Abstract Generative agents are a novel AI agent architecture developed in 2023 using LLMs that can generate believable human behaviors. They are of significant importance in social simulation and human-computer interaction. However, there is a lack of research on the privacy and security issues that may arise from its subsequent application in society. Based on the Generative Agent framework, this paper reintroduces a personal information system, and a shared database, reconstructing the memory and reflection systems to realize a Lightweight Generative Agent framework (LGA). To fill the research gap in the field of privacy for Generative Agents, this paper simulates three scenarios with five agents to explore potential privacy and security issues. Based on observed privacy leaks, three defensive strategies are proposed: behavior guideline regulation, Multi-Agents cooperation, and step-back prompts. These strategies have reduced privacy leaks by more than 50.00% on average. For some privacy leaks with special simulation scenarios, it has even reduced privacy leaks by more than 80.00%. Additionally, through supplementary experiments, we demonstrate that LGA achieves better cost-effectiveness compared to the original Generative Agent framework. The analysis of privacy and security issues provides a reference for subsequent research on the safety of generative agents.

Abstract Wi-Fi networks have become a fundamental component of Internet of Things (IoT) environments, while their open and shared nature also exposes them to a wide range of cyber attacks. This study examines the use of time-series feature engineering combined with classical machine learning techniques for multiclass Wi-Fi intrusion detection using the AWID3 dataset. Network traffic is segmented into multivariate time-series blocks to capture temporal characteristics of wireless communication. From these segments, two complementary feature representations are derived: statistical descriptors that support interpretability and CNN-based features that capture spatial and temporal patterns. The proposed framework is evaluated using K-Nearest Neighbors, Support Vector Machines, XGBoost, and ensemble voting classifiers across 24 experimental configurations, considering different sequence lengths and feature extraction strategies. The experimental results indicate that classical machine learning models, particularly XGBoost combined with statistical time-series features, achieve strong performance in a 14-class intrusion detection task, with accuracy and F1-score exceeding 0.98. These findings demonstrate that carefully designed feature representations, when paired with well-established classifiers, can provide an effective and computationally efficient solution for practical Wi-Fi intrusion detection scenarios.

Abstract Non-interactive zero-knowledge (NIZK) proofs of knowledge have proven to be highly relevant for securely realizing a wide array of applications that rely on both privacy and correctness . They enable a prover to convince any party of the correctness of a public statement for a secret witness . However, most NIZKs do not natively support proving knowledge of a secret witness that is distributed over multiple provers. Previously, collaborative proofs [54] have been proposed to overcome this limitation. We investigate the notion of composability in this setting, following the Commit-and-Prove design of LegoSNARK [19]. Composability allows users to combine different, specialized NIZKs (e.g., one for arithmetic circuits, one for boolean circuits, and one for range proofs) with the aim of reducing the proof generation time. Moreover, it opens the door to efficient realizations of many applications in the collaborative setting such as mutually exclusive prover groups, combining collaborative and single-party proofs and efficiently implementing publicly auditable secure multiparty computing (PA-MPC). We present the first, general definition for collaborative commitand- prove NIZK (CP-NIZK) proofs of knowledge and construct MPC protocols to enable their realization. We implement our protocols for two commonly used NIZKs, Groth16 and Bulletproofs, and evaluate their practicality in a variety of computational settings. Our findings indicate that composability adds only minor overhead, especially for large circuits. We also evaluated our construction in two application settings, one of which shows 18– $$55\times $$ 55 × runtime reduction when compared to prior works while requiring only a fraction ( $$0.2\%$$ 0.2 % ) of the communication.

Abstract Zero Trust is an approach allowing for increased security by providing an object or a subject with the three CIA (Confidentiality, Integrity, Availability) security aspects. To comply with the CIA criteria, access control models need to support functionalities such as: a) safer permission grant and authorization processes, b) policy decision delivery to single or multiple users, and c) policy decision delivery to single or multiple actions or objects. In addition, we need to consider redundancy, conflict detection, different types of permissions to delegate, delegation, and the separation of duties (SoD) with different types. Extensive literature exists with respect to delegation operations on access control models, but most of them do not consider redundancy or partial conflict detection with regard to the standard policies. We address the positive and negative policies resolution as a precursor to the delegation request resolution. We address the resolutions in context of the standard policies that allow or deny an action on the object to a single or multiple subjects. We provide an analysis via multiple case studies using a Python implementation of the HPol (Hierarchical Policy) model. Our analysis demonstrates the ability of the HPol model to handle access control resolution issues discussed, with proof of results in context of the positive and negative (YES & NO) policy requests.

Cyber-physical systems (CPS) combine computational, communication, and physical components to enable real-time monitoring and control in critical infrastructures. Despite their advantages, CPS are highly susceptible to Denial-of-Service (DoS) attacks that disrupt communication and impair state estimation. This study conducted a comprehensive analysis of CPS defense under DoS attacks using learning automata (LA) within a game-theoretic framework. Sensor–attacker interactions were modeled as a two-player zero-sum game, in which the sensor sought to minimize estimation error and communication cost, while the attacker aimed to maximize disruption. Both discrete (DLA) and continuous (CLA) learning automata were employed to adaptively optimize sensor strategies and were integrated with a Kalman filter to achieve accurate state estimation. Simulations in MATLAB/Simulink, including generic CPS and microgrid systems, were performed to evaluate performance under reliable and unreliable channels, as well as under varying attack frequencies and durations. The results demonstrated that CLA achieved smooth convergence and high-accuracy state estimation under stable conditions, whereas DLA adapted more rapidly to abrupt disturbances and dynamic environments. Analyses of attack patterns confirmed the framework’s capacity to maintain system resilience through adaptive strategy allocation. Overall, the study demonstrated that learning automata provided an effective, real-time approach for optimizing CPS defense, balancing estimation accuracy, operational cost, and security, and offered a flexible solution applicable to other cyber threats.

Abstract In multi-user cooperative systems such as social networks, personal data is often jointly created and shared among multiple users. The sensitivity of such data depends on the preferences and relationships of all parties involved, making access control decisions inherently complex and dynamic. This complexity is further exacerbated because such data often forms compound objects, such as photos with multiple tagged users or comments, where access to one object can affect access to related objects. Traditional access control models lack the expressiveness needed to capture joint ownership, evolving social relationships, and time-dependent constraints, which can lead to privacy violations and unintended disclosures. In this work, we propose a fine-grained access control model for multi-user cooperative systems and apply it to social networks. Our model extends attribute-based access control with provenance information to enforce additional constraints and explicitly models compound objects to reflect the interrelated nature of social data. A key contribution is the introduction of temporal constraints in access decision-making, enabling dynamic authorizations based on time-sensitive conditions. We implemented a prototype of the proposed model and conducted an experimental evaluation to assess its feasibility. Our results show that incorporating temporal constraints has minimal impact on performance, demonstrating the practicality of our approach in existing social network environments.