Machine learning’s widespread application across various fields has raised significant privacy concerns, making Membership Inference Attacks (MIAs) a key technique for assessing model privacy vulnerabilities. This paper focuses on improving the LDC-MIA method and proposes the LDC-MIA+ framework. By incorporating a model output stability metric and integrating multiple features such as loss values, calibrated losses, and neighbor cosine similarities, we construct a more effective MIA classifier. Experimental results on CIFAR-10, Adult, and Credit datasets demonstrate that, compared to other MIA techniques, our method significantly enhances the True Positive Rate (TPR) at extremely low False Positive Rates (FPR), with notable improvements in balanced accuracy and Area Under the Curve (AUC). It more accurately infers membership information and exhibits strong robustness across diverse datasets and models, providing a more reliable framework for assessing the privacy risks of machine learning models.

Test-time adaptation (TTA) has gained increasing popularity due to its efficacy in addressing "distribution shift" issue while simultaneously protecting data privacy. However, most prior methods assume that a paired source domain model and target domain sharing the same label space coexist, heavily limiting their applicability. In this paper, we investigate a more general source model capable of adaptation to multiple target domains without needing shared labels. This is achieved by using a pre-trained vision-language model (VLM), e.g., CLIP, that can recognize images through matching with class descriptions. While the zero-shot performance of VLMs is impressive, they struggle to effectively capture the distinctive attributes of a target domain. To that end, we propose a novel method - Context-aware Language-driven TTA (COLA). The proposed method incorporates a lightweight context-aware module that consists of three key components: a task-aware adapter, a context-aware unit, and a residual connection unit for exploring task-specific knowledge, domain-specific knowledge from the VLM and prior knowledge of the VLM, respectively. It is worth noting that the context-aware module can be seamlessly integrated into a frozen VLM, ensuring both minimal effort and parameter efficiency. Additionally, we introduce a Class-Balanced Pseudo-labeling (CBPL) strategy to mitigate the adverse effects caused by class imbalance. We demonstrate the effectiveness of our method not only in TTA scenarios but also in class generalisation tasks. The source code is available at https://github.com/NUDT-Bai-Group/COLA-TTA.

Robust models must demonstrate strong generalizability, even amid environmental changes. However, the complex variability and noise in real-world data often lead to a pronounced performance gap between the training and testing phases. Researchers have recently introduced test-time-domain adaptation (TTA) to address this challenge. TTA methods primarily adapt source-pretrained models to a target domain using only unlabeled test data. This study found that existing TTA methods consider only the largest logit as a pseudo-label and aim to minimize the entropy of test time predictions. This maximizes the predictive confidence of the model. However, this corresponds to the model being overconfident in the local test scenarios. In response, we introduce a novel confidence-calibration loss function called Calibrated Entropy Test-Time Adaptation (CETA), which considers the model's largest logit and the next-highest-ranked one, aiming to strike a balance between overconfidence and underconfidence. This was achieved by incorporating a sample-wise regularization term. We also provide a theoretical foundation for the proposed loss function. Experimentally, our method outperformed existing strategies on benchmark corruption datasets across multiple models, underscoring the efficacy of our approach.

This paper delves into the confidence calibration in prediction when using Graph Neural Networks (GNNs), which has emerged as a notable challenge in the field. Despite their remarkable capabilities in processing graph-structured data, GNNs are prone to exhibit lower confidence in their predictions than what the actual accuracy warrants. Recent advances attempt to address this by minimizing prediction entropy to enhance confidence levels. However, this method inadvertently risks leading to over-confidence in model predictions. Our investigation in this work reveals that most existing GNN calibration methods predominantly focus on the highest logit, thereby neglecting the entire spectrum of prediction probabilities. To alleviate this limitation, we introduce a novel framework called Balanced Calibrated Graph Neural Network (BCGNN), specifically designed to establish a balanced calibration between over-confidence and under-confidence in GNNs' prediction. To theoretically support our proposed method, we further demonstrate the mechanism of the BCGNN framework in effective confidence calibration and significant trustworthiness improvement in prediction. We conduct extensive experiments to examine the developed framework. The empirical results show our method's superior performance in predictive confidence and trustworthiness, affirming its practical applicability and effectiveness in real-world scenarios.

The security of Deep Neural Networks (DNNs) has proven to be critical for their applicabilities in real-world scenarios. However, DNNs are well-known to be vulnerable against adversarial attacks, such as adding artificially designed imperceptible magnitude perturbation to the benign input. Therefore, adversarial robustness is essential for DNNs to defend against malicious attacks. Stochastic Neural Networks (SNNs) have recently shown effective performance on enhancing adversarial robustness by injecting uncertainty into models. Nevertheless, existing SNNs are still limited for adversarial defense, as their insufficient representation capability from the fixed uncertainty. In this paper, to elevate feature representation capability of SNNs, we propose a novel yet practical stochastic neural network that maximizes feature distribution variance (MFDV-SNN). In addition, we provide theoretical insights to support the adversarial resistance of MFDV, which primarily derived from the stochastic noise we injected into DNNs. Our research demonstrates that by gradually increasing the level of stochastic noise in a DNN, the model naturally becomes more resistant to input perturbations. Since adversarial training is not required, MFDV-SNN does not compromise clean data accuracy and saves up to 7.5 times computation time. Extensive experiments on various attacks demonstrate that MFDV-SNN improves adversarial robustness significantly compared to other methods.

This article introduces a command and control (C2) organizational structure model, which can analyze military C2 organization structure from three perspectives: organizational elements, organizational relationships, and organizational processes. In the organizational processes, a decision-making problem of recommending course of action (COA) is described. Then, a planning method is proposed to solve this decision-making problem, which includes six evaluation metrics of COA effectiveness and a tabu search algorithm implemented within the C2 organization structure model. On this basis, we provide two application examples to demonstrate the usability of the proposed C2 organizational structure model and its COA planning method for operational decision-making.

Deep Neural Networks (DNNs) are known to be vulnerable to adversarial attacks. Recently, Stochastic Neural Networks (SNNs) have been proposed to enhance adversarial robustness by injecting uncertainty into the models. However, existing SNNs often inspired by intuition and rely on adversarial training, which is computationally costly. To address this issue, we propose a novel SNN called the Weight-based Stochastic Neural Network (WB-SNN), which is based on optimizing an error upper bound of adversarial robustness from the perspective of weight distribution. To the best of our knowledge, we are the first to propose a theoretically guaranteed weight-based stochastic neural network without relying on adversarial training. In comparison to normal adversarial training, our method saves about three times the computation cost. Extensive experiments on various datasets, networks, and adversarial attacks have demonstrated the effectiveness of the proposed method.

The vulnerability of deep learning algorithms to malicious attack has garnered significant attention from researchers in recent years. In order to provide more reliable services for safety-sensitive applications, prior studies have introduced Stochastic Neural Networks (SNNs) as a means of improving adversarial robustness. However, existing SNNs are not designed from the perspective of optimizing the adversarial decision boundary and rely on complex and expensive adversarial training. To find an appropriate decision boundary, we propose a simple and effective stochastic neural network that incorporates a regularization term into the objective function. Our approach maximizes the variance of the feature distribution in low-dimensional space and forces the feature direction to align with the eigenvectors of the covariance matrix. Due to no need of adversarial training, our method requires lower computational cost and does not sacrifice accuracy on normal examples, making it suitable for use with a variety of models. Extensive experiments against various well-known white- and black-box attacks show that our proposed method outperforms state-of-the-art methods.

Set query is a fundamental problem in computer systems. Plenty of applications rely on the query results of membership, association, and multiplicity. A traditional method that addresses such a fundamental problem is derived from Bloom filter. However, such methods may fail to support element deletion, require additional filters or apriori knowledge, making them unamenable to a high-performance implementation for dynamic set representation and query. In this paper, we envision a novel sketch framework that is multi-functional, non-parametric, space efficient, and deletable. As far as we know, none of the existing designs can guarantee such features simultaneously. To this end, we present a general shifting framework to represent auxiliary information (such as multiplicity, association) with the offset. Thereafter, we specify such design philosophy for a hash table horizontally at the slot level, as well as vertically at the bucket level. Theoretical and experimental results jointly demonstrate that our design works exceptionally well with three types of set queries under small memory.

The growing expansion of data availability in medical fields could help improve the performance of machine learning methods. However, with healthcare data, using multi-institutional datasets is challenging due to privacy and security concerns. Therefore, privacy-preserving machine learning methods are required. Thus, we use a federated learning model to train a shared global model, which is a central server that does not contain private data, and all clients maintain the sensitive data in their own institutions. The scattered training data are connected to improve model performance, while preserving data privacy. However, in the federated training procedure, data errors or noise can reduce learning performance. Therefore, we introduce the self-paced learning, which can effectively select high-confidence samples and drop high noisy samples to improve the performances of the training model and reduce the risk of data privacy leakage. We propose the federated self-paced learning (FedSPL), which combines the advantage of federated learning and self-paced learning. The proposed FedSPL model was evaluated on gene expression data distributed across different institutions where the privacy concerns must be considered. The results demonstrate that the proposed FedSPL model is secure, i.e. it does not expose the original record to other parties, and the computational overhead during training is acceptable. Compared with learning methods based on the local data of all parties, the proposed model can significantly improve the predicted F1-score by approximately 4.3%. We believe that the proposed method has the potential to benefit clinicians in gene selections and disease prognosis.

As an important model of machine learning, Bayesian networks (BNs) have received a lot of attentions since they can be used for classification via probabilistic inference. However, since it is a complicated combination optimization problem, BN structure learning cannot be solved with classic convex optimization algorithms. Hence, evolutionary algorithms provide an alternative way to find a global solution to BN structure learning problem. In this paper, we improve the biased random-key genetic algorithm to solve the BN structure learning problem. Meanwhile, we apply a local optimization model as its decoder to improve the performance of the proposed algorithm. Finally, we conduct our experiments on nine benchmark networks and a real dataset of cross-site scripting (XSS) attack. Experimental results show that the proposed algorithm can obtain more accurate solutions than other state-of-the-art algorithms and achieve a good performance in XSS attack detection for web security.

Cuckoo filter (CF) and its variants are emerging as replacements of Bloom filters in various networking and distributed systems to support efficient set representation and membership testing. Cuckoo filters store item fingerprints directly with two candidate buckets and a reallocation scheme is implemented to mitigate the bucket overflow problem for higher space utilization. Such a reallocation scheme, once triggered, however, can be time-consuming. This shortcoming makes the existing CFs not applicable for insertion-intensive scenarios such as online applications wherein the items join and leave frequently. To this end, in this paper, we propose the Vertical Cuckoo filter (VCF) which extends the standard Cuckoo filter by providing more candidate buckets to each item. Another challenging issue with such a design is how to ensure that the candidate buckets can be indexed by each other such that no additional hash computation and item access are necessary during fingerprint reallocation. Therefore, we present the vertical hashing, which indexes the candidate buckets with the fingerprint and given bitmasks. We further generalize and improve the VCF by realizing$k$(≥ 4) candidate buckets and avoiding unnecessary computation. The comprehensive experiments indicate that VCF outperforms its same kinds in terms of space utilization and insertion throughput, with a slight compromise of lookup speed.

Stochastic neural network (SNN) has attracted increasing attention in recent years, which benefits several important tasks by modeling samples uncertainly, such as adversarial defense, label noise robustness, and model calibration. The current implementations of existing stochastic neural networks are mainly Gaussian noise injection, e.g., deep Variational Information Bottleneck (VIB) uses fixed Gaussian prior to derive noise injection, simple and effective stochastic neural network (SE-SNN) uses a non-informative Gaussian prior to implement it. However, Gaussian distribution assumption is insufficient to model more complex distributions of data in practical, such as the skewed distribution or multi-modal distribution. In this paper, we relax the strict Gaussian prior assumption, and propose a novel distribution calibrated stochastic neural network (DCSNN) which integrates two successive steps. These two steps are as follows: 1) The trained feature vector is preprocessed to make its feature distribution closer to the Gaussian-like distribution. 2) Gaussian distribution’s mean and variance are used to model the sample’s activation indeterminacy. The experimental results show that, compared with the existing methods, our proposed method can achieve state-of-the-art results in a variety of datasets, backbone architectures and multiple applications.

Class imbalance is a widely existed and challenging problem in real-world applications such as disease diagnosis, fraud detection, network intrusion detection and so on. Due to the scarce of data, it could significantly deteriorate the accuracy of classification. To address this challenge, we propose a novel Imbalanced Data Augmentation Generative Adversarial Networks (GAN) named IDA-GAN as an augmentation tool to deal with the imbalanced dataset. This is a great challenge because it is hard to train a GAN model under this situation. We address this issue by coupling variational autoencoder along with GAN training. In this paper, specifically, we introduce the variational autoencoder to learn the majority and minority class distributions in the latent space, and use the generative model to utilize each class distribution for the subsequent GAN training. The generative model learns useful features to generate target minority-class samples. Compared with the state-of-the-art GAN model, the experimental results demonstrate that our proposed IDA-GAN could generate more diverse minority samples with better qualities, and it could benefits the imbalanced classification task in terms of several widely-used evaluation metrics on five benchmark datasets: MNIST, Fashion-MNIST, SVHN, CIFAR-10 and GTSRB.

Training gene expression data with supervised learning approaches can provide an alarm sign for early treatment of lung cancer to decrease death rates. However, the samples of gene features involve lots of noises in a realistic environment. In this study, we present a random forest with self-paced learning bootstrap for improvement of lung cancer classification and prognosis based on gene expression data. To be specific, we propose an ensemble learning with random forest approach to improving the model classification performance by selecting multi-classifiers. Then, we investigate the sampling strategy by gradually embedding from high- to low-quality samples by self-paced learning. The experimental results based on five public lung cancer datasets show that our proposed method could select significant genes exactly, which improves classification performance compared to that of existing approaches. We believe that our proposed method has the potential to assist doctors in gene selections and lung cancer prognosis.

Stealth malware, a representative tool of advanced persistent threat (APT) attacks, in particular poses an increased threat to cyber-physical systems (CPS). Due to the use of stealthy and evasive techniques (e.g., zero-day exploits, obfuscation techniques), stealth malwares usually render conventional heavyweight countermeasures (e.g., exploits patching, specialized ant-malware program) inapplicable. Light-weight countermeasures (e.g., containment techniques), on the other hand, can help retard the spread of stealth malwares, but the ensuing side effects might violate the primary safety requirement of CPS. Hence, defenders need to find a balance between the gain and loss of deploying light-weight countermeasures. To address this challenge, we model the persistent anti-malware process as a shortest-path tree interdiction (SPTI) Stackelberg game, and safety requirements of CPS are introduced as constraints in the defender's decision model. Specifically, we first propose a static game (SSPTI), and then extend it to a multi-stage dynamic game (DSPTI) to meet the need of real-time decision making. Both games are modelled as bi-level integer programs, and proved to be NP-hard. We then develop a Benders decomposition algorithm to achieve the Stackelberg Equilibrium of SSPTI. Finally, we design a model predictive control strategy to solve DSPTI approximately by sequentially solving an approximation of SSPTI. The extensive simulation results demonstrate that the proposed dynamic defense strategy can achieve a balance between fail-secure ability and fail-safe ability while retarding the stealth malware propagation in CPS.

Graphical models are widely used in cyber security analysis to capture relationships among variables in attack scenarios. However, most models are difficult to build due to the greatly imbalanced data in cyber attacks. To solve this problem, we propose a two-level Bayesian network framework in this paper. We firstly classify the cyber attacks into two levels, one is used to identify general types of attacks and the other is used to classify specific forms. Then we train Bayesian networks for each level. To help administrators cope with threats in time, we propose an analysis method. This method finds the important node's Markov blanket and sorts nodes in it by their influence on each specific form, which could help to understand key threats in cyberspace.

In this paper we propose a novel method for modeling the shape of vessel trajectories in a manner which may facilitate the application of machine learning techniques. This is achieved by transforming the topological feature of vessel trajectories into vectors. More specifically, we calculate scale-invariance indicators for every vessel trajectory as shape characteristics, and other indicators to denote the trajectory area. The proposed method is validated using both synthetic trajectories and real-world AIS datasets. We demonstrate that it can achieve good time efficiency and may support vessel trajectory related analysis.

Key element protection of combat system is a challenging problem in modern combat. Effectively evaluating the key elements would be of great help in force deployment in crisis situations. This paper proposes a new evaluation method that combines expert evaluation, PCA (Principal Component Analysis) and TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution). Specifically, we first use expert evaluation to get attributes' values of each element. Then, we apply PCA to extract principal components from elements' attributes. Next, we use TOPSIS to calculate elements' proximities to ideal solutions under different principal components. Finally, we synthesize each element's proximity under different principal component to get the relative superiority. In addition, we test the proposed method in an experimental combat scenario, and show the plausibility of this method.