Paper Title

A MACHINE LEARNING APPROACH TO OPTIMIZE EFFICIENCY AND PRIVACY IN SECURE MULTI-PARTY COMPUTATION

Abstract

Secure Multi-Party Computation (MPC) protocols allow a set of mutually-distrusting parties to jointly evaluate a commonly known function over their inputs, while maintaining correctness of the outputs and the security of their inputs. Privacy-preserving machine learning (PPML) and Secure Multi-party Computation (MPC) has gained momentum in the recent past. As its deployment increasingly depends on data from multiple entities, ensuring privacy for these contributors becomes paramount for the integrity and fairness of machine learning endeavors. We substantiate our theoretical claims through improvement in benchmarks of the aforementioned algorithms when compared with the current best framework ABY3. All the protocols are implemented over a 64-bit ring in LAN and WAN. Albeit its potential, the practicality of MPC is hindered by the difficulty to implement applications on top of the underlying cryptographic protocols. This is because their manual construction requires expertise in cryptography and hardware design. The latter is required as functionalities in MPC are commonly expressed by Boolean and Arithmetic circuits, whose creation is a complex, error-prone, and time-consuming task. We begin with an introduction into compilation and optimization of circuits with minimal size, which is required for constant round MPC protocols over Boolean circuits, such as Yao’s Garbled Circuits protocol.
KEYWORDS : Secure Multi-Party Computation (MPC) protocols, Privacy-preserving machine learning (PPML), cryptographic protocols, LAN and WAN, mutually-distrusting.

Paper Title

SIMULATION AND EXPERIMENTAL STUDY ON THE MACHINABILITY OF TUNGSTEN HEAVY ALLOYS USING FEM

Abstract

Laser beam machining of various materials has found wide applications in the industry due to its advantages of high-speed machining, no tool wear and no vibration, precision and accuracy, low cost of machining, etc. Investigations into the laser beam machining of uncommon alloy are still limited and more research is needed in this field. In this paper, an analysis of the laser beam machining of tungsten alloy was performed, for cutting and drilling machining processes. First, an experimental analysis of micro hardness and microstructure on the laser-cut samples was performed, and then the numerical simulation of the laser beam drilling process and its experimental validation was carried out. The fragmentation and penetration of materials having different mechanical characteristics are studied for the development of tungsten heavy alloy fragile projectile. An FEM analysis model for tungsten heavy alloy fragile projectile is established. Simulation of the process of penetration of tungsten heavy alloy with different ratios of tension to compression against the target was carried out with the dynamic nonlinear finite element analysis software MSC-Dytran. Rules of fragmentation and penetration of tungsten heavy alloy are obtained. Analysis of simulation results shows that the power of fragmentation of projectiles increases as tension intensity falls, but the power of penetration increases as tension intensity increases. Numerical simulation and analysis of results offer a theoretical reference for the engineering design of tungsten heavy alloy fragile projectiles.
KEYWORDS : laser beam machining, tungsten heavy alloy, simulation results, FEM analysis

Paper Title

INTEGRATING AI AND FUSION METHODS FOR HIGH-QUALITY MULTI-FOCUS IMAGING

Abstract

Multi-focus image fusion methods can be mainly divided into two categories: transform domain methods and spatial domain methods. Recent emerged deep learning (DL)-based methods actually satisfy this taxonomy as well. Visual quality evaluation with many numbers of statistical quality metrics were used to evaluate the fusion results with different algorithms. This image fusion approach was also extended to study the texture analysis of final fused image and proposed to incorporate the laws of texture-based energy operator for the image fusion process. Multi-focus image fusion synthesizes a sharp image from multiple partially focused images. However, traditional fused images usually suffer from blurring effects and pixel distortions. EWCM calculates the weights at each position in a content-adaptive manner, suppressing the effects of vignetting artifacts near the edges to preserve more edge details. Specifically, a residual architecture that includes a multi-scale feature extraction module and a dual-attention module is designed as the basic unit of a deep convolutional network, which is firstly used to obtain an initial fused image from the source images. As an important branch in the field of image fusion, the multi-focus image fusion technique can effectively solve the problem of optical lens depth of field, making two or more partially focused images fuse into a fully focused image.
KEYWORDS : Multi-focus image fusion, image structure-driven, Visual quality, EWCM, dual-attention module, blurred images.

Paper Title

DEVELOPMENT OF HIGH-EFFICIENCY SHOCK ABSORBERS FOR IMPROVED RIDE COMFORT IN FOUR-WHEELERS

Abstract

The energy source of vehicles is changing rapidly and significantly in recent years with the increase in renewable energy technologies especially in the case of electric vehicles (EVs). A smart solution has emerged in which the wasted energy in a vehicle’s shock absorber is converted to an alternative energy for the cars themselves, and this is called an energy regenerative shock absorber. Whereas existing regenerative shock absorbers mainly focus on the methods of energy harvesting, there is no such regenerative shock absorber for use in extended range EVs. we present a novel high-efficiency energy regenerative shock absorber using super capacitors that is applied to extend the battery endurance of an EV. A renewable energy application scheme using regenerative shock absorbers for range extended EVs is designed and proposed for the first time. This system collects the wasted suspension power from the moving vehicle by replacing the conventional shock absorbers as these energies are normally dissipated through friction and heat. The proposed system consists of four main components the vibration of the suspension input module, transmission module, generator module and power storage module. The suspension vibration induced by the road roughness acts as the system excitation to the energy regenerative shock absorber. The vibration is then transmitted through the mechanical transmission module, which changes bidirectional vibration into unidirectional rotation based on gears and a rack to drive the generator module. The power storage module stores the regenerative energy of the shock absorber in the super capacitor, which is applied to the EV to improve the cruising mileage.
KEYWORDS : high-efficiency, shock absorbers, energy harvesting, electric vehicles

Paper Title

DEEP LEARNING FOR SARCASM IDENTIFICATION IN TEXT DATA

Abstract

Sarcasm is a form of expression in which individuals convey positive or negative sentiments using words that contradict their actual intent. This style of communication is increasingly prevalent in news headlines and on social media, making it difficult for readers to accurately identify sarcasm. To address this challenge, it is crucial to develop intelligent systems capable of detecting sarcasm in headlines and news articles. This research paper proposes a deep learning-based model for sarcasm detection in news headlines. The model is designed to achieve three primary objectives: (1) to understand the underlying meaning of the text or headline, (2) to learn the characteristics of sarcasm, and (3) to accurately identify sarcasm in the text. Previous studies on sarcasm detection have largely relied on tweet datasets, using hashtags to distinguish between sarcastic and non-sarcastic content. However, such datasets are often noisy due to inconsistent language use and tagging. In contrast, this study leverages multiple datasets to provide a more comprehensive understanding of sarcasm in online communication. By incorporating various forms of sarcasm from the Sarcasm Corpus V2, as well as sarcastic news headlines from The Onion and HuffPost, the proposed model is intended to generalize effectively across different contexts. The model employs Long Short-Term Memory (LSTM) networks to capture temporal dependencies in text and utilizes a GlobalMaxPool1D layer for enhanced feature extraction. Evaluation on training and test datasets yielded accuracy scores of 0.982 and 0.952, respectively, demonstrating the model’s strong performance in sarcasm detection.
KEYWORDS : sarcasm detection; sarcasm; sentiment analysis; text data;

Paper Title

INTELLIGENT SYSTEMS FOR ENHANCED ANOMALY DETECTION IN MODERN CYBERSECURITY

Abstract

Machine learning (ML) models have become indispensable for enhancing the effectiveness of cybersecurity countermeasures, particularly in IDS deployments. However, recent studies have demonstrated that these models are highly susceptible to adversarial attacks. By introducing subtle perturbations into malicious network traffic features, attackers can successfully evade ML-based IDS mechanisms. Consequently, the development of robust defences against such adversarial manipulations has become an urgent priority. Achieving adversarial resilience in cybersecurity systems, however, involves multiple challenges. One of the most significant barriers is maintaining an appropriate balance between an ML model’s robustness and its operational efficiency. Another challenge lies in designing defence techniques that generalise effectively across diverse adversarial attack strategies. Most existing defence mechanisms proposed in contemporary research are predominantly tailored for computer vision domains and are seldom validated using cybersecurity-specific datasets. Unlike images, audio, or video streams, network traffic data possess different temporal characteristics and structural properties. The NSL-KDD dataset is employed to evaluate the proposed HyAD-F against two widely used gradient-based adversarial attack methods—Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD). To assess model behaviour under both adversarial and non-adversarial conditions, three machine learning classifiers—Logistic Regression, Gradient Boosting Classifier, and Multi-Layer Perceptron—are utilised. The integrated defence mechanism yields a substantial improvement in adversarial accuracy for both PGD- and FGSM-generated perturbations, while maintaining minimal degradation in standard cyberattack detection performance. The findings underscore the necessity of conducting rigorous security evaluations on intrusion detection models prior to their deployment in operational environments.
KEYWORDS :