The emergence and potential of large-scale quantum computing is a threat to existing cryptographic infrastructures but also a powerful tool for optimization. Yet, the near-term utility of quantum devices is severely limited due to noise, decoherence, and lack of confidentiality in untrusted cloud execution environments. In this paper, the authors present an algorithm named Noise-Adaptive Secure Optimization (NASA), which is a single approach that solves the crucial intersection between large-scale optimization, real-time noise adaptation, and cryptographic confidentiality of post-quantum security architectures. NASA combines four core innovations: (1) a real-time noise profiling module, which is a continuous characterization of device decoherence times, gate error rates and readout fidelities; (2) a reinforcement learning agent, which dynamically selects optimal adaptation strategies, including pulse-level corrections, Hamiltonian remapping and ansatz switching, based on instantaneous noise conditions; (3) a lightweight quantum one-time pad confidentiality layer, which obfuscates the circuit structure and measurement outcomes with minimal runtime overhead; and (4) a noise-aware objective function that balances energy minimization with fidelity estimates to ensure robust optimization. The algorithm is validated using extensive simulations and hardware experiments on standard MaxCut benchmarks and cryptanalytically-relevant Learning with Errors (LWE) instances. Results show that the success rates of the algorithms (86% on MaxCut, 79% LWE at n=10) are superior for the algorithms developed by the National Aeronautics and Space Administration (NASA) compared to non-adaptive baselines, the circuit fidelity is high (0.89), and the additional runtime overhead of the algorithms is 18% while they deliver 128 bits of cryptographic security. Furthermore, the number of qubits that can be processed in the benchmarks is scaled by the size of the problem (up to 35 qubits) that is much larger than any baseline, proving that noise-adaptive techniques are required to get meaningful results from NISQ devices in security-sensitive applications.

Localization is essential in Wireless Sensor Networks (WSNs) for communication protocols such as geographic routing and applications including person tracking, battlefield monitoring, and environmental surveillance. This paper proposes an improved localization method called SGOL (Social Group Optimization for Node Localization), which combines the Social Group Optimization (SGO) algorithm with K-means clustering to minimize the error between estimated and real positions of unknown sensors. To address the collinearity problem of anchor nodes, K-means clustering selects geometrically well-distributed anchor nodes. A weighted fitness function assigns different precision values to original anchor nodes versus estimated ones obtained through iterative localization stages. A multi-stage iterative process enables isolated nodes (those with fewer than three visible anchors) to be localized using previously localized nodes as reference points. Extensive simulations over 50 independent runs with 100 nodes, 15 anchors, and 5% noise show that SGOL achieves an average localization error of 0.870 meters, representing improvements of 6.5%, 11.2%, 31.1%, and 75.1% compared to Jaya NL-WSN [1], PSO [2], FOA-L [3], and CSO [4], respectively. SGOL also achieves a higher localization rate than all the compared algorithms due to its multi-stage architecture. Notably, K-means clustering becomes increasingly effective as network size grows, contributing up to 19% error reduction in large-scale deployments.

Security of IoT devices is a critical concern due to the widespread use of interconnected devices. These devices, ranging from smart thermostats to security cameras, often lack robust security measures, creating vulnerabilities that malicious actors can exploit. The primary risk is the potential for unauthorized access, allowing attackers to take control of devices, steal sensitive data, and use them to launch network attacks. In this paper, an ideal framework is proposed to detect and classify malware in IoT networks by analyzing traffic features using optimized Machine Learning (ML) algorithms. The recent IoT-23 dataset is applied to evaluate the model’s performance. Traditional approaches often struggle with handling large datasets and accurately categorizing attack traffic. The proposed model not only detects malware but also classifies it into multiple categories. Additionally, upon detecting malware, an alert is generated, ensuring that timely action can be taken for its prevention and mitigation. This enhances security by enabling swift responses to potential threats. The proposed model detects malware with an impressive accuracy of 99%. This exceptional performance and early-stage notification demonstrate the model's efficiency in minimizing the risks of malware and further prevention of threats.

The goal of influence maximization is to find a group of seed users who can increase the spread of information in social networks. Most current influence maximization methods focus on maximizing influence spread, often assuming that all social relationships are trustworthy. However, in reality, social networks are sensitive to trust. Users may trust, distrust, or even reject information depending on how trustworthy they see its source. Influence maximization techniques that neglect trust and distrust relationships risk propagating untrustworthy influence, which can be detrimental to overall performance. Therefore, this paper proposes a Trust-Aware Deep Reinforcement Learning (TA-DRL) framework for influence maximization in signed social networks. The social network is represented as a directed and weighted trust graph, where edges encode trust relationships. A trust-aware Graph Neural Network (GNN) is employed to learn node representations via trust-weighted message passing. The learned node representations serve as inputs to a deep reinforcement learning agent that sequentially selects seed nodes under a fixed budget constraint. The influence maximization problem is formulated as a Markov Decision Process (MDP), and a trust-aware reward function is designed to maximize trustworthy influence while minimizing negative influence. Extensive experiments are conducted on four real-world signed social network datasets: Epinions, Slashdot Zoo, Wiki-RfA, and Bitcoin-Alpha. The experimental results confirm that the proposed TA-DRL framework outperforms existing methods in terms of influence maximization, trust-aware influence maximization, and negative influence suppression. Runtime and scalability analyses further verify the effectiveness of the framework in achieving a favorable trade-off between performance and computational efficiency.

The recent increase in the number of Internet of Things (IoT) devices has vastly increased the cyber-attack surface of the latest networks. Even though intrusion detection using machine learning (ML) has demonstrated encouraging results in terms of detection accuracy, most of the current methodologies are based on computationally intensive or black box models that are not transparent and do not fit the resource-limited IoT worlds. Furthermore, the lack of explanation reduces trust and makes it difficult to apply in the real world. This paper offers an effective explainable ML model for reliable intrusion detection to resolve all these challenges that are specific to the resource-constrained IoT setting. Our proposed framework combines lightweight feature engineering, an effective machine learning-driven detection model, and a layer of Explainable Artificial Intelligence (XAI), offering an insight into the feature-based part of intrusion detection decisions. The framework allows improvement in transparency, trust, and usability by combining detection accuracy and interpretability. The efficiency of the suggested system has been assessed using one of the popular standard datasets, NSL-KDD, allowing consideration of performance in both traditional network traffic and real IoT attacks. The experimental findings indicate that the suggested framework can be used to achieve competitive intrusion detection performance, enhance interpretability, and lower the computation overhead significantly. These findings affirm that reliable intrusion detection can be performed effectively and interpretably without compromising efficiency, and that the proposed framework is viable for real-time implementation on resource-limited IoT systems.

Modern manufacturing sectors require continuous machining operations in the competitive business world of today. Unexpected machining failures result in significant production losses. Therefore, switching from expensive reactive repairs to optimal preventive maintenance schedules requires an accurate reliability assessment. Even though complex machinery includes several subsystems, traditional reliability studies treat it as a single work machine. It obscures the unique failure patterns of individual parts, resulting in erroneous estimates of the machine's total lifespan. This paper proposes a thorough reliability assessment model for a Computer Numerical Machine (CNC) grinding machine with an emphasis on sub-systems in order to fill this research gap. The primary goals of this work are to accurately characterize different sub-system failure behaviors and predict total machine reliability. 252 empirical failure data points from a manufacturing plant's records are used in the study. Machine reliability is evaluated using a thorough series system model that accounts for the mechanical and electrical subsystems. To analyze the unique degradation characteristics of each sub-system, a two-parameter Weibull model is proposed. The least-squares method is used to estimate the shape and scale parameters in order to quantitatively characterize the overall system reliability function. To validate the model, a direct comparison is made between the simulated system Mean Time Between failure (MTBF) and the actual observed MTBF from the raw failure data. The relative error percentage is used to validate the mathematical framework. The machine's overall error was 10.52%. The proposed study's practical implications are crucial for failure and maintenance planners since they enable them to develop maintenance schedules based on these findings.