Gongcheng Kexue Yu Jishu/Advanced Engineering Science

In digital signal processing, Finite Impulse Response (FIR) filters are fundamental components that govern behavior by specific coefficients. Protecting these coefficients from adversarial discovery is crucial, and prior techniques have introduced obfuscation through decoys or key bits. However, these methods remain vulnerable to query attacks capable of revealing secret keys. This work introduces a novel hybrid protection technique that combines hardware obfuscation and logic locking, utilizing a point function to secure parallel direct and transposed forms of FIR filters. Additionally, a Proportional-Integral-Derivative (PID) controller is integrated to enhance system performance and resilience. An innovative unlock function is also developed to ensure controlled access to the protected coefficients. Experimental results demonstrate that the proposed technique offers superior security, protecting against prominent query-based and logic attacks while maintaining competitive hardware complexity. The protected multiplier blocks and FIR filters, including the integrated PID controller, show enhanced security and efficiency compared to traditional methods.

The explosive growth of data generated on social media platforms like Twitter presents both opportunities and challenges for real-time anomaly detection. Traditional approaches struggle to scale with the velocity, volume, and variety of such data. This paper proposes a scalable framework for anomaly detection using stream mining techniques built on Apache Spark and its machine learning library, MLlib. The system is designed to process high-throughput tweet streams in real time, detect anomalous patterns, and evaluate the performance of various anomaly detection algorithms including Streaming K-Means and Isolation Forests. Twitter data is used as a dynamic, high-velocity input source to simulate real-world streaming environments. The framework leverages Spark Streaming for real-time ingestion and distributed processing, while MLlib enables scalable model training and inference. Comprehensive performance analysis is conducted using key metrics: speedup to measure parallel processing efficiency across Spark clusters, throughput to quantify the volume of tweets processed per second, and accuracy to evaluate detection effectiveness. Experimental results demonstrate that the proposed framework can achieve near-linear speedup, high throughput, and competitive accuracy in identifying tweet-based anomalies, such as sudden spikes in sentiment, bot-like behavior, or misinformation bursts. The system proves to be robust, scalable, and suitable for deployment in domains requiring low-latency insights from continuous data streams.

The objective of this study is to develop a robust and scalable framework for detecting fake product reviews in large-scale e-commerce platforms using Graph Neural Networks (GNNs). The research aims to improve the accuracy of review authenticity classification by modeling user-product-review interactions as graph structures, capturing relational and behavioral patterns beyond traditional feature-based methods.The proposed approach constructs a heterogeneous graph from an e-commerce dataset where nodes represent users, products, and reviews, and edges encode their relationships. We apply a Graph Neural Network model that leverages both structural and semantic features extracted from the graph. Attention mechanisms are incorporated to focus on influential connections. The model is trained and evaluated on publicly available datasets such as Amazon and Yelp, with preprocessing steps including review vectorization, graph formation, and normalization. Benchmark comparisons are made against conventional machine learning classifiers and deep learning models. Experimental results demonstrate that the GNN-based model significantly outperforms baseline methods, achieving a higher F1-score, precision, and recall across multiple datasets. The graph-based approach exhibits superior ability to capture deceptive behavior patterns and uncover collusive review activities. The model also shows scalability when applied to datasets with millions of reviews.

A Mobile Ad Hoc Network (MANET) is a decentralized wireless system where nodes communicate dynamically without relying on fixed infrastructure. Due to their self-organizing nature, MANETs have become a crucial research area, particularly in addressing challenges related to mobility and bandwidth constraints. Data transmission occurs through multiple network hops, with each node acting as a router to forward packets toward their destination. Traditional routing protocols designed for fixed networks struggle to adapt to MANETs' dynamic topology, necessitating specialized routing mechanisms. Routing protocols are often classified into reactive, proactive, and hybrid models, with research focusing on their efficiency in handling network disruptions. Studies highlight the importance of adaptive routing strategies to ensure reliable communication in MANET environments.

The goal of sentiment analysis is to automatically determine the tone of a piece of written content. Monitoring social media, analyzing product evaluations, and evaluating consumer feedback are just a few of the numerous applications where it is growing increasingly. Machine learning algorithms have greatly enhanced the performance and accuracy of sentiment analysis. This paper provides an extensive overview of machine learning techniques for document and sentence-level aspect sentiment analysis. The shortcomings of more conventional machine learning methods for sentiment analysis are outlined. After that, numerous machine learning architectures that have been effectively used for this purpose was investigated. Furthermore, the difficulties of handling various types of data, including visual as well as multimodal data, along with how both methods have been modified to overcome these obstacles was considered. In addition, the ways sentiment analysis may be used in various fields, such as product evaluations and social media was investigated. Lastly, possible future research areas and draw attention to the present limits of machine learning algorithms for sentiment analysis was discussed. This survey is designed to give academics and practitioners a thorough grasp of the latest machine learning algorithms used for sentiment analysis while explaining how they work in practice.

New financial innovations and formidable regulatory hurdles, especially with regard to Anti-Money Laundering (AML) compliance, have emerged with the advent of Bitcoin. The decentralised, anonymous, and ever-changing nature of blockchain transactions makes traditional rule-based anti-money laundering solutions sluggish. This study offers a new framework that uses Reinforcement Learning (RL) to ensure that cryptocurrency transactions are always compliant with anti-money-laundering regulations. Robot learning agents may learn the best ways to detect, report, and flag suspicious actions in real-time by modelling anti-money laundering compliance as a sequential decision-making process. In order to identify developing patterns of illegal activity including structure, layering, and integration, the suggested RL model communicates with a virtual blockchain setting. The agent adapts its rules on the fly according to transaction details, user behaviour, and compliance results using a combination of continuous learning and incentive feedback. The system's capacity to adapt allows it to surpass static rule-based methods, particularly when it comes to identifying new methods of laundering and reducing the number of false positives. Furthermore, the framework uses explainable RL approaches to make compliance judgements more transparent and easier to understand, which is important for regulators to approve the system. A combination of supervised pretraining with labelled transaction data and temporal-difference learning allows the system to optimise policies to a finer degree. Additionally, it incorporates a compliance risk score system to prioritise transaction review according to changing behavioural risk instead of fixed criteria. Initial experimental findings show that as compared to traditional AML infrastructure, the RL-based method considerably enhances detection precision and flexibility. In addition to improving productivity, it optimises compliance procedures and decreases the need for human interventions. An intelligent, scalable compliance network that can respond to changing crypto dangers is possible, according to this study, and reinforcement learning may help with both ensuring compliance with AML and this endeavour. This work adds to the expanding body of research in artificial intelligence-driven regulatory technology (RegTech) by providing a thoughtful and proactive strategy for protecting the honesty of decentralised financial transactions. Connectivity with cross-chain compliance standards, real-world deployment issues, and multi-agent cooperation will be the topics of future study.

In recent years, abstractive text summarization using multimodal inputs has garnered significant attention from researchers due to its ability to synthesize information from various sources into concise summaries. Text summarization creates a concise version of the original document by identifying key information, but it is considered a general approach as it doesn’t capture the distribution of opinions or sentiments. In contrast, review summarization offers a detailed breakdown of product aspects and associated sentiments, helping online shoppers make informed decisions. Due to the informal style, short length, and unstructured nature of reviews, review summarization is challenging. This study introduces an aspect-based abstractive summarization method for customer reviews, utilizing an encoder-decoder model with attention and pointer generator networks. A Bi-GRU encoder-decoder ensures that adjacent words contribute to the summary's coherence. The proposed automatic text summarization is compared over the existing models in terms of performance measures like ROUGE metrics achieves high scores as R1 score 43.61, R2 score 22.64, R3 score 44.95 and RL score is 44.27 on Benchmark DUC datasets.

This paper presents a fuzzy-based energy management and Fault Ride-Through (FRT) enhancement approach for a hybrid grid-connected power integrating Wind Energy Conversion System (WECS), Photovoltaic (PV) machine, and piezoelectronics. The proposed fuzzy controller dynamically regulates electricity distribution and inverter management, making sure seamless renewable energy source (RES) integration whilst preserving grid stability throughout abnormal situations. Compared to traditional PI and droop control strategies, the proposed machine improves performance to 96%, reduces FRT response time to 18 ms, and achieves complete low voltage ride-through (LVRT) and high voltage ride-through (HVRT) capabilities, allowing uninterrupted grid connection below voltage sag and transient disturbances. The fuzzy-based technique enhances voltage regulation, transient stability (by using 35%), and decreases voltage sag (via 45%), extensively outperforming conventional manipulate strategies. Simulation consequences verify that the proposed strategy guarantees superior grid synchronization, improved dynamic reaction, and high reliability in hybrid renewable power systems. This has a look at demonstrates that fuzzy based FRT enhancement is a promising solution for modern smart grids, presenting improved suppleness and efficiency in managing grid disturbances and power fluctuations.

Dynamic window systems play a essential position in improving strength-green thermal and visible consolation in lecture rooms. Focusing on Ranipet Government School in Tamil Nadu, India, the research evaluates the effectiveness of fenestration designs across unique seasons—summer, monsoon, and wintry weather. By making use of simulation equipment and on-website online measurements, the look at objectives to evaluate how dynamic windows can optimize natural lighting fixtures and control heat advantage/loss whilst selling strength efficiency in school room environments. The impact of seasonal variations on thermal consolation, glare discount, and natural lighting fixtures is explored, with findings indicating that the current constant window design fails to fulfill most appropriate consolation standards. The research also highlights the importance of adapting window structures to particular climatic situations to improve gaining knowledge of environments. The proposed sustainable window designs goal to balance strength efficiency with comfort, reducing reliance on synthetic cooling and lighting fixtures even as creating a conducive atmosphere for college students. The findings provide valuable insights for designing power-green classrooms in government colleges, specifically in useful resource-confined areas, and make contributions to the broader aim of sustainable and responsive instructional infrastructure.

The Threat Onboarding and Response (TOR) framework is an innovative approach to enhancing cybersecurity in enterprise networks, moving beyond traditional perimeter-based security models. Focusing on cloud and hybrid environments, this study explores the use, challenges, and effectiveness of TOR in modern business settings. TOR addresses critical security threats by automating threat onboarding and response, with an emphasis on real-time detection, continuous monitoring, and incident response. This report examines the obstacles businesses face in implementing TOR, such as high costs, integration complexities, and resistance to change, by reviewing relevant literature and analyzing case studies. It also identifies strategies employed by organizations to overcome these challenges, leading to improved security and operational efficiency. The findings highlight how TOR’s automation and proactive monitoring can significantly reduce security risks. While adopting TOR presents certain technological hurdles, the study concludes that the framework is essential for establishing a resilient security posture. Future research could investigate the role of AI and machine learning in advancing TOR capabilities.

Sentiment analysis (SA), also known as opinion mining, is a burgeoning field of natural language processing (NLP) that aims to computationally identify, extract, and analyze subjective information expressed in textual data. With the proliferation of social media platforms and online review websites, sentiment analysis has gained immense importance in understanding public opinion, customer sentiment, and social trends. This paper presents a comprehensive overview of sentiment analysis techniques, methodologies, and applications, encompassing both traditional rule-based approaches and modern machine learning algorithms. The paper discusses various aspects of sentiment analysis, including sentiment classification, aspect-based sentiment analysis (ABSA), multiclass sentiment analysis, and cross-domain sentiment analysis. Furthermore, it examines the challenges and opportunities in sentiment analysis, such as handling sarcasm, irony, and negation, dealing with multilingual and multicultural data, and ensuring scalability and efficiency in large-scale text processing. Additionally, the paper explores emerging trends and future directions in sentiment analysis research, including the integration of deep learning techniques, the incorporation of multimodal data sources, and the exploration of sentiment analysis in emerging domains such as healthcare, finance, and politics. By providing a comprehensive overview of sentiment analysis, this paper aims to serve as a valuable resource for researchers, practitioners, and policymakers interested in leveraging sentiment analysis for various applications in the digital age.

Sun spot is a natural phenomenon which impact on various changes in the earth especially in weather, climatic conditions, disasters and diseases etc. The existing methods on prediction of sun spots are majorly focused on mathematical and statistical implications. In order to make more productivity, in this research, we applied Artificial Neural network and Machine Learning Methods on sun spots data set to prediction accuracy of various sunspot data. The results and accuracy are based in the training of dataset with SVM classifiers and Machine Learning Methods. The merits of this research prove the prediction accuracy of sun spot numbers and LSTM models are to be applied for the sunspot predictions which performs better in accuracy. SVM classifiers hold various advanced algorithms to make training on data to achieve better accuracy and performance. Among the various kinds of SVM classifiers in the area of prediction Linear SVM, Quadratic SVM and Cubic SVM are been considered for this performance evaluation especially in Prediction accuracy. The main parameter applied for prediction is Rooted Mean Square Error (RMSE). The experimentation part take place using the MATAB tool. An improved Vanilla LSTM model is proposed to overcome the various challenges in the existing models. The main objective of this model is to make predictions using sun spots. This Vanilla Long-Short Term Memory (LSTM) model is applied with optimized hyper parameters with fine tuning i.e. batch size, epoch, and optimizer etc., Adam optimizer is applied for the optimization during the process. Single layer is used and the optimized hyper parameters provide better results. The prediction process is accomplished by data set and is pre-processed with normalization then the sequence is created and the LSTM architecture is established. Based the training and testing data the prediction process is done. The model evaluates the similarity measures such as Absolute Error, Relative Error and Related Mean Square Error (RMSE). The performance of the model is estimated by comparing with the existing Stacked LSTM and Vanilla LSTM model.