The Trinity Academy of Engineering is a technical education institute in the city of Pune, India. The institute is affiliated with the University of Pune and managed by the KJ's Educational Institutes (KJEI). It has been accredited by the National Board of Accreditation and recognized by the All India Council for Technical Education (AICTE). The institute has also been awarded an "A" Grade by the National Assessment And Accreditation Council (NAAC) and Directorate of Technical Education, Maharashtra.
Modern connected vehicles face increased cyber threats due to expanded IoVattack surfaces. Robust intrusion detection systems are essential to counter vulnerabilities from weak authentication and encryption. Traditional methods are inadequate, therefore, an advanced real-time intrusion detection system (IDS) is needed to detect and mitigate evolving cyber-attacks in connected vehicles. To overcome these complications, optimized PRNN-ENet for robust IDS in IoV networks (IDS-PRNN-ENet-IoV) is proposed. The input data is collected from car hacking dataset. The gathered dataare provided to the preprocessing phase. Here, fast guided median filter is used to normalize the data. Afterward, the pre-processed data are fed into the physically recurrent neural network (PRNN) with EfficientCovNet (PRNN-ENet),which classifies and detects the intrusion asnormal, DoS, fuzzy, gear spoofing and RPM spoofing. Finally, a bitterling fish optimization algorithm is employed to optimize the weight parameters of PRNN-ENet. The IDS-PRNN-ENet-IoV technique is executed in Python and the metrics like accuracy, recall, f1-score is examined. The proposed IDS-PRNN-ENet-IoV achieves 6.14
In this paper, we are suggesting an Explainable Deep Neural Network (DNN) pipeline to identify cyber threats in real-time with the help of Shapley Additive Explanations (SHAP) to understand the models. The field of cybersecurity is crucial, and the timely and proper identification of threats cannot be considered an exception. The deep neural networks prove to be very effective at determining the presence of intricate patterns in data; however, its black-box characteristics make it hard to trust and understand. To solve this, we add SHAP to the DNN pipeline in order to understand model prediction clearly. SHAP provides a value of each feature that reflects its input in the output of the model, which makes the process of decision-making transparent. The suggested solution will utilize real-time network traffic data and feed it through the DNN to classify the possible threats and SHAP to explain why a particular action or a detection took place. The findings show that such an approach does not just provide a good way of detecting cyber threats but also gives the required interpretability to enable system administrators to know and trust the decisions of the model.
This paper presents a real-time thermal face detection and recognition system based on an enhanced Multi-task Cascaded Convolutional Neural Network (MTCNN) framework. Unlike visible-light methods, thermal imaging introduces domain-specific challenges such as low spatial resolution, high noise, and intensity variance due to temperature fluctuations. To address these, we propose a dedicated preprocessing pipeline including normalization, contrast enhancement, and channel replication to adapt single-channel thermal images for CNN-based processing. The modified MTCNN is fine-tuned on thermal datasets to accurately detect facial regions and landmarks. Aligned faces are then processed through a thermal-optimized feature embedding network trained with triplet loss to produce identity-preserving descriptors. Recognition is performed using a lightweight classifier over the feature space. The system is optimized for real-time performance using GPU acceleration and quantized inference. Experimental results on publicly available thermal face datasets demonstrate the effectiveness of our approach in terms of detection accuracy, recognition rate, and processing speed, making it suitable for surveillance and biometric applications under low-light or no-light conditions.
A completely novel method of harnessing wind energy is employed by bladeless wind turbines. The vortex-shedding phenomenon, an aerodynamic feature that has long troubled structural engineers and architects, it is captured using a gadget. The wind’s flow alters as it passes by a permanent object, creating a cyclical vortex pattern. When these forces get sufficiently powerful, the stationary structure begins to oscillate, may even collapse as it enters resonance with the wind’s lateral forces. The Tacoma Narrows Bridge is a well-known academic example that collapsed 3 months after it was opened due to the effects of galloping and flattering as well as the vortex-shedding effect. The goal of the Spanish SME Vortex Bladeless is to create the vortex or vorticity wind turbine, a novel idea for a wind turbine without blades. With the goal of removing or reducing many of the current issues with traditional generators, this design offers a new paradigm in wind energy.
The speedy enlargement of the Internet of Things (IoT) has revolutionized contemporary verbal exchange networks with the useful resource of interconnecting billions of gadgets all through domain names consisting as healthcare, clever cities, and commercial automation. However, this exponential growth has introduced big protection traumatic conditions because of the heterogeneous nature of IoT devices, restricted computational sources, and the use of light-weight communique protocols. Traditional encryption algorithms, even though ordinary, impose excessive computational overhead, making them incorrect for beneficial useful resource-confined IoT environments. To address the barriers, this study proposes an AI-pushed protocol-degree safety size framework that integrates Simplified Advanced Encryption Standard (S-AES) with system learning-based totally completely absolutely optimization. Unlike current-day works that depend completely on algorithmic overall performance or protocol enhancements, the proposed framework introduces an AI-optimised mild-weight encryption scheme that dynamically regulates encryption parameters across multiple IoT communication protocols, which includes MQTT, CoAP, and AMQP. The framework is finished in Python and evaluates the use of a simulated IoT test mattress comprising Raspberry Pi nodes and virtual sensors. Experimental results show that the proposed model achieves an average 22\% reduction in power consumption, 11\% improvement in latency, and a 15\% increase in throughput compared to standard lightweight encryption algorithms, including PRESENT, Midori, and HIGH. These enhancements validate the framework’s adaptability and average overall performance in securing IoT communications below diverse network situations.
