial fidelity. Ablation studies further validate the individual contributions of VAN and ConvLSTM modules. The results, verified through three representative case studies, confirm that VAN‐ConvLSTM outperforms state‐of‐the‐art approaches across multiple evaluation metrics, while offering significantly enhanced computational efficiency. This demonstrates its strong potential as a reliable and scalable alternative to traditional CFD methods in rotor airflow prediction scenarios.

The advent of ubiquitous computing devices in the Internet of Things (IoT) has resulted in an explosion of data. Traditional centralized machine learning models face challenges including limited bandwidth in wireless environments and privacy concerns due to their data aggregation approach. Federated learning addresses these challenges via decentralizing model training across numerous devices, leveraging model updates to enhance privacy and reduce communication overhead. To improve its cost efficiency, current research focuses on minimizing either time or energy costs but rarely both, and does not jointly optimize the parameters of device and training scheduling in the presence of system and data heterogeneity inherent in IoT networks. In our paper, we first introduce a multi-group transmission scheme and propose a comprehensive device scheduling framework, Group Scheduling on Orthogonal Frequency-Division Multiple Access (GS-OFDMA), to address time bottlenecks. Then we formulate a joint optimization problem for device and training scheduling that minimizes the total cost of training while ensuring model convergence. To tackle the resulting mixed integer nonlinear programming problem, we develop an iterative algorithm. Experimental results show that our approach significantly reduces the total cost by at least 35% across various real-world datasets and data distributions in comparison with random participant selection. The proposed GS-OFDMA protocol also exhibits higher time efficiency over other device scheduling schemes.

Federated learning enables decentralized model training across numerous devices without data centralization, leveraging model updates to enhance privacy and reduce communication overhead. Despite its advantages, federated learning systems must be optimized for cost efficiency, considering the limited computational capabilities and battery life of edge devices. Current research often focuses on minimizing either time or energy costs but rarely both, and does not jointly optimize the parameters of device and training scheduling in the presence of system and data heterogeneity. In our paper, we formulate a novel joint optimization problem for device and training scheduling that minimizes the total cost of federated learning while ensuring model convergence. We propose a new device scheduling scheme, Group Scheduling on Orthogonal Frequency-Division Multiple Access (GS-OFDMA), to improve time efficiency and develop an iterative algorithm to tackle the resulting mixed integer nonlinear programming problem. Our experimental results show that our approach significantly reduces the total cost by at least 35 % across different real-world datasets and data distributions in comparison with random participant selection.

Federated learning is a distributed machine learning paradigm that allows multiple edge devices to collaboratively train a shared model without exchanging raw data. However, the training efficiency of federated learning is highly dependent on client selection. Moreover, due to the varying wireless communication environments and various computation latencies among the clients, selecting clients randomly or uniformly may not be optimal for balancing the data diversity and training efficiency. In this article, we formulate a new latency-minimization problem that simultaneously optimizes client selection and training procedures in federated learning, which takes into account the data and latency heterogeneity among the clients. Given the nonconvexity of the problem, we derive a new convergence upper bound for federated learning with probabilistic client selection. To solve the mixed integer nonlinear programming problem, we introduce a hybrid solution that integrates grid search techniques with the polyhedral active set algorithm. Numerical analyses and experiments on real-world data demonstrate that our scheme outperforms the existing ones in terms of overall training latency and achieves up to three times acceleration over random client selection, especially in scenarios with highly heterogeneous data and latencies among the clients.

Deep Neural Networks (DNNs) tend to go deeper and wider, which poses a significant challenge to the training of DNNs, due to the limited memory capacity of DNN accelerators. Existing solutions for memory-efficient DNN training are densely coupled with the application features of DNN workloads, e.g., layer structures or computational graphs of DNNs are necessary for these solutions. This would result in weak versatility for DNNs with sophisticated layer structures or complicated computation graphs. These schemes usually need to be re-implemented or re-adapted due to the new layer structures or the unusual operators in the computational graphs introduced by these DNNs. In this article, we review the memory pressure issues of DNN training from the perspective of runtime systems and model the memory access behaviors of DNN workloads. We identify the iterative, regularity , and extremalization properties of memory access patterns for DNN workloads. Based on these observations, we propose AppObMem, an application-oblivious memory scheduling system. AppObMem automatically traces the memory behaviors of DNN workloads and schedules the memory swapping to reduce the memory pressure of the device accelerators without the perception of high-level information of layer structures or computation graphs. Evaluations on a variety of DNN models show that, AppObMem obtains 40–60% memory savings with acceptable performance loss. AppObMem is also competitive with other open sourced SOTA schemes.

As deep neural networks (DNNs) continue to find applications in ever more domains, the exact nature of the neural network architecture becomes an increasingly sensitive subject, due to either intellectual property protection or risks of adversarial attacks. While prior work has explored aspects of the risk associated with model leakage, exactly which parts of the model are most sensitive and how one infers the full architecture of the DNN when nothing is known about the structure a priori are problems that have been left unexplored. In this paper we address this gap, first by presenting a schema for reasoning about model leakage holistically, and then by proposing and quantitatively evaluating DeepSniffer, a novel learning-based model extraction framework that uses no prior knowledge of the victim model. DeepSniffer is robust to architectural and system noises introduced by the complex memory hierarchy and diverse run-time system optimizations. Taking GPU platforms as a showcase, DeepSniffer performs model extraction by learning both the architecture-level execution features of kernels and the inter-layer temporal association information introduced by the common practice of DNN design. We demonstrate that DeepSniffer works experimentally in the context of an off-the-shelf Nvidia GPU platform running a variety of DNN models and that the extracted models significantly improve attempts at crafting adversarial inputs. The DeepSniffer project has been released inhttps://github.com/xinghu7788/DeepSniffer.

The graph convolutional network (GCN) emerges as a promising direction to learn the inductive representation in graph data commonly used in widespread applications, such as E-commerce, social networks, and knowledge graphs. However, learning from graphs is nontrivial because of its mixed computation model involving both graph analytics and neural network computing. To this end, we decompose the GCN learning into two hierarchical paradigms: 1) graph-level and 2) node-level computing. Such a hierarchical paradigm facilitates the software and hardware accelerations for GCN learning. We propose a lightweight graph reordering methodology, incorporated with a GCN accelerator architecture that equips a customized cache design to fully utilize the graph-level data reuse. We also propose a mapping methodology aware of data reuse and task-level parallelism to handle various graphs inputs effectively. The results show that Rubik accelerator design improves energy efficiency by$26.3\times $–$1375.2\times $than GPU platforms across different datasets and GCN models.

The training of deep neural networks (DNNs) is usually memory-hungry due to the limited device memory capacity of DNN accelerators. Characterizing the memory behaviors of DNN training is critical to optimize the device memory pressures. In this work, we pinpoint the memory behaviors of each device memory block of GPU during training by instrumenting the memory allocators of the runtime system. Our results show that the memory access patterns of device memory blocks are stable and follow an iterative fashion. These observations are useful for the future optimization of memory-efficient training from the perspective of raw memory access patterns.

Adversarial patch attacks that craft the pixels in a confined region of the input images show their powerful attack effectiveness in physical environments even with noises or deformations. Existing certified defenses towards adversarial patch attacks work well on small images like MNIST and CIFAR-10 datasets, but achieve very poor certified accuracy on higher-resolution images like ImageNet. It is urgent to design both robust and effective defenses against such a practical and harmful attack in industry-level larger images. In this work, we propose the certified defense methodology that achieves high provable robustness for high-resolution images and largely improves the practicality for real adoption of the certified defense. The basic insight of our work is that the adversarial patch intends to leverage localized superficial important neurons (SIN) to manipulate the prediction results. Hence, we leverage the SIN-based DNN compression techniques to significantly improve the certified accuracy, by reducing the adversarial region searching overhead and filtering the prediction noises. Our experimental results show that the certified accuracy is increased from 36.3% (the state-of-the-art certified detection) to 60.4%on the ImageNet dataset, largely pushing the certified defenses for practical use.

In the original compression tracking algorithm, the size of the tracking box is fixed. There should be better tracking results for scale-invariant objects, but worse tracking results for scale-variant objects. To overcome this defect, a scale-adaptive compressive tracking (CT) algorithm is proposed. First of all, the imbalance of the gray and texture features in the original CT algorithm is balanced by the multi-feature method, which makes the algorithm more robust. Then, searching different candidate regions by using the method of multi-scale search along with feature normalization makes the features extracted from different scales comparable. Finally, the candidate region with the maximum discriminate degree is selected as the object region. Thus, the tracking-box size is adaptive. The experimental results show that when the object scale changes, the improving CT algorithm has higher accuracy and robustness than the original CT algorithm.

Current Transport Layer Security (TLS) Public-Key Infrastructure (PKI) is a vast and complex system; it consists of processes, policies, and entities that are responsible for a secure certificate management process. Among them, Certificate Authority (CA) is the central and most trusted entity. However, recent compromises of CA result in the desire for some other secure and transparent alternative approaches. To distribute the trust and mitigate the threats and security issues of current PKI, publicly verifiable log-based approaches have been proposed. However, still, these schemes have vulnerabilities and inefficiency problems due to lack of specifying proper monitoring, data structure, and extra latency. We propose Accountable and Transparent TLS Certificate Management: an alternate Public-Key Infrastructure (PKI) with verifiable trusted parties (ATCM) that makes certificate management phases; certificate issuance, registration, revocation, and validation publicly verifiable. It also guarantees strong security by preventing man-in-middle-attack (MitM) when at least one entity is trusted out of all entities taking part in the protocol signing and verification. Accountable and Transparent TLS Certificate Management: an alternate Public-Key Infrastructure (PKI) with verifiable trusted parties (ATCM) can handle CA hierarchy and introduces an improved revocation system and revocation policy. We have compared our performance results with state-of-the-art log-based protocols. The performance results and evaluations show that it is feasible for practical use. Moreover, we have performed formal verification of our proposed protocol to verify its core security properties using Tamarin Prover.

Physical layer (PHY) security has aroused great research interest in recent years, exploiting physical uncertainty of wireless channels to provide communication secrecy without placing any computational restrictions on the adversaries under the information-theoretic security model. Particularly, researches have been focused on investigating Wyner's Wiretap Channel for constructing practical wiretap codes that can achieve simultaneous transmission secrecy and reliability. While theoretically sound, PHY security through the wiretap channel has never been realized in practice, and the feasibility and physical limitations of implementing such channels in the real world are yet to be well understood. In this paper, we design and implement a practical opportunistic secret communication system over the wireless wiretap channel for the first time to our best knowledge. We show that, our system can achieve nearly perfect secrecy given a fixed codeword length by carefully controlling the structure of the parity-check matrix of wiretap codes to strike the proper balance between the transmission rate and secrecy. Our system is implemented and evaluated extensively on a USRP N210-based testbed. The experimental results demonstrate the physical limitations and the feasibility of building practical wiretap channels in both the worst channel case and the case where the sender has only the knowledge of instantaneous channel capacities. Our system design and implementation successfully attempts towards bridging the gap between the theoretical wiretap channel and its practice, alleviating the unrealistic and strong assumptions imposed by the theoretical model.