Existing provenance graph-based advanced persistent threat (APT) attack detection methods mainly focues on identifying isolated attack events and fail to capture the temporal correlations and causal dependencies among multi-stage attack events. To address this issue, the problem of APT attack path reasoning was investigated, which aimed to aggregate related attack events belonging to the same APT campaign into a complete attack chain, and an attack tactics and techniques knowledge-driven APT attack path reasoning method was proposed. Specifically, the proposed method first constructed an anomaly subgraph containing isolated attack events through anomaly detection, attack tactics and techniques identification, and graph pruning, then introduced an ATT&CK-based tactic-technique sequence pattern built from threat intelligence to guide the attack path reasoning, and finally reconstructed complete APT attack chains by integrating graph search with a threat scoring mechanism. Experimental results on a simulated attack dataset collected from kernel logs and the public DARPA TC dataset demonstrate that under the premise of maintaining attack chain integrity, the proposed method improves the reconstruction precision by over 60% compared with existing methods.

Advanced Persistent Threats (APTs) remain difficult to detect due to their stealthy nature and long-term persistence. To tackle this challenge, provenance-based threat hunting has gained traction as a proactive defense mechanism. This technique models audit logs as a whole-system provenance graph and searches for subgraphs that match APT patterns recorded in Cyber Threat Intelligence (CTI) reports. However, several limitations persist: 1) significant memory and time overhead due to the extremely large provenance graphs; 2) imprecise segmentation of APT activities from provenance graphs due to their intricate entanglement with benign operations; and 3) poor alignment of attack representations between CTI-derived query graphs and provenance graphs due to their substantial semantic gaps. To address these limitations, this paper presents ProHunter, an efficient and accurate provenance-based APT hunting system with a platform-independent design. To minimize system overhead, ProHunter creates a compact data structure that efficiently stores long-term provenance graphs using semantic abstraction and bit-level hierarchical encoding strategies. To segment APT behaviors, a heuristic-driven threat graph sampling algorithm is designed, which can extract precise attack patterns from provenance graphs. Furthermore, to bridge the semantic gaps between CTI-derived graphs and provenance graphs, ProHunter proposes adaptive graph representation and feature enhancement methods, enabling the extraction of consistent attack semantics at both localized and globalized levels.Extensive evaluations on real-world APT campaigns from DARPA TC E3, E5 and OpTC datasets demonstrate that ProHunter outperforms state-of-the-art threat hunting systems in terms of efficiency and accuracy. Our code is available at https://github.com/xueboQiu/ProHunter.

Advanced Persistent Threats (APTs) inflict formidable damage on national security and critical infrastructure owing to their covert and sustained nature. Data provenance, which systematically correlates discrete audit logs into a structured representation of system activities (known as a provenance graph), has emerged as a pivotal technique in intrusion detection. Constrained by the scarcity of APT datasets, current provenance-based research predominantly relies on unsupervised anomaly detection approaches. Nonetheless, they suffer from limited semantic expressiveness in modeling system behaviors, mainly attributable to overly simplistic and monotonous pre-training objectives, which culminates in suboptimal detection performance.In this paper, we introduce PRO-TRAP, a provenance-based APT detection system that leverages a domain-adaptive architectural integration of multi-scale contrastive learning and heuristic provenance graph augmentation. First, a multi-scale contrastive learning framework is proposed to capture behavior features across node, edge, and graph levels, enabling the distillation of discriminative representations for detection. Second, to expand the boundaries of contrastive learning within the context of provenance graphs, we formulate a heuristic graph augmentation strategy that capitalizes on their intrinsic temporal dynamics to generate semantic-consistent augmented views. Third, we customize a hierarchical anomaly detection pipeline that initiates with coarse-grained identification of suspicious graphs, followed by fine-grained localization of anomalous nodes, thereby efficiently furnishing actionable intelligence for security analysts. Extensive experiments on benchmark datasets substantiate that PRO-TRAP outperforms state-of-the-art APT detection systems in both detection precision and computational efficiency. Specifically, it achieves 99.70% precision with a low 0.3% false positive rate on StreamSpot and 96.33% precision with a 3.6% false positive rate on Unicorn. Across the realistic DARPA TC dataset, our method delivers an average precision improvement of 5.49% and reduces false alarms by over 50% on average. Code is available at: https://github.com/xueboQiu/PRO-TRAP.

With the widespread adoption of microservice architectures, the security threats they face have become increasingly sophisticated. Existing anomaly detection methods based on system calls exhibit significant limitations in three key aspects: multidimensional data fusion, temporal causality modeling, and forensic analysis of anomalies. This paper proposes MADGuard, a provenance graph-based anomaly detection system for microservices. MADGuard addresses these challenges through three key innovations: (1) It constructs a native provenance graph by integrating multisource services and multidimensional data, employing feature hashing and positional encoding for efficient graph representation; (2) The system introduces a Temporal Graph Network (TGN) model combined with edge reconstruction error and Inverse Document Frequency (IDF) weighting, achieving a 15. 07% improvement in the F1 score compared to existing methods; (3) For the first time in microservice security, an integrated forensic analysis module is implemented, allowing rapid anomaly path reconstruction through aggregated anomaly subgraphs. Comprehensive evaluations on typical microservice benchmarks (TeaStore, RobotShop, SockShop) demonstrate MADGuard’s superior performance: 94.08% detection accuracy, significantly outperforming state-of-the-art approaches while maintaining practical operational efficiency.

Geomagnetic navigation is a passive technique that leverages the spatial distribution of the Earth’s magnetic field to mitigate the susceptibility of Global Navigation Satellite Systems (GNSS) to external interference and reduce cumulative errors in inertial navigation systems, thereby ensuring robust stability. However, its relatively low accuracy has historically limited practical deployment. To address this challenge, this paper proposes an aircraft geomagnetic navigation method via dual-view feature extraction and hybrid multi-criteria adaptive weighting (DHAGN). DHAGN extracts features from two distinct views, adaptively adjusts feature weights using both standard-deviation-based and summation-based criteria, and integrates an loss-feedback mechanism within the summation-based weighting to further enhance navigation accuracy. Experiments on 13 flight routes from the SGL2020 dataset demonstrate that DHAGN achieves an average distance-root-mean-square (DRMS) error reduction of 45.5 meters compared to the state-of-the-art Magnav2C method, validating its effectiveness in enhancing geomagnetic navigation accuracy and facilitating practical implementation.

Provenance graphs offer critical insights into system execution traces for intrusion detection. Recent work predominantly applies Graph Neural Networks (GNNs) to model causal system behaviors in provenance graphs. However, two fundamental limitations constrain their practicality: 1) uniform aggregation mechanisms that fail to account for heterogeneous behavior patterns; and 2) local interaction modeling that struggles to capture long-range dependencies essential for multi-stage attack analysis. To tackle these issues, we propose PG-MoE, a pioneering Provenance Graph Mixture-of-Experts framework tailored for intrusion detection. PG-MoE integrates an ensemble of specialized GNN experts, each designed to handle distinct topological contexts. Through a learnable routing mechanism, nodes are dynamically dispatched to appropriate experts based on their structural and semantic characteristics, thereby enabling adaptive modeling of diverse system behaviors. Furthermore, we introduce a spatio-temporal contrastive learning strategy that enforces the alignment of node behavior representations across multiple temporal and contextual (spatio) resolutions, thereby empowering experts to capture both local and global interaction dynamics while discerning subtle behavior variations. Extensive experiments on three real-world attack datasets demonstrate that PG-MoE outperforms state-of-the-art baselines, exhibits strong robustness against adversarial perturbations, and maintains computational efficiency. The source code is available at https://github.com/xueboQiu/PG-MoE/.

In recent years, Living off the Land (LotL) attacks have been drawing attention due to their flexibility and difficulty in detection. These attacks exploit legitimate tools already in the system to conduct malicious activities, hiding their malicious intent behind normal benign programs. However, detection methods for such attacks largely rely on expert rules. While rule tags can effectively detect known attacks, this also leads to a high false positive rate, resulting in low detection accuracy for the models. To address these issues, we propose a detection system called LOTLDetector, which combines deep learning methods with expert rules to detect malicious command lines in LotL attacks from both data and knowledge perspectives. LOTLDetector learns the semantics of command line text through neural networks and combines rule tags from expert knowledge, enabling a more comprehensive detection of LotL attacks. We extensively evaluated our method, validated it on a Windows dataset containing 27,448 command lines and a Linux dataset containing 27,093 command lines, and compared it with existing methods. The results show that our method significantly outperforms existing methods in detecting malicious command lines. For the Linux dataset, the detection system achieved a detection performance with an accuracy of 0.9728; for the Windows dataset, the system's detection accuracy also reached 0.9598, which is about 8% higher than the best existing method. In addition, our project has been open-sourced at https://github.com/csedikaf/LOTLDetector.

With the rapid proliferation of social networks and the ubiquitous use of portable imaging devices, massive volumes of visual data are generated and shared daily. Much of this data contains personal or sensitive information, heightening the risk of privacy breaches and data misuse. Consequently, visual privacy detection has become a critical area of research at the intersection of computer vision and information security. This review provides a comprehensive examination of visual privacy detection technologies, tracing developments from pixels to privacy, from low-level image features to high-level semantic understanding. It first explores the conceptual foundations of visual privacy and identifies the lack of unified measurement standards as a key barrier to progress. The review then systematically analyzes major privacy-related datasets and categorizes existing detection approaches based on visual content, feature representation, multimodal information integration, and vision–language model-based reasoning. Furthermore, practical applications of visual privacy detection are discussed across diverse domains, including social media platforms, surveillance systems, urban transportation, targeted advertising, and augmented/mixed reality (AR/MR). The review also highlights persistent challenges such as uncertainty in privacy perception, dataset imbalance, annotation subjectivity, and the trade-off between privacy protection and data utility. Finally, it outlines prospective research directions aimed at building robust, generalizable, and ethically responsible visual privacy frameworks.

As Advanced Persistent Threats (APTs) become more complex and destructive, attack investigation has gained importance. Analysts use provenance graphs for causality analysis on Point-Of-Interest (POI) events to capture critical events. However, existing methods suffer from problems of high false positives, high overhead, and high latency due to the vast size of the provenance graph and the rarity of critical events. We propose SParse for constructing critical component graphs (i.e., consisting of critical events) from streaming logs in real time. Our approach is based on two key observations: 1) Critical events exist in suspicious semantic graphs (SSGs) composed of interaction flows between suspicious entities, and 2) Information flows accomplishing the attacker's goal exist as paths. SParse uses a two-stage framework that first constructs the SSG using a state-based mode with semantic transfer rules and storage strategies. Then, it identifies suspicious flow paths (SFPs) related to the POI event and quantifies each path's influence to filter irrelevant events. Evaluation on a large-scale attack dataset shows our system generates a critical component graph (similar to 113 edges) in 1.6 seconds, which is 2014 & times; smaller than the backtracking graph (similar to 227,589 edges). It is also 25 & times; more effective in filtering irrelevant edges compared to other state-of-the-art techniques.

Federated recommendation systems (FedRecs) enable multiple clients to collaboratively train recommendation models without directly sharing private user data, providing a promising solution for privacy protection. However, recent studies reveal that the collaborative mechanism in FedRecs still faces serious privacy leakage risks. To systematically assess these vulnerabilities, this article proposes CiCo-MIA, a Causal inference-based Collaborative Membership Inference Attack designed to quantify the privacy risks of FedRecs through a shadow recommendation model. CiCo-MIA utilizes interaction data from multiple clients to train a shadow model that mirrors the behavior of the target FedRecs. By constructing a causal graph and introducing counterfactual reasoning, we develop a user-item impact elimination module that enhances the robustness and interpretability of the attack. Additionally, a dynamic aggregation strategy is proposed to balance attack effectiveness and recommendation performance. Extensive experiments on six real-world datasets under two FedRec frameworks demonstrate that CiCo-MIA achieves state-of-the-art performance, improving F1 scores by about 20% and HR@10 by at least 20% compared with baseline attacks. Even under effective defense mechanisms such as local differential privacy, the attack remains effective, maintaining around 30% of its original performance. These results highlight that FedRecs are still vulnerable to collaborative inference attacks, emphasizing the urgent need for stronger privacy-preserving defenses.

The inherent complexity of microservices makes them prone to cascading failures, yet locating the precise root cause remains a critical challenge. Existing multimodal methods often fuse logs, metrics, and traces through coarse feature concatenation, which can introduce semantic interference. More over, although some recent studies have advanced from service level localization to metric-level localization, they usually rely on static correlation or causal models and thus struggle to capture dynamic fault propagation. We present MicroInspect, a progressive framework for metric-level root cause localization in microservices. MicroInspect first disentangles multimodal features to reduce semantic interference and dynamically fuses them with a Mixture-of-Experts gate adapted to fault propagation stages, supporting accurate anomaly detection and service-level localization. It then employs a Dynamic Neural Granger Causal ity Network (DNGCN), guided by anomaly detection signals, to infer time-varying metric dependencies and progressively refine root causes from faulty services to metric-level causes. Evalu ation on four public benchmarks (TrainTicket, SocialNetwork, FAMOS-Mall, and GAIA) shows that MicroInspect achieves the best overall performance across anomaly detection, service level localization, and metric-level root cause localization, with statistically significant improvements in F1-score and HR@1 under paired t-tests (p < 0.05).

Data provenance transforms audit logs into provenance graphs that encode intricate dependencies among system entities, forming the analytical foundation for causal reasoning and intrusion investigation. Recent progress in self-supervised representation learning has enabled the modeling of behavioral semantics within provenance graphs for specific downstream tasks. However, such specialized models generalize poorly across heterogeneous provenance problems and demand costly retraining to adapt to emerging scenarios.To bridge this gap, we propose UniProv, a pioneering unified pretraining framework that learns general provenance graph representations transferable across diverse downstream tasks. First, UniProv synergizes a suite of self-supervised learning tasks to extract multi-faceted and complementary semantic insights from topological structure and attributes of provenance graphs. Second, a causality-preserving provenance graph augmentation algorithm is proposed, which generates augmented graph views via heuristic perturbations to strengthen the learning efficacy. Third, Bayesian uncertainty is introduced to dynamically calibrate multi-objective learning priorities, thereby distilling shared behavioral semantics into general representations. Extensive experiments on three provenance tasks across six datasets validate both the generalizability and effectiveness of UniProv. Notably, its F1-score outperforms state-of-the-art models by 4.43% and 3.54% on average for graph- and node-level intrusion detection. Moreover, UniProv exhibits low training overhead and high scalability, with performance steadily improving as the quantity of training data increases. Our code is available at https://github.com/xueboQiu/UniProv.

Advanced persistent threat (APT) attacks are characterized by high stealthiness and long-lasting persistence, and provenance graphs constructed from kernel audit logs have become an important technique for APT attack detection. However, existing provenance graph-based methods mainly focus on identifying isolated attack events and fail to capture the temporal correlations and causal dependencies among multi-stage attack behaviors, resulting in a large number of disconnected suspicious events that greatly increase the difficulty of security analysis and decision-making for administrators. To address this issue, this paper investigates the problem of APT attack path reasoning, which aims to aggregate related attack events belonging to the same APT campaign into a complete attack chain, and proposes a tactic and technique knowledge-driven APT attack path reasoning approach. Specifically, the proposed method first constructs an anomaly subgraph containing isolated attack events through anomaly detection, attack tactic and technique identification, and graph pruning, then introduces an ATT&CK-based tactic-technique sequence pattern built from threat intelligence to guide the attack path reasoning, and finally reconstructs complete APT attack chains by integrating graph search with a threat scoring mechanism. Experimental results on a simulated attack dataset collected from kernel logs and the public DARPA TC dataset demonstrate that the proposed method achieves a 99.9% reduction in graph scale while preserving 91% of critical attack events, and under the premise of maintaining attack chain integrity, improves the reconstruction precision by over 60% compared with existing methods.

Adversarial attacks on machine learning-based malware detection models could enable attackers to craft evasive malware variants that remain functional while bypassing detection. However, existing methods primarily focus on feature-space adversarial attacks, lacking effective mapping from feature-space to code-space (i.e. the modifications of source code of malware to implement the adversarial attacks), which limits its practical application. To address this issue, this paper proposes an large language model (LLM) enhanced code-space adversarial attack method against the dynamic malware detection models, achieving end-to-end generation of executable and source code-level adversarial samples. First, we apply a feature-space adversarial attack algorithm to perturb the feature vectors and pinpoint the changes of dynamic behaviors (e.g. API calls). Second, we propose a prompt mechanism guided by the dynamic behavior changes and code generation hints to automatically modify the source code of malware samples by leveraging an LLM. Experiments on five types of typical malware demonstrate that the proposed method can effectively modify the source code of malware based on the adversarial perturbations, and consistently outperforms existing approaches in terms of evasion capability and functionality retention.

Efficient and reliable data transmission is vital for ensuring Quality of Service (QoS) in edge vehicular networks, especially during emergencies requiring low-latency communication. However, high mobility and unpredictable road conditions make it difficult to maintain stable routes, limiting the effectiveness of conventional routing protocols. This paper proposes FuzzyForward, a novel fuzzy logic-based data forwarding framework tailored to the dynamic and uncertain nature of edge VANETs. Unlike traditional methods, FuzzyForward uses fuzzy reasoning to evaluate imprecise traffic metrics, such as connection quality and turning fitness, for accurate selection of road segments and relay vehicles. Another key contribution is its two-phase fuzzy optimisation process, supported by a relay candidate reduction strategy to minimise overhead and improve responsiveness. Simulation results demonstrate that FuzzyForward significantly enhances packet delivery, reduces end-to-end delay, and lowers communication overhead.

Provenance-based threat hunting identifies Advanced Persistent Threats (APTs) on endpoints by correlating attack patterns described in Cyber Threat Intelligence (CTI) with provenance graphs derived from system audit logs. A fundamental challenge in this paradigm lies in the modality gap-the structural and semantic disconnect between provenance graphs and CTI reports. Prior work addresses this by framing threat hunting as a graph matching task: 1) extracting attack graphs from CTI reports, and 2) aligning them with provenance graphs. However, this pipeline incurs severe information loss during graph extraction and demands intensive manual curation, undermining scalability and effectiveness. In this paper, we present APT-CGLP, a novel cross-modal APT hunting system via Contrastive Graph-Language Pre-training, facilitating end-to-end semantic matching between provenance graphs and CTI reports without human intervention. First, empowered by the Large Language Model (LLM), APT-CGLP mitigates data scarcity by synthesizing high-fidelity provenance graph-CTI report pairs, while simultaneously distilling actionable insights from noisy web-sourced CTIs to improve their operational utility. Second, APT-CGLP incorporates a tailored multi-objective training algorithm that synergizes contrastive learning with inter-modal masked modeling, promoting cross-modal attack semantic alignment at both coarse- and fine-grained levels. Extensive experiments on four real-world APT datasets demonstrate that APT-CGLP consistently outperforms state-of-the-art threat hunting baselines in terms of accuracy and efficiency.

Advanced persistent threats (APTs) are stealthy and multi-stage, making single-point defenses (e.g., malware- or traffic-based detectors) ill-suited to capture long-range and cross-entity attack semantics. Provenance-graph analysis has become a prominent approach for APT detection. However, its practical deployment is hampered by (i) the scarcity of APT samples, (ii) the cost and difficulty of fine-grained APT sample labeling, and (iii) the diversity of attack tactics and techniques. Aiming at these problems, this paper proposes APT-MCL, an intelligent APT detection system based on Multi-view Collaborative provenance graph Learning. It adopts an unsupervised learning strategy to discover APT attacks at the node level via anomaly detection. After that, it creates multiple anomaly detection sub-models based on multi-view features and integrates them within a collaborative learning framework to adapt to diverse attack scenarios. Extensive experiments on three real-world APT datasets validate the approach: (i) multi-view features improve cross-scenario generalization, and (ii) co-training substantially boosts node-level detection under label scarcity, enabling practical deployment on diverse attack scenarios.

The evolution of Internet of Vehicles (IoV) technologies, encompassing wireless communications and Artificial Intelligence (AI), has advanced the collaborative “Pedestrian-Vehicle-Road-Cloud” IoV edge services, enhancing road efficiency and driving safety. Operating in an open-edge environment with vast sensory data, IoV faces significant privacy risks from unauthorized access and data breaches. Consequently, privacy-preserving computation (PPC) is crucial for secure IoV services. This paper reviews PPC techniques in IoV edge services, exploring network characteristics and potential privacy attacks. It categorizes and evaluates techniques such as differential privacy, homomorphic encryption, and secure multi-party computation based on data security, utility, and overhead. Summarizing their pros and cons, the challenges and future research directions for IoV edge services are outlined.

Advanced Persistent Threats (APTs) commonly manifest through a sequence of attack steps, known as Tactics, Techniques, and Procedures (TTPs). Recent studies identify TTPs by converting audit logs into causal provenance graphs and applying expert-driven mappings that correlate low-level system events with high-level TTP patterns. However, these methods face persistent challenges: determining the impact boundaries of TTP activities, adapting to evolving TTP stacks, and recognizing fine-grained TTP semantics for deeper forensic insights. To address these challenges, we present Zoomer, a novel TTP recognition framework that segments provenance graphs into multiple TTP subgraphs with multi-granular annotations (i.e., tactics, techniques, and sub-techniques). First, we devise a heuristic subgraph sampling algorithm guided by anomalous node detection to precisely delineate the scope of TTP activities. Second, we introduce a dual-tower Deep & Wide architecture that integrates contextual behavior semantics from provenance graphs and domain-informed features to learn expressive TTP representations. Finally, we adopt a prototypical network that reformulates TTP recognition as a few-shot pattern matching task, thereby enhancing adaptability and accuracy under limited supervision. To advance future research, we built and released the first TTP-annotated provenance dataset, encompassing the most comprehensive collection of TTP instances to date. Extensive experiments show that Zoomer achieves TTP recognition with 88% accuracy at the sub-technique level and 94% at the tactic level, significantly outperforming state-of-the-art baselines.

Detecting advanced persistent threats (APTs) at a host via data provenance has emerged as a valuable yet challenging task. Compared with attack rule matching, machine learning approaches offer new perspectives for efficiently detecting attacks by leveraging their inherent ability to autonomously learn from data and adapt to dynamic environments. However, the scarcity of APT samples poses a significant limitation, rendering supervised learning methods that have demonstrated remarkable capabilities in other domains (e.g., malware detection) impractical. Therefore, we propose a system called TAGAPT, which is able to automatically generate numerous APT samples with provenance-level granularity. First, we introduce a deep graph generation model to generalize various graph structures that represent new attack patterns. Second, we propose an attack stage division algorithm to divide each generated graph structure into stage subgraphs. Finally, we design a genetic algorithm to find the optimal attack technique explanation for each subgraph and obtain fully instantiated APT samples. Experimental results demonstrate that TAGAPT can learn from existing attack patterns and generalize to novel attack patterns. Furthermore, the generated APT samples 1) exhibit the ability to help with efficient threat hunting and 2) provide additional assistance to the state-of-the-art (SOTA) attack detection system (Kairos) by filtering out 73% of the observed false positives. We have open-sourced the code and the generated samples to support the development of the security community.