Search

This work explains the DeepFace algorithm and the Support Vector Machine (SVM) classifier, which are utilized for unconstrained face identification. Identifying faces without constraints presents significant challenges, such as dealing with pose variations, lighting conditions, occlusions, and other real-world factors. Our proposed method begins with face detection and facial alignment to accurately locate and normalize the faces in the images and ensure their consistency. We then employ deep learning through a deep Convolutional Neural Network (CNN) for feature extraction. The CNN is trained with a large dataset of labeled faces to capture low-level and high-level facial features. The SVM classifier is used to enhance the efficiency of the Deep Face algorithm by utilizing deep-face features extracted from the CNN. A kernel function is then applied to map these features to a high-dimensional space, optimized based on decision boundaries to separate and classify different identities. The second step involves applying a metric learning approach to improve face-matching accuracy. By using an optimized similarity metric, we can determine the similarity or distinctiveness between one face and another, thereby aiding in the verification and identification process. We also evaluate the performance of our proposed method on different datasets, which serve as benchmarks for unconstrained face identification. Our method achieves a higher accuracy of 99.9% compared to traditional face recognition techniques, reaching results comparable to human quality competition. [ABSTRACT FROM AUTHOR]

The smart, automated car washing device, which is primarily based on Internet of Things (IoT) technology, is a new approach to wash cars more efficiently. This device was carried out using the Mega Arduino board and a few Arduino-like-minded components, including relays, gearboxes, pumps, and sensors. Each element has a relay connected to it to govern turning the part on and off, and the related relay is controlled via the Arduino, which is programmed using the incorporated improvement surroundings, where the synchronization approach is used for the three washing degrees and in accordance with the size of the car to be washed. In particular, the car passes through the three gates of the three degrees. In the primary gate, soap liquid is pumped and unique brushes are moved to clean the car, and in the second gate, water is pumped to put off the liquid soap using pumps. In the third gate, the car is dried utilizing three fans. Thus, the showering process is completed, as the car is eliminated from the three ranges through the transmission. This entire process is controlled by IoT generation via the Hypertext Transfer Protocol (HTTP) browser via customers' mobile gadgets, and the procedure is described. The method of washing and the length of time it takes are determined by the client and in line with his/her preferences. This device includes a civil defense system to provide more security to the website and those found on it. A fireplace sensor was used to detect infrared rays and lighting, and an alarm bell became extensively utilized to alert those within the website. All the system parts are connected to communique networks to send a text message to the worried government, which is involved with reporting the fire situation through the Global System for Mobile Communications (GSM) generation. [ABSTRACT FROM AUTHOR]

Cancellable biometrics has emerged as a promising research area for enhancing security and ensuring the anonymity of biometric signals transmitted over the internet. With the rapid advancement in biometrics, numerous research efforts on cancellable biometrics have been made to maximize security. However, there is no definitive evidence demonstrating that one cancellable biometric technique is better than another, thus highlighting the need to analyze the current state-of-the-art. This study thoroughly examines current progress in cancellable biometrics, categorizing the research based on the approaches used. These approaches include various strategies such as minutiae-based, transformation-based, encryption-based, deep learning-based, and other methods, which are used in the generation of cancellable biometric templates. The analysis of the current state-of-the-art focuses on studying the significant contributions, methodologies employed, analyzed datasets, and performance indicators used in each study, emphasizing the advancements achieved in the respective field. Furthermore, this paper addresses the challenges and potential areas for further investigation in the field of cancellable biometrics. These include several issues such as scalability, robustness, usability, standards, ethics, and long-term revocability. Ultimately, the goal is to provide insights into the current status of cancellable biometrics and highlight potential areas for future progress. [ABSTRACT FROM AUTHOR]

Securing computer networks against cyberattacks is paramount for contemporary organizations, with Intrusion Detection Systems (IDSs) serving as indispensable defenses. This proposed system investigates the potential of Machine Learning (ML) to bolster IDS efficacy, leveraging ML's adeptness in detecting network threats. Despite ML's prowess, handling large datasets presents a challenge, necessitating the selection of pertinent information to enhance efficiency. Additionally, ML-based IDSs often contend with elevated false-positive rates and reduced accuracy on imbalanced datasets. Our evaluation encompasses ML methodologies, including Decision Trees (DT), Random Forest (RF), and Extreme Gradient Boosting (XGBoost), to address these hurdles. Results demonstrate that the XGBoost model, coupled with Particle Swarm Optimization (PSO) for feature selection, achieves unparalleled accuracy (99.51%) in threat identification. Furthermore, a notable reduction in false positive and false negative rates to 0.004 and 0.006, respectively, was observed for the UNSW-NB15 datasets. [ABSTRACT FROM AUTHOR]

This research aims to combine BiLSTM and CNN to provide a new model for applications in image classification. A BiLSTM network is a long-term recurrent neural network (RNN) that can recall dependencies. The main potential of BiLSTM, when used in a multi-layer configuration, is to enhance a CNN's feature extraction capabilities. CNN is able to glean useful information from models that BiLSTM can selectively remember for a long time. The model proposed has performance superior to the standard CNN classifier in image classification. This is due to the data passing through two stages and several layers, the first through BiLSTM and the second through the CNN stage, each of which contains several layers and filters, resulting in high accuracy compared to previous studies. This is because convolutional and recurrent neural networks work together to form the suggested model. This model can be trusted by working great with neural networks for all types of classification tasks. Researchers validated these results by testing our model on a real dataset. The proposed model's training accuracy is (0.9982), and its testing accuracy is (0.9633), indicating its success and suitability compared to other models. [ABSTRACT FROM AUTHOR]

In some cases, sensitive images may be shared for research, consultation, or collaboration purposes. These images often need to be stored for extended periods, sometimes decades. Sensitive medical images, financial documents, or confidential government data benefit from encryption to prevent unauthorized individuals or hacking attempts from accessing them. By combining the benefits of Piece-Wise Linear Chaotic Maps (PWLCM) features with Message Digest (MD5), this research provides a novel approach for producing safe key images. In the beginning, we emphasize the necessity of safe key images for cryptography and multimedia applications. Highlighting the PWLCM-based and MD5-based techniques that are currently in use and pointing out their security and complexity shortcomings. The fundamental information section introduces the concepts of PWLCM and MD5 hashing, along with their mathematical underpinnings. Our proposed model is then presented, detailing the steps involved in extracting PWLCM features, incorporating randomness through chaotic maps, and generating a robust key image using MD5 hashing. The program was written using the Python Pi programming language. In a 256*256 size image, there will be 256 potential keys for each pixel. For our proposed method using a 512*512 image size, the key space will be 2512*8=24096. The results section showcases the experimental evaluation of our method, demonstrating its effectiveness in terms of key space, and resistance against common attacks. Finally, the conclusion summarizes the contributions of our work and paves the way for future research directions in secure key image generation. [ABSTRACT FROM AUTHOR]

Applications for sensor networks are many and have a broad range of possibilities. But for sensor network systems to function well in practical applications, some problems must be solved. Energy saving is a major issue for sensor networks because most sensors lack rechargeable batteries, which have a limited lifespan. Dynamically scheduling sensors' work and sleep cycles is a frequent strategy for extending the lifespan of a sensor network (or duty cycles). Additionally, cluster heads in cluster-based networks may rotate among the sensors to maintain an even distribution of energy use and are often chosen to reduce overall energy usage. In general, these energy-saving topology configuration methods (also called scheduling mechanisms) need to meet certain application needs while saving energy. In this article, we give a summary of energy-efficient scheduling techniques for sensor networks with distinct design specifications from those of conventional wireless networks. We classify these techniques based on their priority traffic sensitivity, clustering, and Time division. In addition to having the same goal of maximizing network longevity, each mechanism may also have unique design goals based on the applications it is intended to serve. [ABSTRACT FROM AUTHOR]

The task of recognizing and categorizing vehicles in videos and images as objects poses a considerable challenge in terms of appearance-based representation. However, it holds great importance in the practical implementation of Intelligent Transportation Systems (ITS), particularly in real-time applications. The fast advancement of deep learning has resulted in an increasing need within the computer vision field for the development of efficient, robust, and outstanding services across diverse domains. This paper provides an extensive analysis of various methodologies for vehicle detection and classification, along with their utilization in real-time targets, estimating the density of traffic, and related domains through the implementation of Deep Learning techniques. The major findings of our survey highlight crucial insights obtained from an extensive analysis of existing literature, shedding light on the current state-of-the-art techniques. Through a comprehensive review of deep learning methodologies, performance metrics, benchmark datasets, and a comprehensive exploration of the challenges encountered, our survey offers valuable contributions to the field. By synthesizing and presenting the collective knowledge in this domain, our paper serves as a key resource for researchers and practitioners alike, providing a holistic understanding of the advancements and challenges in vehicle identification and categorization within deep learning architectures. [ABSTRACT FROM AUTHOR]

Recently the requirement for Appling AI models at edge devices without connection to clouds is increased. For this reason. A lot of companies produced embedded systems with a smaller size and low power consumption like the popular Raspberry Pi and Nvidia's Jetson Nano for acting as compute edge nodes. Put all these systems required more processing time than desktop or pc laptops for this reason companies support this device with hardware and software resources to reduce processing time.in this paper, in this work NVIDIA jetson nano use as the edge device and siamapn++ as Siamese single object tracking PyTorch model, and TensorRT as optimization firmware to optimize the model. After optimizing we found that the speed is increased x2 times than the original PyTorch model. [ABSTRACT FROM AUTHOR]

Wireless Sensor Networks (WSNs) have been increasingly popular in recent years for various applications in the real world, like tracking, monitoring, transportation, military, and healthcare. As a result, maintaining the sensor node's lifetime is critical in WSNs. The battery's energy is the most important part of the sensor nodes, influencing the WSN's lifespan. Because the sensor nodes are powered by a limited battery, energy conservation is critical. At sensor nodes, energy is required for several tasks, including data reception and transmission, data processing, sensing, and so on. Data transmission consumes the greatest energy of all of them, although data processing consumes significantly less. As a result, reducing the amount of data communication (transmitting and receiving) is crucial to conserving energy and extending the lifetime of WSNs.DABOEP Data aggregation based on Extrema Point is proposed in this paper, which works at the sensor node level. Data collection, data aggregation, and data transmission are the three stages of the suggested techniques' operation periodically. The goal of these techniques is to reduce the size of transmitted data by aggregating sensed redundant data before transmitting to the base station (BS), reducing the amount of energy consumed and thus extending the network lifetime while maintaining an acceptable level of accuracy for the data received at the BS. Extensive simulation experiments are used in the evaluation of suggested techniques. When compared to the flat network (normal case), the suggested technique reduces the quantity of data left by 77%–90%. Energy consumption is reduced by 78%–99%. The ratio of lost data is ranging between (1.11%-9.01%) and the amount of data sent by 96.551% to 98.424%. [ABSTRACT FROM AUTHOR]

Encryption techniques are important on the Internet. The challenges of color-coding are still hot topics. Many papers were published to develop encryption algorithms to prevent security attacks. Cryptography is the process of converting plaintext into ciphertext. It is classified into two types: symmetric encryption, which uses the same key for encryption and decryption processes; and asymmetric encryption, which uses private and public keys. Text to image color coding needs to encrypt text data into colors. Researchers proposed different techniques to implement this approach to safeguard text data against attacks. This work compares various types of text to image encryption and decryption techniques based on the RGB and CMYK color values and provides a comparison between them using various parameters such as application, program language, type of image, number of keys used, image dimension used, encryption and decryption processing time, advantage and disadvantage, image size used, evaluation measures, and so on. [ABSTRACT FROM AUTHOR]

A Sensor network compromises more than one sensor that sends sensed data towards a base station (BS) in a cooperative way. There is a constraint in every sensor node, and that can be the requirement of power consuming. Every sensor node carries restricted sources of power that cannot be replaced. In the scenario of real-life as those nodes are spread into all kinds of environments, they might not be nearly put to the base station. Actually, those very nodes might be very distant from one another. Since every sensor node is battery-driven, a skillful use of power-consumption can be considered necessary for decreasing data traffic in each sensor network, hence decreasing the data required to be sent towards the (BS) thereby improving the lifetime of networks. The major aim of the data aggregation algorithms is to gather and aggregate data in an efficient energy manner, in that case, the network lifetime enhances. Here, in this work we present a comprehensive review of energy-efficient methods in wireless sensor networks. We also submit a few significant aspects of wireless sensor networks (WSNs) in relation to data aggregation as well as different data aggregation techniques. This work shows that these methods not only reduce resource consumption but also extend network life. [ABSTRACT FROM AUTHOR]

Cryptanalysis is the investigation of cipher content, ciphers and cryptosystems with the point of seeing how they work and finding and improving procedures for crushing or debilitating. Elliptic Curve Cryptography is one of the newest fields in cryptography field. This paper has proposed a new technique where classical techniques are replaced with new technique by changing characters with affine points. This technique avoids costly mapping. [ABSTRACT FROM AUTHOR]