Network intrusion detection is a crucial line of defense for protecting network security. Despite the significant advancements made by deep learning in this field, existing methods are primarily based on closed-set classification and are ineffective in detecting unknown attacks. To address this research gap, we propose an open-set recognition-based network intrusion detection method. We first provide a network traffic classification model based on open-set recognition, OpenPN , to classify known classes of network traffic and recognize unknown network traffic. Then we introduce a novel attack detection algorithm involving expert intervention, which reduces manual costs through expert verification and utilizes a density-based k-reciprocal nearest neighbor clustering algorithm for optimization. Finally, we perform continuous learning for the classes that have been verified as novel attacks. Extensive experiments conducted on three public datasets demonstrate that the proposed method outperforms existing methods in both closed-set classification and open-set recognition. In addition, the impact of each critical parameter on the performance of the relevant algorithms is comprehensively analyzed.

Recent advances in diffusion generative models have yielded remarkable progress. While the quality of generated content continues to improve, these models have grown considerably in size and complexity. This increasing computational burden poses significant challenges, particularly in resource-constrained deployment scenarios such as mobile devices. The combination of model pruning and knowledge distillation has emerged as a promising solution to reduce computational demands while preserving generation quality. However, this technique inadvertently propagates undesirable behaviors, including the generation of copyrighted content and unsafe concepts, even when such instances are absent from the fine-tuning dataset. In this paper, we propose a novel bilevel optimization framework for pruned diffusion models that consolidates the fine-tuning and unlearning processes into a unified phase. Our approach maintains the principal advantages of distillation—namely, efficient convergence and style transfer capabilities—while selectively suppressing the generation of unwanted content. This plug-in framework is compatible with various pruning and concept unlearning methods, facilitating efficient, safe deployment of diffusion models in controlled environments. Code is available here.

Shuffling-type gradient methods are favored in practice for their simplicity and rapid empirical performance. Despite extensive development of convergence guarantees under various assumptions in recent years, most require the Lipschitz smoothness condition, which is often not met in common machine learning models. We highlight this issue with specific counterexamples. To address this gap, we revisit the convergence rates of shuffling-type gradient methods without assuming Lipschitz smoothness. Using our stepsize strategy, the shuffling-type gradient algorithm not only converges under weaker assumptions but also match the current best-known convergence rates, thereby broadening its applicability. We prove the convergence rates for nonconvex, strongly convex, and non-strongly convex cases, each under both random reshuffling and arbitrary shuffling schemes, under a general bounded variance condition. Numerical experiments further validate the performance of our shuffling-type gradient algorithm, underscoring its practical efficacy.

We study cost-aware routing for large language models across diverse and dynamic pools of models. Existing approaches often overlook prompt-specific context, rely on expensive model profiling, assume a fixed set of experts, or use inefficient trial-and-error strategies. We introduce Cost-Spectrum Contrastive Routing (CSCR), a lightweight framework that maps both prompts and models into a shared embedding space to enable fast, cost-sensitive selection. CSCR uses compact, fast-to-compute logit footprints for open-source models and perplexity fingerprints for black-box APIs. A contrastive encoder is trained to favor the cheapest accurate expert within adaptive cost bands. At inference time, routing reduces to a single $k$‑NN lookup via a FAISS index, requiring no retraining when the expert pool changes and enabling microsecond latency. Across multiple benchmarks, CSCR consistently outperforms baselines, improving the accuracy–cost tradeoff by up to 25\%, while generalizing robustly to unseen LLMs and out-of-distribution prompts.

Fine-tuning pre-trained Large Language Models (LLMs) for downstream tasks using First-Order (FO) optimizers presents significant computational challenges. Parameter-Efficient Fine-Tuning~(PEFT) methods have been proposed to address these challenges by freezing most model parameters and training only a small subset. While PEFT is efficient, it may not outperform full fine-tuning when high task-specific performance is required. Zeroth-Order (ZO) methods offer an alternative for fine-tuning the entire pre-trained model by approximating gradients using only the forward pass, thus eliminating the computational burden of back-propagation, % in first-order methods, but they converge painfully slowly and are very sensitive to the choice of task prompts. We bridge these worlds with Bilevel‑ZOFO, a penalty‑based bilevel formulation that treats adapter parameters as a lower‑level learner coupled to an upper‑level ZO optimizer of the full backbone. This double-loop optimization strategy only requires the gradient of the PEFT model and the forward pass of the base model. We provide theoretical convergence guarantees for Bilevel ZOFO. Empirically, we demonstrate that Bilevel-ZOFO significantly outperforms existing ZO methods, achieves 2–4$\times$ faster training, and reduces sensitivity to prompts. Bilevel-ZOFO also outperforms FO PEFT methods while maintaining similar memory efficiency. Additionally, we show its strong potential for meta learning.

Millimeter-wave (mmWave) communication inside high-speed railway (HSR) intrawagon is essential for future high-data-rate and low-latency onboard services. Accurate and efficient channel modeling is essential for system design and evaluation. Ray-tracing (RT) techniques are capable of accurately capturing significant multipath components (MPCs). However, RT becomes complex and time-consuming in harsh environments, where dense multipath components (DMCs) exist and the reverberation effects occur. To address this issue, this paper analyzes the mmWave propagation channel inside HSR intrawagon and proposes a Quasi-Deterministic (QD) modeling approach. Channel measurements and RT simulations were conducted at the 26 GHz band to track major specular reflection paths and reveal the underlying mmWave propagation mechanisms. By combining the accuracy of RT with the computational efficiency of stochastic models, the proposed method accurately characterizes dominant MPCs and efficiently models DMCs. Experimental results demonstrate that the measured data aligns well with the proposed QD model. The proposed approach improves computational efficiency while maintaining modeling accuracy.

Bilevel optimization problems appear in many widely used machine learning tasks. Bilevel optimization models are sensitive to small changes, and bilevel training tasks typically involve limited datasets. Therefore, overfitting is a common challenge in bilevel training tasks. This paper considers the use of dropout to address this problem. We propose a bilevel optimization model that depends on the distribution of dropout masks. We investigate how the dropout rate affects the hypergradient of this model. We propose a dropout bilevel method to solve the dropout bilevel optimization model. Subsequently, we analyze the resulting dropout bilevel method from an optimization perspective. Analyzing the optimization properties of methods with dropout is essential because it provides convergence guarantees for methods using dropout. However, there has been limited investigation in this research direction. We provide the complexity of the resulting dropout bilevel method in terms of reaching an $\epsilon$ stationary point of the proposed stochastic bilevel model. Empirically, we demonstrate that overfitting occurs in data cleaning problems, and the method proposed in this work mitigates this issue.

Recently, DDoS attack has developed rapidly and become one of the most important threats to the Internet. Traditional machine learning and deep learning methods can-not train a satisfactory model based on the data of a single client. Moreover, in the real scenes, there are a large number of devices used for traffic collection, these devices often do not want to share data between each other depending on the research and analysis value of the attack traffic, which limits the accuracy of the model. Therefore, to solve these problems, we design a DDoS attack detection model based on federated learning named FLDDoS, so that the local model can learn the data of each client without sharing the data. In addition, considering that the distribution of attack detection datasets is extremely imbalanced and the proportion of attack samples is very small, we propose a hierarchical aggregation algorithm based on K-Means and a data resampling method based on SMOTEENN. The result shows that our model improves the accuracy by 4% compared with the traditional method, and reduces the number of communication rounds by 40%.

Configuration errors are among the dominant causes of system faults for the industrial automation and control systems (IACS). It is difficult to detect and correct such errors of IACS as there are various kinds of systems and devices with miscellaneous configuration specifications. In this article, we first propose a streaming algorithm to keep all the configuration changes in the limited memory space. When making a new configuration change, another novel streaming algorithm is proposed to search and return all the similar historical changes, which can be used to validate this new one. So far, we are the first to model the configuration changes of IACS as a data stream and apply the streaming similarity search in correcting configuration errors while overcoming the inherent unbounded-memory bottleneck. The theoretical correctness and complexity analyses are presented. Experiments with real and synthetic datasets confirm the theoretical analyses and demonstrate the effectiveness of the proposed method in preventing misconfigurations of IACS.