Shanmugam Padmapriya
@prathyusha.edu.in
PROFESSOR/CSE
Prathyusha Engineering College
EDUCATION
PhD(Computer Science)
RESEARCH INTERESTS
AI, IoT, MACHINE LEARNING
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
S. Padmapriya, R. Shankar, R. Thiagarajan, N. Partheeban, A. Daniel, and S. Arun
Computers, Materials and Continua (Tech Science Press)
B. S. Liya, S. Arun, and S. Padmapriya
AIP Publishing
Padmapriya Shanmugam, Balajivijayan Venkateswarulu, Rajalakshmi Dharmadurai, Thiagarajan Ranganathan, Mohan Indiran, and Manikandan Nanjappan
Wiley
Mobile malware is malicious software designed specifically for targeting various mobile gadgets like tablets, smartphones, and so forth, in which any type of malicious code affecting the mobile devices without the knowledge of the user. The increasing number of users encourages the hacker for generating various malware applications. Therefore, in this paper, we utilized three vital phases namely the pre‐processing process, Feature extraction process as well as classification process in which the malicious data are detected. In the pre‐processing phase, an Androguard tool is used for decompiling and disassembling the android applications. The API call features are extracted in the feature extraction phase and in the classification phase, long short term memory based electro search optimization (LSTM‐ESO) is employed to detect the unknown mobile applications as benign or malicious. The malicious mobile detecting accuracy deals in requesting permission and exhibiting malicious code applications. In order to enhance the identification of various malware applications, this paper utilized frequency analysis and permissions of API calls. Finally, the experimental analysis is performed by evaluating the performance measures like accuracy, precision, recall, and F‐measure. From the evaluation outcome, it is observed that the classification accuracy obtained is 97.69%.
Vineet Kumar Singh, M Manoj Kumar, J. Yuvaraj, T. Rubeshkumar, Sumit Kumar, and S. Padmapriya
IEEE
The IoT - Internet of Things is essential in a diversity of sectors. The highest rainfall afflicts most places in the winter period of each year. As a result, climate and other ecological and environmental catastrophes must be monitored. The primary goal of our system is to propose an IoT-based automatic weather system to give farmers an idea about whether cultivating or harvesting crops would have been profitable and to avoid and notify disasters. The system suggested in this paper offers a way to track weather patterns precisely and make the data obtainable to any person in the world. The Internet of Items (IoT) is the innovation behind all this, a cost-efficient solution for linking devices to the internet and linking the entire universe in a network. Temperature, Humidity, Rain Drop, Pressure, Light Intensity, and other sensors were used extensively in this architecture. Other relevant variables, such as the dew point is determined with the assistance of humidity and temperature. The data updated by the established system is available on the internet and may be accessed from anywhere. In MATLAB, the system design topology was implemented. The units are linked to the laptop and have been extensively examined in a software development environment.
Padmapriya S, Bagavathi C, and Jagadeeswari M
IEEE
In recent days, the unauthorized replication problem causes many copyright ownership disputes regarding the originality of the digital contents. Digital watermarking is the best solution for this problem. The main challenges in designing a digital watermarking are robustness, speed and tradeoffs involved. A watermark inclusion and extraction system is designed without affecting the quality of the image considering the security. To overcome the unaccepted delays a speedy embedding technique is required. A novel water marking algorithm for images using Discrete Wavelet Transform (DWT) is proposed which minimizes the complexity, robustness, resource utilization and security for data processing. The proposed watermark embedding and extraction algorithm is implemented using Xilinx System Generator.
Jagadeeswari M, Padmapriya S, and Selvaganesh M
IEEE
The paper focusses on an efficient reconfigurable Finite Impulse Response Filter (FIR) design based on the sub-expressions for low power applications. The reversible gates are used instead of the logic gates for the design of an efficient low power reconfigurable filter. Across the adjacent coefficients, sub-expression elimination and the variable sub-expression elimination were applied within each coefficient. The proposed Reversible Binary Common Sub-Expression Elimination (BCSE) algorithm is implemented in FPGA and also in Cadence EDA tool. The results shows that, the proposed algorithm power consumption is reduced by 140558.856 nano watts along with an improvement in the Power saving ratios to 7.6%, 20.6% and 28% with the existing techniques respectively.
Kannan Chakrapani K. Chakrapani, P. Malathi, U. Iniyan, R. Thiagarajan, S. Padmapriya, and R. Krishnamoorthy
IEEE
Due to the improvement in digital technologies, the security of the data gets into drastic trouble. Several type of encryption-based algorithm has been initiated which tended to be insecure and procure threats to the digital world. To avoid the flaws in the digital era, the researchers brought up new encryption algorithm in existing framework. For a better accuracy purpose, support vector machine algorithm has been developed to provide image-oriented encryption. Some of the parameters are been detected to categorize the image dataset. Machine learning can be used to process large amount of vast data. SVM is used to classify each dataset in different aspects. Here in this paper, the enhanced homo morphic encryption (EHE) is used. Logistic regression provides less precise loss by classifying it. Data gets supervised since logistic regression does not have closed solution. Based on classification, it determines whether the data is been secured or violated. EHE is used for computation on the already encrypted data without assuming the value. For the efficient way of securing message transmission in mobile ad hoc network EHE is used.
G. Saritha, S. Famitha, P. Malathi, A. Priyadharshini, S. Arun, and S. Padmapriya
IEEE
Recently, many micro blogs are used by people to share their perspective and opinions instantly. In this updating world, each and everyone's view is very important to implement any social projects, solve social issues etc. So, we came with discrete sharing Emotions on Covid-19 using deep learning, which is a subfield of artificial intelligence and gives continuous quality improvement in a large and complex process. In this study, we use Latent Dirichlet Allocation to do topic modelling on the COVID-19 Open Research Dataset (CORD-19). The CORD-19 data is the biggest collection of coronavirus studies to date, making it difficult to gain a full understanding of the dataset. This is when topic modelling might come in handy. In general, topic modelling relates to the procedure of building a predictive method characterising the probability of topics to a dataset. LDA is a type of unsupervised generating mode. This paper is an effective means of discovering the polarity of people's opinions and displaying accurate information. Further the accuracy is also found using CNN and we have achieved 81% of accuracy. This concept can be implemented in various domains like Finance, Medicine, public actions to predict the values and worth of their products or projects among the people to enhance their further process.
V Anbumani, S Padmapriya, S Soviya, S Sneha, and L Saran
IEEE
In image applications, noise removal is one of the important requirement. Impulse noise is removed by median filters using image processing. Median filter is the good non-linear filter used for edge preservation in images and in real-time hardware implementation, sorting network is used. Sorting network must have an efficient implementation in terms of area, delay and power. Row three cell sorter, column three cell sorter, and right diagonal three cell sorter comprise the median filter. The median value of the inputs is the output of the median filter. Three data comparators give the maximum, minimum and middle value in the three cell sorter. The comparator returns the highest and lowest value of given inputs. The middle value is calculated using a modified shear sorting method. This paper compares two existing data comparators and suggests a modified version of those two existing comparators in factors of delay and area. Full adders and full subtractors are used to build the proposed comparators using MUX. The two’s complement based data comparator consumes less time. As a result, an implementation using two’s complement is suited for high-speed implementations. The carry select data comparator based implementation consumes smaller area, hence it can be employed in reduced area applications. The median filter is intended to be used with a two’s complement based data comparator. In the parallel architecture of modified shear sorting for constructing median filter, different architectures for data comparators were used.
C. Bagavathi, S. Padmapriya, and H. Mangalam
IEEE
Evolutionary Algorithms (EA) are standard search mechanisms that use Natural Selection and Survival of the Best as the fundamental algorithm progressing mechanism. The parameterless portfolio is a special technique designed to resolve various categories of problems without any prior requirement of parameter setting. This technique involves an increase in computational effort that can be considered acceptable. In this work, parameterless swarm algorithm using the method of Particle Swarm Optimization has been defined for the application of Traveling Salesman Problem. The performance of the algorithm through the application of parameterless portfolio has been analysed and it can be deduced that the effort of making the Evolutionary Process parameterless can be justified through the benefits discussed in this work.
Ayan Das Gupta, K. Sathiyasekar, R. Krishnamoorthy, S. Arun, R. Thiyagarajan, and S. Padmapriya
IEEE
Wireless sensor networks generally consist of low energy-consuming devices which are eventually distributed in isolated environments. WSN plays a vital role in the sensors as it uses the wireless medium in frameworks. These WSN are used to gather the data or information in a systematic way by using the interference in an environment. Based on the usage of WSN, the sensing of the data is analyzed using the networking. To increase the network capability and its lifetime, the network optimization of the WSN techniques has to be handled in an efficient way. By reducing the cluster of nodes, it can reduce the energy consumption. In this paper, a proposed genetic (GA) algorithm is used to resolve the issues in characterizing the cluster of nodes during the network optimization. The network feasibility and energy consumption can be reduced using heterogeneous heed and celrp protocol. Due to the heed protocol and celrp, the lifetime of the network gets increased due to the minimal loss. The heed and celrp protocol minimize the threshold range by improving the effectiveness in the network. After evaluating it, the network optimization increased the network capability with the low-cost efficiency.
R. Manikandan, T. Mathumathi, C. Ramesh, S. Arun, R. Krishnamoorthy, and S. Padmapriya
IEEE
The embedded type of de vices in IOT generally depends upon the resource constraints which include memory capabilities, low power consumption and reliable in cost. The constrained devices such as edge server are handled at the end nodes. The end nodes such as sensors and actuators are connected using the gateway devices which connect the IOT cloud-based platform. A wireless device which has the limited set of processing and storage-based capability which runs based on the wireless medium or batteries is the resource constrained device. Resource constrained devices provides the efficient way of limited processing with the maximal data output along with the minimal power as input. These are generally cost effective as it consumes less energy and power consumption in devices. The edge server is a type of resource-constrained devices which is the entry point of the network and application. In this paper, the research is based upon the proposal model of resource constrained devices by reducing the parameters using the DNN. The DNN model parameters reduce the memory, execution latency by attaining the higher accuracy. To preserve the higher accuracy in the device computation, the Knowledge Distillation Method is proposed. The knowledge distillation method determines the output predictions of larger DNN into the smaller DNN trained sets. This methodology reduces the trained model by compressing the model accordingly. These smaller DNN predicts the output and behaviors similar to the larger DNN. Smaller DNN predicts approximately equal to the larger DNN. Knowledge Distillation Method is used in several applications in machine learning such as natural language processing, AI, Detection of objects and neural networks graph correspondingly.
S Ramachandra, Sathyajee Srivastava, M Roshini, S. Arun, S. Padmapriya, and R. Krishnamoorthy
IEEE
The scheduling in cloud is considered as a major task in cloud computing environment as it is associated with severe constraints while deploying or scheduling a task. It is hence necessary to model a schedule that effectively deploys the schedule the task based on the constraints. In this paper, we develop a DenseNet model to improve the rate of scheduling in cloud computing environment. This algorithm helps in the task assignment and execution in the cloud. The simulation is conducted in cloudsim simulator, where the communication costs is lesser than existing methods. Similarly, the execution of task takes lesser time in proposed than in other methods.
J. S. Bhatti
Informa UK Limited
Cloud computing is the accessibility of computer resources which includes storing of data and power to compute the data, when there is a demand of it for several applications. It is used to represent the data centers to users in the end over the Internet. The user cannot directly manage it actively. The challenging problem is the integrity of outsourced data. The encryption and decryption of data must be secure to overcome this problem. Here we use MD5 encryption algorithm to secure the outsourced data. Data replication is avoided by using de-duplication technique; thus the efficiency of memory storage is enhanced and time for processing is minimized.
R. Krishnamoorthy, B. S Liya, S. Arun, S Padmapriya, Gunasundari B, and R Thiagarajan
IEEE
As the age of the elderly patients are getting increased, the health factors are also getting affected. By the drastic increase of the human being population, the health factor has to be checked and monitored accordingly. One of the main factors which have to be monitored is the heart rate. Heart rate is said to be the rate of which the heart beat per minute. Heart arrhythmia/ irregular heartbeat can cause heart stroke which causes fatal too. Heart stroke occurs based upon the vibratory state of the heartbeat. If the Heartbeat is mild when compared to the normal heart beat, then the pulse rate is low for the patient. If the Heartbeat is irregular which can vibrate in high stroke then it is strong heart stroke/ severe cardiac arrest where the blood flow is excessive to the artery. Based upon the heart stroke, the range of cardiac arrest can be detected. Blood pressure is also maintained identify the flow of excessive flow. By detecting both the heart rate and blood pressure state, the stroke can be detected. In this novel paper, describe the way of identifying different variations in the heartbeat and blood pressure using machine learning approach. Supervised learning classifies the data in comparative analysis by using SVM, random regression and other techniques. These compares the statistical data of the previous data report by classifying and comparing the data into a statistical report via application.
P. Kalyani, Sathyajee Srivastava, A Reddyprasad, R. Krishnamoorthy, S. Arun, and S. Padmapriya
IEEE
All subfields of medical image analysis, such as classification finds a greater acceptability for Convolutional Neural Network (CNN), since it offers flexible finding of the instances based on the input query. The issues with CNN owing to limited labels and scarce data are solved by employing CNN. In this paper, we develop an increased processing CNN design are opening the door to better results for massive datasets. The classification of images required the utilisation of the three fully connected layers to retrieve features. The architecture for medical image retrieval was put to the test using widely known measures such as precision and recall. This success would lead to better computer-aided detection and diagnosis systems in the long run.
S Padmapriya, Shankar R, R Thiagarajan, S. Arun, B.S Liya, and Gunasundari B
IEEE
Smart Heath incorporates big data and IOT in recent years. The wearable devices such as gadgets watch are used to assist with the health activity and constant monitoring of the health. By using the integration of using wireless sensor networks, the large amount of data is associated with the patient’s health condition. As the data are stored in the cloud, privacy over preserving the data plays a vital role as the communication channel gets compromised. To increase the secure aspects over the communication range, Data-CAAC mechanism is used. Data-CAAC incorporates with the smart watch to constantly monitor the access over the data mechanism both in space and time context mechanism. LSH is used to enhance the secure mechanism by enclosing the confidentiality data along with it. To encrypt these data, Data-CAAC encryption technique is used to encrypt over the access control where the cuckoo filter carries out the data structure to avoid getting compromised. The automated SVO logic analysis is used to determine the correctness in accuracy where it achieves high efficiency and low computational cost. It attains high reliability with good accuracy performance.
K Lokesh, Sathyajee Srivastava, M. Praveen Kumar, S. Arun, S. Padmapriya, and R. Krishnamoorthy
IEEE
In this paper, we develop a deep learning model using dense neural network (DenseNet) to detect the gastric cancer in stomach region using computerised tomography (CT) imaging. The image is initially pre-processed and the features are extracted, where these features are used for training the DenseNet to develop a model. The model is tested and validated against various stomach cancer datasets to check the efficacy of the model. The simulation is validated in terms of various performance metrics that include accuracy, precision, recall and f-measure. The results show that the proposed method is effective in improving the rate of detection over various CT images than other methods.
R Thiagarajan, Gunasundari B, S Padmapriya, B.S Liya, Shankar R, and S. Arun
IEEE
Mobile Adhoc Networks (MANET) is one of the key research areas in networks. More interest is created towards this field due to the reason that it is infrastructure less and no central administration is needed. There are so many routing protocols present in MANETS. The transmission of packets from source to destination without any delay, packet loss by maintaining the energy level of the node is the critical issue existing. In this paper the approach for sorting out these issues are found out. Entire research is carried out in two phases. Firstly the comparison of routing protocols is carried out. The comparisons of DSR, DSDV and AODV protocols are done by considering the key parameters. The results are compared and understood that AODV performance results better when compared to other two routing protocols. Further to improve the shortcomings of AODV multipath routing protocol is taken into consideration i.e. AOMDV. Various parameters are compared which results in improvement of performance of the overall network. This protocol uses the messages in between the nodes to maintain the hop information which sometimes results in flooding. To sort out this issue as future enhancement the QoS parameters can be incorporated.
R. Sangeeta and S. Padmapriya
Springer Singapore
Kowsalya N, Lavanya S, Sri Madhava Raja N, and S. Padmapriya
IEEE
Due to various reasons, most of the humans are suffering due to several brain disorders. If the brain illness is identified in the pre-mature level, it is possible to suggest treatment process to cure/control the disease/disorders. One of such disease is schizophrenia. It is a brain disorder which can affect the regular behaviour of a person. The symptoms of schizophrenia are problems with thought process, decision making and emotional instability. Signal (EEG) and images based techniques are used to identify the schizophrenia. The proposed work aims to detect schizophrenia using brain MRI images. The work executes a series of image processing techniques including pre-processing, segmentation and performance validation. The experimental results implies that, proposed method provides improved results on the MRI images.
S. Padmapriya, R. Valli, and M. Jayekumar
IEEE
Vehicular Adhoc Networks (VANETs) ensures road safety by communicating with a set of smart vehicles. VANET is a subset of Mobile Adhoc Networks (MANETs). VANET enabled vehicles helps in establishing communication services among one another or with the Road Side Unit (RSU). Information transmitted in VANET is distributed in an open access environment and hence security is one of the most critical issues related to VANET. Although each vehicle is not a source of all communications, most contact depends on the information that other vehicles receive from it. That vehicle must be able to assess, determine and respond locally on the information obtained from other vehicles to protect VANET from malicious act. Of this reason, message verification in VANET is more difficult due to the protection and privacy issues of the participating vehicles. To overcome security threats, we propose Monitoring Algorithm that detects malicious nodes based on the pre-selected threshold value. The threshold value is compared with the distrust value which is inherently tagged with each vehicle. The proposed Monitoring Algorithm not only detects malicious vehicles, but also isolates the malicious vehicles from the network. The proposed technique is simulated using Network Simulator2 (NS2) tool. The simulation result illustrated that the proposed Monitoring Algorithm outperforms the existing algorithms in terms of malicious node detection, network delay, packet delivery ratio and throughput, thereby uplifting the overall performance of the network.
M. Akshaya, R. Nithushaa, N.Sri Madhava Raja, and S. Padmapriya
IEEE
Back Propagation Network is the most commonly used algorithm in training neural networks. It is employed in processing the image and data to implement an automated kidney stone classification. The conventional technique for medical resonance kidney images classification and stone detection is by human examination. This method is not accurate since it is impractical to handle large amount of data. Magnetic Resonance (MR) Images may inherently possess noise caused by operator errors. This causes earnest inaccuracies in classification features/ diseases in image processing. However, usage of artificial intelligent based methods along with neural networks and feature extraction has shown great potential in extracting the region of interest using back propagation network algorithm in this field. In this work, the Back Propagation Network was applied for the objective of kidney stone detection. Decision making is carried out in two stages: 1.Feature extraction 2.Image classification. The feature extraction is done using the principal component analysis and the image classification is done using Back Propagation Network (BPN). This work presents segmentation method using Fuzzy C-Mean (FCM) clustering algorithm. The performance of the BPN classifier was estimated in terms of training execution and classification accuracies. Back Propagation Network gives precise classification when compared to other methods based on neural networks.
B Vijayalakshmi, S Anusha, S Padmapriya, C Ramkumar, S Prasanth Bharadhwaaj, and R Priyanka
IEEE
SLEEP APNEA is a potentially serious disorder associated with pausing or stopping of breathing repeatedly. The monitoring of sleep apnea and its detection is very important for the society as it aids in improvement of health and also causes decrease in mortality rate. The current technologies in order to diagnose OSA requires the patients to undergo Polysomnography (PSG), a very complicated and invasive test method to be performed in a specialized center which involves many sensors and wires. Accordingly, each patient is required to stay in the same position throughout the duration of one night, thus restricting their movements and causing disturbance in sleep patterns. This paper proposes an easy, very cheap, and portable approach for the monitoring of patients with OSA using IoT (Internet of Things). The project concludes with highlighting the pros and cons of the current technologies of the current technologies which can set a map for researchers and clinicians in this developing field of study.