International Journal of Intelligent Engineering and Systems

1

Authors: Fauza Khair, I Wayan Mustika, Dodi Zulherman, Fakhriy Hario P

Light fidelity (Li-Fi) offers communication services with a wide bandwidth and is one of the options for telecommunications services in the future, especially indoor communication. However, Li-Fi requires accurate device positioning for line of sight (LOS) conditions to meet the main requirement for light communication. Conventional Li-Fi design models require a more dynamic transmitter and receiver device settings but can still serve LOS communication well. This study aims to design a Li-Fi communication model by varying the transmitting devices, such as the use of fixed and movable LEDs, and receiving devices, such as the addition of optical rectangle filters, low-pass filters, and trans-impedance amplifiers. The focus of observation includes the effect of changes in wavelength, data transmission speed, transmitter half-angle, irradiance angle, and incidence angle, as well as field of view (FOV) on received power parameters, signal-to-noise ratio, Q-factor, and bit error rate (BER). On the basis of the test results, the performance of all wavelength variations meets the ITU-T standard up to a variation of the transmission distance of 8 m, where the use of fixed and movable LEDs does not result in a significant difference in performance. The test results indicate that the wavelength parameter affects the signal quality due to the spectral response and LEDs emission. For short distances, the wavelength of 450 nm has better performance, whereas for distances up to 8 m, the wavelength of 650 nm has a better performance than other variations. Furthermore, increasing the transmitter half-angle increased the BER value and drastically decreased the Q factor. The use of a smaller FOV is recommended, as shown from the increase in distance resulting in a significant decrease in performance. The proposed system model can also be used for bit rates up to 30 Mbps, although it is only reliable up to 6 m. As for the 40 Mbps variation, the 75° half-angle transmitter, 60° FOV, and 90° FOV variations do not meet the standard. The two system models do not have a significant difference in performance. Thus, using movable LED instead of fixed LED will make the system more dynamic with a well-maintained LOS.

24

Authors: Gurusiddayya Hiremath, Jose Alex Mathew, Naveen Kumar Boraiah

Cancers are aggressive which is coupled with higher mortality rates and remain a major life-threatening factor in humans. Thus, early detection of cancer among patients is important as it showed a great survival chance. Therefore, early detection provides the patients with greater survival chances. The diagnosis performed for the mammography images are quite expensive and also radiations produced during the process were harmful to the patients. Thermography is a cost-effective and invasive method compared to mammography images and thus has reached its popularity. The present research is aimed to create the Machine Learning (ML) models using Convolutional Neural Networks (CNN) approaches which were created based on the machine learning models. The present research work utilized the DMR-IR dataset for the results evaluation and performance evaluation of the model which has been verified with datasets. The feature extraction process utilizes a machine learning algorithm to overcome the problems. The developed models were engaged in sophisticated ways to extract the features to improve the classification of the model. The Hybrid statistical and texture feature extraction technique extracts the features better in turn improved the model training. The results showed that the proposed hybrid feature extraction with the Dense Net 121 model obtained better accuracy of 98.97 %, the precision of 99.45%, Recall of 98.35%, F-score of 96.85%, Sensitivity of 99.4%, and Specificity of 97.98 % when compared to the existing Multi-input CNN that obtained accuracy of 93.8%, precision of 94.1%, Recall of 97.7%, F-score of 91.4%, Sensitivity of 88.9%, and specificity of 96.7%.

50

Authors: Maha Kadhim Kabier, Ali A. Yassin, Zaid Ameen Abduljabbar

During the COVID-19 pandemic, online electronic educational systems have been used in most schools and universities as they were forced to move their operations from classrooms to online settings. However, these systems face a serious security issue. Access control considers the core of data security for any implemented system. This paper presents the well-known role-based access control (RBAC) approach to enhance system security and improve user role and system privilege. This study also addresses the issues faced by extant schemes, such as security risk tolerance, by proposing a privacy-preserving educational system that utilizes RBAC and smart multi-factor authentication. This approach uses an asymmetric cryptosystem based on the Elgamal digital signature operation to provide multi-factor authentication while relying on low-complexity cryptographic hash functions. RBAC manages system security via the “user classification, role authorization, and unified management” approach. By limiting the amount of data that users can access, RBAC is particularly suited for multi-level applications. This approach also uses informal analysis and the Scyther tool to conduct extensive formal security proofs. RBAC offers many benefits, including mutual authentication, identity anonymity, forward secrecy, key management, and high resistance to well-known attacks, such as phishing, replay, Man-In-The-Middle (MITM), and insider attacks. Compared with other schemes, RBAC offers more security features and boasts higher cost effectiveness in processing and communication. Furthermore, our work achieves a good balance between performance and security complexity when compared to the state-of-the-art. So, we get good results at a cost of 0.253 ms for computing and 1326 bits for communication.

74

Authors: Mustafa Abdul Salam, Nashwa Nageh Ahmed, Ahmed Elshamy, Abdel Wahab Said Hassan, Mostafa Sami

This study presents an enhanced model based on a new swarm intelligence algorithm called jellyfish search optimizer (JSO). The suggested model is called chaotic jellyfish search with enhanced swap operator (CJSESOS). The CJSESOS algorithm has two modifications to the original JSO algorithm, the first is the chaotic sequence generated by iterating a logistic map which is named CJSO. This enhancement discovers a creative solution by directing particles to different location of the search space, the second is enhanced swap sequence operator which increased the CJSO algorithm's ability to escape from local minimum by diversifying the results. The performance of the CJSESOS is evaluated using fourteen benchmark functions. The proposed model is applied to solve a discreate real life problem, named team formation (TF) problem which considered one of the most significant problems in computer science and optimization. TF problem is defined as creating the most effective team of experts in social network to carry out a task with the lowest possible cost. The proposed CJSESOS algorithm was tested for solving the TF problem with varying number of skills in different datasets. In addition, the proposed algorithm is compared to well-known optimization algorithms such as jellyfish search optimizer (JSO), particle swarm optimization (PSO), genetic algorithm (GA), grey wolf optimization (GWO), heap based optimizer (HBO), aquila optimizer (AO) and pelican optimization algorithm (POA). The simulation results show that the proposed model outperformed all the compared algorithms on the term of efficiency and accuracy.

100

Authors: Furkan Rabee, Israa Jazaery, Kamlesh Kumar

The optimization method with the genetic algorithm (GA) is well-known in different science fields. The crossover is the primary important operand of the genetic algorithm, so many researchers worked with this operand to get the optimal solutions. This article presents a new crossover operand for the genetic algorithm by supposing a new organization for genes inside the chromosome. The new gene supposed that contains two alleles, and each allele refers to an attitude; thus, each gene has two attitudes. One attitude of the gene supposes to be preferred over another. Therefore, two crossovers were presented, the first crossover applying to intense the preferred one on another as enhancing method of genes to a single individual’s chromosome before mating. After the first crossover, the mating process will come by the second crossover, which mating the parents or individuals. The second crossovers' output is four children; eliminate the two lowest fitness, and select the others as new offspring, and they will be part of the next generation. This process is repeated for each population until it reaches the last generation. The new crossover is called the quaternary-child crossover. The proposed is supposed to be applicable for real-time scheduling and is not limited to that. Empirical experiments have tested the proposed work as an optimization method with different mathematical and scheduling equations that had been used in optimization by researchers, and the performance was satisfactory.

125

Authors: Peddi Kishor, Sammulal Porika

Association rule mining (ARM) is to discover entire association rules with the confidence and support that are accordingly equivalent or above user-specific lower confidence and support. The existing suggested ARM methods can do the operations on the static datasets only, where the operation of the complete dataset must be accessible in the initial process of mining. Classical ARM methods are designed well on working with static data, and so, they cannot be used instantaneously for mining the association rules in data. To accomplish this, a novel framework is developed for association rule mining with the aid of an adaptive concept. Initially, the required data is collected from the benchmark datasets. The next step is to develop the novel association rule mining algorithm (NARMA), where the parameters are tuned by Modified Scavenger rate-based Dingo Billiards Optimization (MCR-DBO) by the hybridization of Dingo Optimizer (DOX) and Billiards-Inspired Optimization (BIO). This algorithm is used to identify the distinct itemsets of data. Consequently, the resultant dataset is given for finding the most eminent factors, where the itemsets are categorized into positive association rules as well as negative association rules. When the minimal threshold values are fixed with a minimum support count of 0.3, the proposed model achieved an average 81% confidence for generated top 1000 association rules from WebDocs dataset by minimizing the execution time and memory requirements. The outcomes disclose that the designed approach competently mines both the positive and negative rules and outperformed other recommended techniques.

152

Authors: Parvati Kadli, Vidyavathi B M

Sarcasm detection is a challenge in review based e-commerce product recommendation systems. Failure to detect sarcasm causes assignment of directionally opposite ratings to reviews and faults the recommendation systems. This work proposes a cross domain hybrid feature fusion based sarcastic opinion recognition system. Hybrid deep feature integrating content and context is formed and feature vector volume is expanded using cross domain adversarial learning. The hybrid features are used for sarcasm detection in two combinations. First combination is hybrid features with bi-directional long short term memory (LSTM)and second is hybrid features with traditional machine learning classifiers. Hybrid feature with bi-directional LSTM is able to capture temporal relation between sentences of the review for better identification of sarcasm. Traditional machine learning classifiers trained with these hybrid features are also able to provide higher accuracy of sarcastic detection in reviews. The effectiveness of the proposed cross domain hybrid feature fusion features is tested against various experimental setups with Bi-directional LSTM and traditional machine learning classifiers. The proposed solution is able to achieve peak accuracy of 95% in classifying e-commerce reviews.

176

Authors: Moch. Syahrir, Anthony Anggrawan

The Apriori algorithm is one of the most widely developed and used association rule algorithms because it can produce optimal rules. The association rule a priori algorithm has three main problems: the old dataset scan process, less than optimal rules formed, and the use of large memory. Therefore, this study aims to improve the performance of the a priori algorithm in the itemset frequency search process with optimal rules, small memory usage, and fast dataset scans. The method used in this study is TID-List Vertical and data partitioning. This study result indicates that the TPQ-Apriori method improves the performance of the dataset scan process up to two times faster in scanning datasets compared to the study's results using the traditional a priori method. The performance of the dataset scan process from the results of this study is also superior in processing speed by up to 7% on scanning datasets compared to the ETDPC-Apriori method on testing three data sets (mushroom, chess, and c20d10k).

204

Authors: Jayaseelan Loganathan, Sureshkumar Chinnaiyan

The compression and encryption of images is essential to securely transmit high quality images over the optical network. In earlier studies, an optical image encryption using Loxodromic cat map with improved double random phase encoding (LCMIDRPE) has been developed. Hilbert huang transform (HHT) used in this encoding model decrease deviations of encrypted images. However, efficient image compression technique is required for compression. An efficient optical image compression and encryption (OICE) technique is developed in this paper by Direction-adaptive discrete wavelet transform (DA-DWT) along with directional lifting and LCMIDRPE. The original image is compressed using the DA-DWT with directional lifting and then encrypted by LCMIDRPE. The inverse version of LCMIDRPE and DA-DWT regain the original image. The missed information’s due to lossy compression, decompression, encryption, decryption and transmission of images degrade the quality of regained images. The regained image is enhanced using intensified convolutional neural network (ICNNet). The whole process is termed to be LCMIDRPE-DADWT-ICNNet. Finally, the experimental results exhibits that the LCMIDRPE-DADWT-ICNNet model achieves correlation coefficient (CC) of 0.98 , Peak signal-to-noise ratio (PSNR) of 97db, Mean absolute difference (MAD) of 20 and mean square error (MSE) of 0.45 which is higher than OICE methods like Compressive Sensing and Rivest-Shamir-Adleman method (CS-RSA), Fresnel Diffraction and Discrete Wavelet Transform (FD-DWT), 2Dimentional Sparse Representation and Chaotic Map (2DSR-CM), 2Dimensional Compressive Sensing and Hyperchaotic System (2DCS-HS), Diffractive-Imaging-Based Encryption (DIBE) and Content-Adaptive Image Compression and Encryption using Optimized Compressive Sensing with Double Random Phase Encoding (CAIE-OCS-DPRE) methods.

228

Authors: S.V. Nethaji, M. Chidambaram

The internet of things (IoT) affects everyday life because digital technology improves. IoT is a group of devices with sensors that talk to each other to reach a goal. IoT systems have traditionally been built on top of cloud computing (CC). IoT devices are slow because CC data centres are separated from them. This slows down the speed at which real-time applications respond. In addition, IoT devices send a lot of data to the cloud to be processed, which overloads the cloud. Edge computing can stop IoT devices from being slow or overloaded. Fog computing (FC) is a way to get services at the edge of a network. With locationawareness, the FC cuts down on latency and overloading. Bandwidth and jitter must be looked at during the process of allocating resources. In this work, the Lotka-Voltera load balancer and elman hebbian-recurrent neural network cache (LV-EHRCC) are proposed for allocating resources in an FC context. LV-EHRCC is made up of load balancing and allocating resources. First, the LotkaVoltera traffic load balancer model is used to increase the amount of Bandwidth available for load balancing. Second, an Elman Hebbian-Recurrent Neural Network model for allocating cached resources efficiently is made for the best load-balanced FC context. Simulations test what will happen. The load balancing capacity of the proposed scheme is 93.25 and attains the highest Bandwidth of 45. In FC simulations, the LV-EHRCC method improves the efficiency of load balancing in terms of Bandwidth, makespan, and jitter rate. The simulation results back up our study and show that LV-EHRCC is better than the benchmark approaches when they are compared.

255

Authors: Mohamed S. Sawah, Shereen A. Taie, Mohamed Hasan Ibrahim, Shereen A. Hussein

In highway management, intelligent vehicle detection and counting are becoming increasingly important as an accurate estimation of traffic density on road congestion reduction. Traffic density estimation is affected by the difficulties of perspective distortion, size change, significant occlusion, and background interference in traffic images. To address the previous issues, this article develops a novel model that enhances the quality of estimating traffic density. The efficientNet fine-tuning architecture is used then, followed by the development of seven dilated convolutional layers to extract the deeper features in the images that maintain the output’s resolution to generate a high-quality density map. Finally, the vehicle count will be calculated from the high-quality density map. The experimental results indicate that the suggested approach significantly enhances the accuracy of traffic density estimation compared to the existing ones. It achieves 5.23 as a mean absolute error (MAE) on the TRANCOS dataset.

279

Authors: Ahmad M. Awaad, Hesham Hefny

We present a new parallel density-based spatial clustering of applications with noise (DBSCAN) algorithm for spark on the google cloud platform (GCP). Statistical analysis is applied to determine DBSCAN's optimal parameters to enhance clustering performance. for scalability cost-based, R-tree partitioning is selected based on the distribution of the dataset into balanced workloads. Parallel DBSCAN consists of three parts: local DBSCAN, partitioning, and merging. Optimizing the partitioning of parallel DBSCAN is important to save time and space compared to serial DBSCAN. This approach can improve the performance and time cost of large datasets. the modified statistical cost-based (SCbs-DBSCAN) is applied to the UCI (university of california irvine) standard datasets, basic benchmark clustering and large different scales of data. For clustering performance and time cost, the experimental results show that the proposed algorithm achieve 10~15% more efficiently, and can run about 1.5x~3x faster than alternative Parallel DBSCAN method on Spark without sacrificing clustering quality.

307

Authors: Ekhlas K. Gbashi, Abeer Tariq Maolood, Yaseen Naser Jurn

Fog computing is an efficient paradigm to diminish latency and congestion of networks, in which the applications and data are no longer deployed in the cloud alone. Alternatively, the cloud server is improved with devices working near the edge layer and end users such as mobile applications and sensors respectively in internet of things (IoT) scenarios. Video surveillance system is a critical issue needs considerable computing and network instruments to provide the substantial quality of data. Multimedia communication includes videos transferring between fog and cloud server via secure channel is considered a critical problem needs to resolve. This paper proposes a light-weight cryptographic scheme to provide the secrecy and privacy for the transferred video data within edge-fog-cloud platform. Specifically, the framework of the proposed cryptographic scheme involves three main phases; secret key (SK) generation algorithm and distribution processes, pre-processing of video frames based on frame region of interest extraction FROI approach, and video data encryption/decryption phase using 2D chaotic map. The major parameters that effect on video encryption performance within edge-fog-cloud infrastructure are latency, accuracy and practicability. The secret key generation process is implemented into central fog node (CFN), while keys distribution and management processes are implemented into cloud computing layer for higher security distribution procedure. Common evaluation benchmarks were utilized in the experimental results to analyse and evaluate the proposed cryptographic scheme. The performance evaluation of the proposed cryptographic scheme is carried out based on some of metrics which are memory consumption rate, encryption time, rate decryption time, key sensitive analysis and computation complexity via using urban surveillance video dataset (USVD). The results of these metrics prove that the proposed scheme has been very suitable for securing transmit video data for edge-fog-cloud structure against intruders. Scientific comparisons against stat of art methods are conducted in our experiments to highlight the performance evaluation of the proposed scheme. Finally, the finding results based on scientific comparisons show that the proposed scheme is more appropriate for transfer encryption video frames from fog-layer to cloud-computing layer.

332

Authors: Hussein Mahdi Ali, Nidhal K. El Abbadi

The retina is the visible component of the nervous system that links directly to the brain; hence, analyzing and extracting features of the retinal fundus image is essential in ophthalmology. Unfortunately, most retinal fundus images suffer from degradation or distortions, resulting from different imaging qualities during capture and infection by a disease of the eye, such as Glaucoma, Retinoblastoma, Diabetic Retinopathy (DR), Myopia, and Macular Edema. As a result, these changes make it difficult to identify regions of the retina that assist in diagnosing the disease. For instance, many retina diseases diagnosis is impossible to detect without detecting the optic disc (OD) correctly. All these issues motivate us to suggest an algorithm for the localization of the OD that is important for assisting in diagnosing some eye diseases. The proposed algorithm includes several stages. Firstly, we focus on gathering seventeen datasets that make the suggested algorithm more popular, where a group of images belongs to four datasets used in training and the rest in testing. Secondly, the quality of the image improves in three steps, crops the Field of View (FOV), which carries essential information in the retinal image, removing pixels at the top and bottom of the FOV because they represent redundant information, and we suggested a new method for enhancing the illumination and contrast of the retinal fundus image. Thirdly, we proposed using a You only look once (YOLO) for the localization of OD. The method locates the OD exactly (drawing box around the OD and each side of the box touches the OD), not the region around it as in the previous methods. The current method works excellently with all the challenges that may face in the detection and localization of the OD. The proposed model tested with about 6000 different images from various datasets and challenges The accuracy obtained was 100% with an average time of 0.19 seconds. These results were very good compared with previous works in terms of accuracy, average time, and the number of tested images.

361

Authors: Antono Adhi, Nurhadi Siswanto, Budi Santosa

Maritime Inventory Routing Problem (MIRP) is an important issue in the optimization of maritime distribution and transportation. This problem is related to planned ship’s routing and scheduling in delivery of goods from the depot to some demand points by minimizing associated costs such as transportation and inventory management costs. This paper discusses how to solve MIRP with multi ships and limited undedicated compartments used to deliver bulk products from the depot port to the several consumption ports. The new hybrid metaheuristics are used to find the optimal assignment routes and schedule of ships along time horizon. This research modifies several metaheuristics algorithms called Modified Hybrid Particle Swarm Optimization (MHPSO) to find the best solution for MIRP. The algorithm of this method is developed from the combination of Particle Swarm Optimization (PSO), Nahwaz-Enscore-Ham (NEH), and 3-Opt. Some metaheuristic methods such as Genetic Algorithm (GA), Tabu Search (TS), Particle Swarm Optimization (PSO), Hybrid Genetic Algorithm (HGA), and Hybrid Tabu Search (HTS) were also developed in the same way to test and compare with the proposed method. Based on the ten test data instances, it can be concluded that MHPSO provides 0.64% effectiveness better results than other metaheuristic methods.

389

Authors: Janani M, Jebakumar R

Nitrogen nutrient concentration is predominant in crop field monitoring. Noticing the deficiency pattern, providing sufficient amounts of nutrients will help to recover the crop. Sometimes it may not be possible to reverse the crop in cases of deep deficiency. It will affect the quality and quantity of the crop product in yield time. Hence, timely identification is important. Therefore, detection of chlorophyll deficiency in groundnut leaves (DCDGL) system is proposed here for nutrient deficit recognition, before a notable pattern emerges. In first phase, groundnut leaf images are collected and clustered based on chlorophyll measurement. Then, different image processing techniques like image compression, resizing, and noise reduction filters are applied and analyzed on leaf images to enhance the images and secure the predominant features. The median filter is fixed because it provides high quality output with a low noise percentage. The quality of an image is measured in decibels with the help of the peak signal to noise ratio (PSNR). Then, in the next phase, colour features and grey level co-occurrence matrix texture features are extracted and given to developed intensified multinomial classification (IMC) model to detect nutrient deficiencies in groundnut leaves. The obtained average accuracy of IMC model is 97%. The DCDGL system’s performance is tested with state-of-the-art techniques like K-nearest neighbour and support vector machine. The developed model performance is superior to state-of-the-art techniques regarding accuracy. The appropriate image quality analysis and pre-processing approach helped achieve less processing time to predict nutrient deficiency before tangible symptoms.

414

Authors: Dony Hutabarat, Muhammad Rivai, Muhammad Agung Nursyeha, Djoko Purwanto, Sheva Aulia

Robots have been widely used to overcome problems in the environment. In this case, mobile robots were used to find the source of gas leakage. Each robot was equipped with gas sensors in the form of a stereo-nose and Light Detection and Ranging (LiDAR). The robot formation in finding the gas leak source had the advantage of expanding the gas detection area, meaning the robots could not easily lose the pathway. However, detecting through formation required a longer execution time. To speed up the robots in finding the gas leak source’s location, this study involved the heuristic model. This model allowed each robot to change positions in a formation. The experiment results showed that the robot group using the heuristic model was 62.2% faster than a single robot. This method was also 19.4% faster than the robot group with a fixed formation.

440

Authors: Purba Daru Kusuma, Astri Novianty

Order allocation problem is a crucial problem in the manufacturing system. Many studies have been carried out to overcome this problem. On the other hand, future studies related to this problem are still available due to its complexity, circumstance, and methods. This work develops a new metaheuristic called a multiple interaction optimizer (MIO). MIO has distinct mechanics in finding the optimal solution. MIO consists of two phases. In the first phase, each agent interacts with some randomly selected agents in the population. The guided search is conducted in every interaction. In the second phase, each agent carries out a local search which linearly reduces the search space during the iteration. Three tests are carried out on the performance of MIO. In the first test, the MIO is challenged to solve 23 functions. The second test is performed as the investigation of the hyper parameters. In the third test, MIO is challenged to solve the order allocation problem with the objective is minimizing the total cost, total lateness, and total defect. In this test, MIO is benchmarked with five metaheuristics: particle swarm optimization (PSO), marine predator algorithm (MPA), grey wolf optimizer (GWO), slime mould algorithm (SMA), and golden search optimizer (GSO). The result indicates that MIO is superior in solving both 23 functions and order allocation problems. In the theoretical test, MIO outperforms PSO, MPA, GWO, SMA, and GSO in 22, 21, 22, 19, and 18 functions respectively. MIO is also superior in achieving minimum cost, minimum lateness, and minimum total defect in solving order allocation problems.

464

Authors: Radhwan Jawad Kadhim, Hussein K. Khafaji

With the rapid advancement of the digital network, information technology, digital libraries, and especially the services of the world wide web, many types of information can be retrieved at any time. Thus, the issue of security has become one of the most important problems in distributing new information. It is essential to protect this information while it is passing through unsecured channels. Steganography offers a powerful approach to hiding confidential data in suitable media vectors such as images, audio files, text files, and video files. In the field of steganography, the most common measurements of hiding the intended data in a cover media are hiding capacity, cracking probability, payload, and bit per nucleotide (bpn), and these measurements are considered the most of the prime challenges in the field of steganography. The main objective of this research is to provide a novel data hiding schema using protein-based steganography that provides good security performance measurement and outperforms DNA-based steganography approaches, where protein sequences are not utilized for steganography purposes. According to the proposed method, each byte of the secret message is partitioned into two 4-bit parts, and then these parts are converted into the decimal system. Finally, the decimal values of the secret message are randomly included in the cover protein sites using a pseudo random number generator (PRNG) for each cover protein base value instead of being sequentially embedded in the cover protein bases. It is considered that this random distribution increases system security. The proposed method works with a one-digit protein decimal coding rule (PDCR), and the byte of the message will be embedded in two bases of amino acids. From the experimental results, it has been found that the proposed method preserves the reference protein's original function (zero modification rate in the original protein sequence) and achieves a very high hiding capacity of 4 bpn when the sequence's bases are entirely data-embedded, a very low cracking probability, and a payload of zero. Furthermore, the proposed method saves 50% of the bandwidth compared with the other existing technique.

486

Authors: Hrushikesh Joshi, Uttam Patil, Kuldeep Sambrekar

The Internet of Things (IoT) has become an infrastructure for various domain applications such as smart homes, smart cities, wearables, smart grids, etc. Due to these applications, the internet of things search engines has attained some attention from various researchers, industry, and users. The search throughput of IoT search engines is crucial because they must execute hundreds of spatial-time-keyword inquiries in a second. Moreover, IoT search engines use cloud resources to execute tasks. Some of the tasks might have a small workflow but some of the tasks might have a larger workflow and require more virtual machines. Instead of using the virtual machine the best option is to utilize a caching mechanism. In recent years, there are only a few methods that use the caching mechanism for the execution of the task. Furthermore, these methods are not reliable, and efficient and have not considered multi-core edge-cloud computing. To overcome this issue, this paper proposes a high-reliability method using the cache-failure mechanism for the multi-core edge-cloud computing architecture. The proposed High Reliability through cache-failure minimization (HRCFM) technique's main aim was to provide better tradeoffs during the execution of the task using the caching mechanism and reduce energy consumption. The results have been compared with the existing energy minimizing scheduling (EMS) and reliability (REL) technique in terms of execution time, power sum, power average, and energy consumption. The results have been compared and it shows that the proposed HRCFM technique reduces the time, energy consumption, power sum, and average by 83.3%, 90.92%, 38.44%, and 91.59% for EMS technique and 81.91%, 91.4%, 41.24% and 91.85% for REL technique respectively.

506

Authors: Waad Q. Shiaa, Amjad B. Abdulghafour, Omar H. Hassoon

In intervention procedures or planning for joint replacement, the models developed must be precise and represent the natural look of the bones. For instance, the function of the knee joint is dependent on the motion of the femur. Therefore, precisely reconstructing the femur model assists in the design of the components for joint replacement. Moreover, when accurate measurements are taken for implant design, these models have a high effect on their accuracy. As a result, obtaining accurate 3D models of human bone tissue is important. Therefore, this paper proposes a novel method for accurate bone tissue reconstruction from CT data. The method is based on histogram modelling, interpolation, and reverse engineering techniques. Initially, soft tissues and unwanted artifacts are eliminated in order to extract the ROI (bone). After that, feature points are extracted based on pixel variance from each slice and then interpolated to create a 3D point cloud. Finally, a 3D model is created from point cloud data. Additionally, the model is printed using a 3D printer based on the FDM technique. Furthermore, the method is tested on CT stacks of femur and sheep bone in order to evaluate its strength and performance. Results indicated that the feature points collected by the proposed method could accurately describe the bone geometry. The error percentage between the proposed method and the actual bone was approximately 1.62%. The results indicate that the proposed method provides promising outcomes. Furthermore, the 3D model of bone is more compatible with the real bone.

526

Authors: Seresane Venkata Krishna Reddy, Jayanthi Keshava Murthy

Wireless sensor networks (WSN) is self-organizing network that has numerous tiny sensor nodes used to track and monitor applications in an extensive range. However, security and energy consumption are considered two important issues, because of the open medium and restricted energy resources. In this paper, the multiobjective-trust centric reptile search algorithm (M-TCRSA) is proposed to perform the secure cluster based routing for WSN. The M-TCRSA is used to ensure the secure cluster head (SCH) and secure route discovery to achieve reliable communication over the WSN. Hence, the developed M-TCRSA method provides improved security against malicious attacks while enhancing energy efficiency. The main objective of the M-TCRSA is to achieve secure data transmission while increasing the life expectancy of the WSN. The performance of M-TCRSA is analyzed by means of packet delivery ratio (PDR), throughput, end to end delay (EED), jitter and life expectancy. The existing research such as trust-based routing using adaptive genetic algorithm (TAGA), trusted cluster based energy and lifetime aware routing (TCELR), secure clustering and routing using IMFO (SCRIMFO) and cost centric cuckoo search algorithm (CCCSA) are compared with the M-TCRSA. The PDR of the M-TCRSA is 99.27% for 400 nodes which is high when compared to the TAGA, TCELR, SCRIMFO and CCCSA.

546

Authors: Nur Nafi’iyah, Chastine Fatichah, Darlis Herumurti, Eha Renwi Astuti, Ramadhan Hardani Putra

Mandibular segmentation is an important step in gender identification and age estimation, which aims to segment the mandible from intact and complete panoramic radiograph. One of the main drawbacks of most existing mandibular segmentation methods is that they cannot completely represent the mandible. When conducting several segmentation experiments with several methods, namely U-Net, MobileNetV2, ResNet18, ResNet50, Xception, InceptionResNet V2, MobileNetV2 turned out to be superior. However, if you only use the MobileNetV2 method, the results still need to be clarified on the coronoid and mandibular condyles. Then it is necessary to add an ensemble so that the mandibular segmentation results become more intact and complete. In contrast to the usual MobileNetV2, the mandibular segmentation results are assembled to achieve a complete and intact performance. Finally, this method experimented with 38 panoramic radiographs verified by radiologists. The experimental results show that the proposed MobileNetV2 ensemble for segmentation was superior to the usual MobileNetV2 method with a dice value of 0.9655.

571

Authors: Sukesh Hoskote Aswathanarayana, Sundeep Kumar Kanipakapatnam

Skin cancer is one of the most commonly occurring cancer and it causes hundreds to thousands of yearly deaths worldwide. Early identification of skin cancer significantly increases the recovery chances from skin cancer. However, precise skin cancer classification is a challenging task because of the ineffective segmentation of skin cancer. In this paper, the saliency-based level set with an improved boundary indicator function (SLSIBIF) is proposed for the effective segmentation of skin cancer. An improved boundary indicator function is used in the segmentation to detect the skin cancer boundaries even under the constraints of low intensity and illumination. The features from the segmented images are extracted by using the GoogLeNet which uses sparse connections to extract an optimal feature. Further, the classification is done using a multi-class support vector machine (MSVM). The performances of the proposed SLSIBIF-MSVM are evaluated using accuracy, sensitivity, specificity, positive predictive value (PPV), error rate, jacard, and dice coefficient. The existing approaches such as deep-learning system (DLS), ResNet-50, K-means with grasshopper optimization algorithm (GOA) and Region-based CNN (RCNN) and Fuzzy K-means (FKM) are used to compare the SLSIBIF-MSVM. The classification accuracy of SLSIBIF-MSVM for ISIC-2017 dataset is 98.74%, which is high when compared to the DLS, ResNet-50, K means GOA and RCNN-FKM.

591

Authors: Oday Ahmed Al-Ghanimi, Hussein K. Khafaji

Frequent itemsets (FIs) mining is the main challenge in analyzing association rules since the complexity of association rule mining is determined mainly by identifying all frequent itemsets. There are many approaches to FIs mining; each approach includes many algorithms, but there is a major weakness in the logic-design-based FIs approach because there is a limited number of methods for logic circuit or expression optimization, each of which suffers from some drawbacks that are transferred to the mining process when they are utilized for this purpose. We propose LCOFI, a new algorithm for frequent itemset mining based on the LCO algorithm, in this paper. LCOA is a new algorithm for logic circuit optimization. LCOFI utilizes the suggested operations for a bipartite graph in LCOA. The proposed algorithm is simple and efficient and supports a large number of input items. It scans the transaction database only once to construct the bipartite graphs of frequent 1-itemsets and avoids the generation of candidate itemsets. It does not require complex data structures such as trees and hash tables to validate the frequency of FIs. When applied to a variety of datasets with varying characteristics, the proposed algorithm, LCOFI, outperformed the Apriori and Fp-growth algorithms in all the experiments.