International Conference on Innovative Computing and Communications

Table of Contents

1. Detection of Epileptic Seizure Using a Combination of Discrete Wavelet Transform and Power Spectral Density

   Puja Dhar, Vijay Kumar Garg

   Abstract

   Epileptic seizure is detected by reading the electroencephalogram (EEG) signals which are obtained from the electrical activities of the brain which are containing information about the brain. Epileptic seizure is known as the abrupt abnormal activity of a bunch of neurons which results in an electric surge in the brain. India is also one of the countries on the globe which is having about 10 million people suffering from a seizure. In this paper, the combination of discrete wavelet transform along with power spectral density is proposed for the classification and feature extraction process to detect epileptic seizures. To achieve high accuracy of seizure detection rate and explore relevant knowledge from the EEG processed dataset, deep learning has been used. The result shows that the detection of epileptic seizures using the proposed method gives an accuracy of 90.1%. This system would be useful for clinical analysis of epileptic seizures, and appropriate action would be taken against epileptic seizures.

2. Combination of Oversampling and Undersampling Techniques on Imbalanced Datasets

   Ankita Bansal, Ayush Verma, Sarabjot Singh, Yashonam Jain

   Abstract

   Many practical classification datasets are unbalanced, meaning that one of the classes is in the majority when compared to the others. In various real-world circumstances, class-imbalanced datasets arise, where the number of data samples in a class is not equal to the other class. To develop good classification models based on present level calculations, using these datasets is difficult, particularly for separating minority classes from the majority class. To address the issue of class imbalance, under/oversampling procedures are used to minimize and enhance the quantities of data examined in minority and majority class. This paper explores the utilization of combination of both undersampling and oversampling techniques mainly synthetic minority oversampling technique (SMOTE) and neighborhood cleaning rule (NCL) to balance the datasets. The performance has been evaluated using two machine learning algorithms. The results are then classified using recall measure and geometric mean which showed improved performance of the algorithms.

3. Comparative Analysis on Effect of Different SVM Kernel Functions for Classification

   Deepali Virmani, Himakshi Pandey

   Abstract

   Besides linear classification, Support Vector Machine (SVM) is proficient in non-linear classification by deploying kernel tricks that implicitly maps and transform input features to high dimensional feature space. Kernel-SVM, can be utilized to secure progressively complex connections on datasets with no push to do changes all alone. In this paper, 5 different SVM kernel functions are implemented on 4 datasets, viz., IRIS, Breast Cancer Wisconsin (diagnostic), Mushroom and Letter Recognition Dataset. The five kernel functions considered in this paper are: Linear kernel, Gaussian Radial Basis Function (RBF) kernel, Laplacian kernel, Polynomial kernel and Sigmoid kernel. Our goal is to locate the best non-linear kernel. The outcomes show that the precision of expectation for Laplacian kernel is most extreme with a forecast scope of (max 100%, min 97.53%) and least for the sigmoid kernel with a forecast scope of (max 100%, min 47.28%).

4. Two-Phase Image Denoising Using Hough Transform

   Shaveta Rani, Yogesh Chhabra, Kamal Malik

   Abstract

   Advancement in applications requires more efficient image denoising techniques as it is still an unexplored area for the researchers. The crucial step in image denoising process is detection of noisy pixel, where over and under detection may affect the desire outcomes. Random impulse noise is difficult to remove because it appears randomly on the image, and most filters fail to identify all damaged pixels. To remove random impulse noise, we proposed two-step sequential algorithm, in the first step, ROAD-TGM filter ensure accurate noise estimation by avoiding over and under selection of noisy pixels. The second step performs image restoration using Hough transform method. Hough transform technique is used to extract features form images and by using these features, we restore the damaged pixel by the mean of undamaged pixel. In comparison with other well-known methods, the results of proposed method indicate that our restored images show a substantial change.

5. Modern Four-Port MIMO Antenna Design Using Bended Curves for 5G Communications

   Kolli Venkatrao, Yadavalli Sai Sundara Sriramam, N. Suguna, Nalini Prasad Tirumani, Ch. Rama Krishna, Ch. Murali Krishna

   Abstract

   In this paper, modern four-port Multiple-Input Multiple-Output (MIMO) antenna design using bended curves for 5G communications is presented. The overall dimensions (L_sub × W_sub × H) of the antenna are about 19.5 mm × 19.5 mm × 1.575 mm on Rogers RT/Duroid 5880 substrate with dielectric constant of 2.2 and loss tangent of 0.009. A similar four-element MIMO antenna is designed, and to enhance the parameters of the antenna, circular arcs are loaded. The proposed design can obtain a frequency range from 27.65 to 36.98 GHz which supports 5G communication. The proposed antenna covers an impedance bandwidth (IBW) of 9.33 GHz, and the resonant frequency is 34.60 GHz. The overall peak gain of the antenna is 6.93 dB. Some of the antenna parameters such as envelope correlation coefficient (ECC) < 0.0004, diversity gain (DG) ≈ 10 dB, total active reflection coefficient (TARC) < 0.38, channel capacity loss (CCL) <0 .2 bits/Hz and mean effective gain (MEG) is varied in between − 2.80 to − 4 dB. By using HFSS software, the proposed design is designed and simulated.

6. Supervised Question Classification on SelQA Dataset Using Variational Quantum Classifiers

   Pragya Katyayan, Nisheeth Joshi

   Abstract

   Machine learning and quantum computing fuse together to form quantum machine learning. Although the phenomenon is new, it has already proved its worth in various fields like finance and chemistry. The potential of quantum computing and its extraordinary properties enable us to process data in a way classical computer can never think of. When machine learning gets the power of quantum computing, information processing is enhanced significantly. In this paper, we have used variational quantum classifiers to classify questions from two domains of SelQA dataset. We keep the focus on the implications of circuit-depth in different experiments and analyze the results. VQC performs well with 11 features on lowest circuit depths and gives a testing accuracy of 58%.

7. SMOR-Smart Mirror for College Department

   Deepak Sharma, Abhishek Khanna, Devesh Chaudhary, Anjali Jain, Archika Malhotra, Aayushi Rohilla, Risheek Kumar, Anuradha Bhasin

   Abstract

   This paper aims to present SMOR—Smart mirror for College department. Smart mirror which has mirror-like reflective properties but display information in the form of a widget. SMOR is designed using Raspberry Pi, LCD screen, two-way mirror, and ultrasonic sensor. It contains a voice activated chatbot system which can be easily customized. It is designed specifically for the college department as it will show all the information related to that department like schedules, notices, announcements, etc. It fetches data from the cloud database and displays it according to the request. It also displays current weather, date, time, and updated news about science and technology.

8. Food Classification Using Deep Learning Algorithm

   R. V. Jamnekar, R. R. Keole, S. W. Mohod, T. R. Mahore, Sagar Pande

   Abstract

   Monitoring of food plays a significant role in leading health-related issues and tasks. With its multiple applications and features, image processing emerges to be an interesting field in the process of identifying food items. In this paper, a technique has been presented for classifying the food image using the You Only Look Once (YOLO) algorithm. Unlike the conventional artificial neural networks, the YOLO algorithm has more efficiency, and it has been trained on a loss function that corresponds straight to detection, and the complete model is trained with 6000 epochs. Due to the high variance in the alike domain of food images, food classification becomes a difficult task but it has a significant role in lives at the present time as it can be utilized by numerous sources. In this paper, a comparison of the working of the YOLO algorithm with other techniques that are used in image processing such as ResNet-50, VGG-16, ImageNet, and Inception has been elaborated. In this work, the famous dataset from Kaggle is used for implementation purposes. The dataset consists of 4000 Indian Food Image 80 different categories or classes. The proposed model is giving 99% accuracy for classifying the food.

9. Applying Machine Learning Algorithms on Urban Heat Island (UHI) Dataset

   Mujtaba Shafi, Amit Jain, Majid Zaman

   Abstract

   Climate change worldwide is a huge challenge, and urban heat island (UHI) is being explored as one of the contributors to this challenge. UHI is an urban or rural area with a temperature variance than its neighbouring areas. Researchers can model the UHI data and predict the temperature change using various relative parameters of UHI. The land surface temperature (LST) data and its co-related parameter of the study area, i.e. Srinagar City, JK, India, has been extracted from satellite imageries. LST data of the study area is assessed to understand the evolution to help analyse the UHI effect and its variance. The LST data was extracted through MODIS Satellite, from 2001 to 2020, with an 8-day revisit time/peak month of the season. In having a voluminous dataset, i.e. 16 sampled LST data/each km²/year measured in Kelvin(k), various machine learning algorithms were applied on LST data to establish relations for UHI modelling. Unsupervised machine learning algorithms were used on continuous LST data to define clusters and further standardized/compared with existing scientific classifications of the study area. The number of clusters was tweaked to determine the best-case scenario. Additionally, correlation and regression were applied to determine if there is multicollinearity amongst the LST data. The outcome of two analyses was used to build a UHI framework on a structured UHI dataset. Performance of algorithms in predicting UHI parameters like urban, vegetation and wetlands zones varied considerably. Naive Bayes and support vector machine did considerably well in predicting wetlands but failed to perform impressive accuracy for urban and vegetation zones. Random forest, gradient boost tree and probabilistic neural networks failed in predicting wetlands. Neural networks have performed worst in predicting wetlands, having a prediction accuracy of around meagre 5%, while the decision tree algorithm has performed well in all three zones.

10. A Novel DDOS Attack Detection and Prevention Using DSA-DPI Method

    V. Deeban Chakravarthy, K L. N. C. Prakash, Kadiyala Ramana, Thippa Reddy Gadekallu

    Abstract

    In the current Internet world, connection of computers, IoT devices, and mobile devices together becomes common activity. Because of the enormous advantages available with the Internet, many applications are connected to it even without the proper authentication from the user end. The same activity happens at the public network also enable the user device get hacked by the third-party attack holders. Distributed denial of service (DDoS) attacks act as the one of the common malfunctions happen in the systems. Detection of such attack and defending mechanism against it is much more important. Software defines networks have the facility to configure the network platforms with the preventive measures from the DDoS attacks. It is mandatory to design a preventive system for DDoS attacks and developing an analysis module to test the pattern of activity happens during the attack is important. The proposed system is focused on implementing such module that detects and prevents the DDoS attacks over the Internet. DDoS is the type of attack that overloads the firewall by unwanted malware scripts. The system provides the robust preventing mechanism called digital signature algorithm (DSA) collaborated with deep packet inspection (DPI), together called as DSA-DPI model to prevent the DDoS attacks. Our proposed design provides preventive alters on infrastructure before the malware attack get happens.

11. Dynamic Decentralized Group Signature Scheme for Privacy Protection in Blockchain

    S. Devidas, N. Rukma Rekha, Y. V. Subba Rao

    Abstract

    Group signature schemes can play a key role in privacy protection of blockchain-based applications because of its security properties like unforgeability, anonymity, unlinkability, traceability, etc. But the malicious group manager may collude with some group members leading to biased decisions. This issue was addressed by Devidas et al. [12] with a static decentralized group signature scheme(DGSS) by decentralizing the group manager. However, the limitation of DGSS is that it works only for static domains, where group members are fixed and will not allow new members to join and existing members to leave the network. In this paper, DGSS is extended to propose a dynamic decentralized group signature scheme which allows the group members to join and revoke at run-time. The performance of Devidas et al. [12] scheme is also improved by reducing the number of multiplications to make it suitable for user identity privacy protection in lightweight blockchains or memory constraint devices. The security analysis and proof of correctness for the proposed scheme are also discussed in this paper.

12. An Adaptive Scheme for Detection of Attack in Energy-Aware Dual-Path Geographic Routing (EDGR)

    M. Sridhar, P. B. Pankajavalli

    Abstract

    The geographic routing (GR) is one of the wireless sensor network (WSN) routing protocol. During routing, it is susceptible to diversified attacks, namely wormhole and blackhole attacks, which are very hard to identify and defend. In the context of wormhole, intruder overhears the information passed over the transmission area, and in the scenario of blackhole attack, information can be reprogrammed to block exchange of information. If any data transmit through attacked section that makes failure of transmission with huge delay in delivery and drop in packet. To identify the attack and mitigate, an adaptive scheme is initiated into the network, and it enriches the process of data transmission. In this research, attacked area is investigated by the adaptive approach using energy-aware dual-path geographic routing (EDGR) protocol. The performance of the network recovery approaches investigated by simulation, and it is identified that the proposed scheme shows promising performance by attaining high minimal delay and packet drop.

13. Backmatter