Second International Conference on Networks and Advances in Computational Technologies

Malicious software has had a massive impact on the current digital world. Malware attacks have been skyrocketed in numbers in the last decade. Among different malware variants, ransomware became very popular as it locks up the victim’s system and demands ransom payment for regaining access. The growth of malware occurs exponentially, but the mitigation strategies are not very successful as there are still stories of attacks in the world. To discriminate between new unknown threat and threat caused by mere variants of known malware, four different methods discussed in this paper are Signature based detection, Behavior based approach, Honeypot based approach, and Hybrid approach. The quality of the detection mechanism is determined by the technique it uses. It is necessary to study different techniques and to understand its strengths and limitations. This survey examines through various techniques that are used by the industry in mitigating the same.

Online product/service reviews are mostly used by people for their purchasing decisions. Recently, the review system has contained fake reviews that are written by people who are hired by an organization either to promote their product or to demote their competitors’ product. Nowadays, these fake reviewers work in groups to magnify their activities and to gain maximum profit. These review spammer groups occur more frequently and are even more harmful than individual review spammers in current review spamming scenarios. To detect such review spammer groups, several pieces of research have been carried out. This survey introduces a taxonomy of review spammer group detection techniques and evaluates the performance of each technique.

In every single day, a large number of digital images are uploading on Internet, but the authenticity of these images is uncertain. The propagation of the fake images via Internet may cause unwanted political and social unrest. So it is very essential to validate the genuineness of the digital images. Conventionally images are manipulated by copy move or image splicing operations and there are effective forgery detection methods based on machine learning techniques are available for identifying these types of image forgeries. Recently, a new kind of fake images called deepfake images are generated using generative adversarial networks (GANs). These deepfake images are more dangerous due to its realistic appearances. So in this paper, we review various methods to detect deepfake images generated by GANs.

Code coverage assessment is an essential activity for ensuring the testing adequacy and qualification of any critical software. Flight software in Indian Space Research Organization (ISRO) launch vehicles is one such critical software element, where every functionality and portion of software has to be tested for clearing the software for a mission. The integrated system-level tests are very important for the qualification of the flight software. Since software testing is one of the time-consuming steps in software development life cycle, it is necessary to have automated coverage analyzers that can help in finding weakly and frequently executed code segments. They can be utilized for designing adequate number of test cases and optimizing the software for performance improvements. The automated coverage analyzers instrument the code for retrieving coverage information during execution of a test case. Coverage assessment of launch vehicle flight software in the integrated system test is a difficult task as code instrumentation affects the real-time performance of the system and the integrated tests fail to meet its objectives. Hence the software with instrumentation is executed in a standalone mode making use of the inputs tapped from the integrated tests without considering the real-time performance of timing but matching the functional outputs. Here open-source tools are used to carry out the code instrumentation. The end-to-end development carried out for assessing software coverage information of the integrated system test is described in this paper.

Internet censorship exists in more than 65 counties of the world including India. The government of these countries is closely monitoring the internet activity of its citizens though they publicly deny it. India uses selective filtering to safeguard its political, social, and security interests. Indian authorities realized the importance of censorship only after Mumbai bombings of 2008, which was organized with the help of internet. There is no fixed criteria or rule to censor the internet. However, removing/blocking sensitive content has become more common in recent years. This type of censoring was not completely successful as citizens found various ways to bypass them including alternate DNS servers, mirror/proxy websites, and virtual private networks. Irreversible damage would have already occurred by the time authorities completely remove malicious content manually. Removing entire malicious content manually takes time within which irreversible damage would have occurred. This paper proposes a censorware based on machine learning. It uses a modified version of the Naive Bayes algorithm to analyze captured packets and censor it. It is designed as a continuous learning tool to adapt itself to new types of circumvention techniques. The development is still in early stages. The results achieved demonstrates the effectiveness of proposed techniques compared to traditional methods. This improved version of censorware will be more helpful to the government as well as the citizens.

Sentiment analysis, also known as opinion mining, is a new emerging trend in understanding the emotions and sentiments of various people in various situations in their day-to-day life. Social media analysis, or social network analysis, is a hot topic nowadays where the information passed by the users will be used for the analysis procedure. The evolution of various natural language processing algorithms or tools also helped in a positive way. Basically, the data are classified into binary classification (positive and negative) and ternary classification (positive, negative, and neutral). Moreover, in multiclass classification, the data are divided into various classes based on the polarities by various classification algorithms. Emoticons also follow the same pre-processing, classification, and analysis steps as text.

Distributed Denial of Service (DDoS) ranks among the top ten threats to the cloud computing environment. DDoS mainly targets limited resources of cloud like bandwidth and CPU thereby denying access to legitimate clients. DDoS attacks are initiated by a vast network of remotely controlled nodes called zombies. New forms of DDoS are invented every day. Therefore, DDoS preventive measures do not fully guarantee its mitigation. Detecting an attack and defending it as early as possible is critical for reducing the attack impact. The real solution to mitigate any attack is tracing back the attacker and punishing him. However, a real attacker will masquerade his identity using a spoofed address to avoid being traced back. The routing mechanism used on the internet does not have any memory of its own making traceback further difficult. Many businesses are reluctant to enter the cloud due to these DDoS vulnerabilities of the cloud. DDoS will affect network performance and may disrupt configuration information available in the system. In the event of DDoS, businesses will have to suffer reputation damage, customer agitation, and legal repercussions. Unless cloud is made secure, we cannot benefit from its full potential. Research on DDoS attacks and defense is in its infancy. DDoS defense and traceback is still an open and challenging problem. This paper presents basic types of DDoS and focuses more on DDoS prevention, detection, and traceback techniques.

Research management in academia and industry is a field which has so many opportunities due to the advancement in information and storage technologies. It also hoards as much challenges as that of the opportunities. Like academic collaborations through co-authorship, collaboration of inventors through co-inventorship of patents exists in almost all industrial areas in varying degrees. Understanding the nature of inventor collaboration network is vital for many applications including policy related ones. Identification of suitable inventors for collaboration will be decisive for inventors in different phases of their career, making collaboration management one of the concerns of research management. In this work, through a network based predictive approach, a preliminary design of an inventor collaboration recommendation system is proposed. As a case study, patent data from ‘Wireless Power Transmission’ is analysed and various implications are discussed.

A WordCloud by definition is a computer visualization technique used in the text mining methods of document summarization. WordClouds being simple and user-friendly are widely used in various real-life applications irrespective of the domain. The key design principle behind wordclouds is to project the most prominent keywords from a given raw text or document on a visual canvas. The prominence of each keyword is proportionately depicted using the size it occupies on the canvas. The traditional random wordclouds have the keywords being positioned at random locations on the canvas irrespective of the word semantic relationships. On the contrary, semantic wordclouds are an enhanced version of random wordclouds, where the words are positioned on the canvas by maintaining the appropriate relationships among them. In this article we provide a tutorial on the architecture and design components to generate such a system. Such an overview can be really helpful for aspiring researchers in the prerequisite analysis of semantic wordclouds.

Education has always been a strenuous task for visually impaired students. Courtesy to Braille’s system for being a primary technique for blind education since 1824. In this paper, we propose a technology-aided system – named THIRD EYE: A System to Help the Visually Impaired Students in Academics, which functions on voice commands especially designed to help visually impaired students. This system will convert the text in text books into audio files using text-to-speech conversion technique. It also converts the textbooks in English languages to audio in Malayalam language and sets up a digital library to store these converted audio files, which will help the students in their studies. This also reduces human efforts and also proves to be quite handy in case of change in syllabi. This system also helps the students to write their examinations without the help of a scribe, by automatically recording the dictated answer and converting the speech into corresponding text format. A provision to replay the answer is also provided to make corrections if required. This is an android and web-based application, which requires a headset with microphone as an external device. The system proposed will mark a beginning of new hope and an emancipation from the above-said requisites.

Cloud computing technology allows us to outsource our data to the cloud. The main advantage of this technology is its storage capacity. To maintain privacy, everything should be encrypted before outsourcing into the cloud and only authorized users can retrieve their data. But this process is complicated in the case of encrypted data, because traditional data storage system works only on plain data. Therefore dedicated information retrieval systems were developed to handle the encrypted data. Different search methods over cloud data have been proposed for efficient information retrieval. The main objective of this paper is to provide an overview of some of the existing searching techniques on the encrypted data and make a comparative study.

The avionics system, a critical part of a launch vehicle, consists of varied hardware and software systems. Thorough validation of the onboard software is essential to ensure complete success of missions. In this work, a part of the onboard software, the scheduler, is represented in the SPIN model checker. Different tasks in the software are modeled with minimal data transfer and error checking. Assertion statements are inserted in the model to check for setting of error flags. On simulation and verification, the assertion was found to be violated in one case. The error trail given by the tool is a realistic execution sequence that cannot be detected through traditional testing techniques. This effort thus establishes the effectiveness of formal verification in early detection of errors in mission-critical embedded software.

The infant mortality rate is the number of newborns death under 1 year of age occurring among the live births in a given region during a given year. Electrocardiogram is generally used for finding the cardiovascular variation. The infant mortality rate can be drastically reduced by adopting this proposed diagnosis technique for fetus. This proposed work provides an indication of fetal health and heart information. Sometimes newborn babies will be affected by heart diseases like tachycardia and bradycardia. By diagnosing these diseases, we could reduce the death rate of the newborns. This method proposes an early detection of fetal ECG and the arrhythmia of the fetus. So this will dynamically reduce the infant mortality rate. It brings amazing changes in the fields of medical industries and medical research fields. The main aim of this work is to detect the variations in the heart rate. The variations in the heart can be detected by using feature vector that is extracted from the signal and a classification method is also used to classify the signal as abnormal and normal. This work shows a result of accuracy about 94.11% and sensitivity of 88.88%.

The technique of data hiding reversibly in encrypted images may be considered as an efficient method that can be used to hide data within an image that is being encrypted. In prevailing methods of data hiding, as in any other encryption standards, an image that has to be transmitted is encrypted first. This encryption is done by the sender by using any encryption key that is unknown to anyone accessing the image. It can be possible to introduce or cover information within the image that has been encrypted. This can be done without knowing any information about the image or about its encryption. On the receiver side, a recipient can decode the image if he has the encryption key or can retrieve the hidden message if he has the data hiding key. The methods must be able to perform a high payload compensating a high quality of image that has been reconstructed. Here various techniques that use reversible data hiding are reviewed in this paper and this leads to bitwise prediction data hiding.

Data hiding and encryption are the two successful techniques to protect the information in public domain while transmission. Conceal information in a cover image and reconstruction of the same cover image after extracting the hidden information without any alteration is recognized as reversible data hiding (RDH) technology. Reversible data hiding is mainly found application in military and medical field applications. Many such types of schemes are technologically advanced to ensure the security, genuineness, and integrity of the images on transmission through Internet or through any other medium. The cover image used here is a gray image and it is preprocessed for prediction error detection. After the detection of possible prediction error pixels, the position of the predicted error pixel is stored in a binary location map. For prediction error detection the median edge detector (MED) is used. By using information in the binary location map, the error position information is inserted to the encrypted image by most significant bit (MSB) substitution method. Chaotic generator used in this work for encryption process is two chaotic logistic map, which is used to produce pseudo-random binary sequence. The encryption algorithm used here is Coupled Chaotic Logistic Map, to generate pseudo-random binary sequence to encrypt the image pixel by pixel. The secret message is embedded into the prediction error highlighted encrypted image by MSB substitution method. MSB prediction method is used in the decoding phase as it is essential for the successful reconstruction of cover image. The embedded image can be productively extracted and the original cover image can absolutely recovered and its PSNR and SSIM value are infinity and 1.00, respectively. By using the coupled logistic chaotic map, high visual security, better NPCR, UACI, and correlation values are achieved.

Telemedicine is the medical technology that enables patients from villages and rural hospitals to get expert advice from specialist doctors located anywhere in the world. These systems work primarily by sharing audio-, visual-, and text-based information over suitable communication medium such as the internet. Close examination camera is a handheld device used to examine body parts such as skin, eye, etc. It is usually operated by a nurse or a doctor. It is seen that due to lack of good focusing system and/or image stabilization techniques, the output from these handheld devices may not always be of acceptable quality. Motion blur and out-of-focus blur are commonly seen issues, particularly if the nurse operating the device is older in age. This can result in doctor being unable to make diagnosis or even worse—make wrong diagnosis. Due to this reason, quality detection of the transmitted images is quite important. For selecting the best image, we need to rank all the images that are taken by the camera in a given duration based on a Quality Metric. This paper describes a system that will automatically analyze the quality of the acquired images and then transmit and present the visually good images to the remote specialist’s side. Proposed method provides high correlation with human judgment when compared on the same set of images. Real and simulated databases were used to verify the correlation.

A 32-bit successor has been designed for the indigenous 16-bit processor used in the on-board computers of the operational satellite launch vehicles of ISRO. Significant improvements have been made in the areas of computational power, memory addressing capability, registers and data types. The instruction set is tailored for optimising performance in space-based applications. Verifying that the design is flawless prior to tape-out is a challenging task for the microprocessor design team. Zero tolerance of bugs is a pre-requisite for the processor, given the unforgiving nature of space applications. This paper describes the exhaustive validations carried out for evaluating the design of the 32-bit processor in the designer lab.

Potential of complex network analysis to address complex systems such as stock markets is steadily gaining recognition. In this study, an approach for data mining of stock market based on complex networks is done as a preliminary for development of stock recommendation and/or portfolio management systems. Lobbying power of players is identified based on unweighted and weighted stock market networks that are created from United States stock data dynamics. Also, a criterion to check whether strength of correlation can significantly impact the assessment of local influence and lobbying power of players is devised. Portfolio analysis based on lobbying power and weighted lobbying power is carried out to reveal crucial industrial sections of the market. Our study revealed the affordability of offering financial services for firms belonging to such industrial sections for systemic risk reduction. Weighted lobby analysis is found to reveal in-depth structural insights for portfolio analysis than lobby centrality.

Most researches done in the area of wireless communication is in vehicular ad-hoc networks (VANETs). Traffic accidents in roads are marked as the 9th major reason for loss of humans lives in the world. For solving this issue, VANET is developed and it forms the pillar of Intelligent Transportation System (ITS). Transportation efficiency and road safety are the main applications of VANET. For these applications, an efficient routing protocol is required in order to handle the high mobility and rapidly changing network topology. Many routing protocols have developed based on topology and position. VANET routing protocol also varies according to the area of communication, rural or urban. Since VANET is large in terms of amount of information generated and number of nodes, geographic routing protocols are appropriate for urban road network. But the main forwarding strategy applied in geographic routing is to select nodes that are nearer to the destination which can lead to increase in the probability of link breakage. Hence fuzzy logic based geographic routing protocol is most appropriate because it uses artificial intelligence in selecting the route. This paper focuses on the geographic based routing protocols and the geographic routing protocol based on fuzzy logic developed in urban scenarios.…

Security of data before adopting into the cloud is important because of the security challenges in clouds. The data security of the cloud server must be analyzed before implementing new security features. The data of the cloud owner have problems of attacks externally as well as internally. These attacks can be addressed, and tampering of server data can be handled to ensure confidentiality and integrity in cloud data. The proper use of cryptographic algorithms makes the cloud owner and the data on the cloud premises secured. This paper analyzes how data can be securely stored in the cloud using hybrid cryptographic technique. It also describes how to handle the cloud data security issues. The performances of the proposed algorithm are analyzed with other similar algorithms.

This paper presents a deep learning based approach to Malayalam Parts of Speech (POS) tagging. We applied two neural sequence labelling models long short-term memory (LSTM) and Convolution Neural Network (CNN). The proposed model is an end-to-end deep neural network and that benefits from both word and character level representations. We have studied the performance of a six different combinations of neural sequence labelling models on the ILCI Phase II Malayalam dataset and achieved accuracy up to 87.05% for POS tagging. The proposed Word LSTM model with character LSTM and Softmax gives little improvement than character LSTM and Conditional random Field (CRF) models. Also we demonstrated the effect of word and character embeddings together for Malayalam POS Tagging. The proposed approach can be extended to other languages as well as other sequence labelling tasks like Chunking and Named Entity Recognition, etc.

Global System for Mobile communication (GSM) phones use A5/1 stream cipher for secure communication. It encrypts the information transmitted from a mobile user. Initially, we have studied all existing attacks on A5/1 cipher and performed a comparative analysis. In this paper, we propose an alternative approach for constructing rainbow tables for cryptanalysis of A5/1 algorithm. These pre-computed tables are used to construct the initial state of A5/1 cipher which can be used to recover the session key. These two methods are parallelized and implemented on 5 TeraFlop High Performance Computing (HPC) facility.

The advent of opportunities to shop online has made inroads into virtually all modes of businesses worldwide. The success and future of online businesses depend on how well they adopt newer technologies. The online platforms are presented with a much faster and suitable option to buy, but have not been able to deliver the experience of touch and feel shopping to its customers. Technologies such as Augmented Reality (AR) steps in this direction—bringing an entirely new aspect to the online selling platforms. Our project presents a smartphone-based augmented reality application which makes it possible to enhance the whole-in store shopping experience with low cost, easy to use smartphone-based augmented reality platform.

The task of conversion of grayscale images into colorful ones is a complex problem because the same intensity of light can correspond to different colors in three channel color models. This usually requires manual modifications to attain artifact-free quality. This paper looks into this problem and aims at the conversion of grayscale image to colorful ones. In the proposed approaches, support vector regression (SVR) is trained upon the features using SLIC and fuzzy c-means algorithm. Scores of trials and tests demonstrate that our algorithm gives an excellent performance in terms of quality, speed, and several feature learning benchmarks.

Time-series data has generally been difficult to predict, due to unseen hidden patterns in the data and the unpredictability of values in them. Classical methods such as autoregressive moving average (ARMA), autoregressive integrated moving average (ARIMA), and artificial neural networks (ANNs) have been used to obtain predictions. However, the accuracy obtained was distorted due to significant misprediction rates. Deep learning, a part of machine learning, has recently seen a surge in usage to obtain accurate predictions from time-series data. Areas such as stock market, petroleum production, solar energy production, electric load, etc. make use of deep learning. Recurrent neural networks are used in most cases due to their ability to recall past information and apply it for future predictions. Variations of classic RNNs such as long short-term memory (LSTM), gated recurrent unit (GRU), and modifications of the previous two networks with other types of networks have been observed. A survey of usage of variations of RNNs and their modified versions has been conducted to understand the strengths of each type of RNN and to find out which type of RNN is the most versatile.

Social media has become a prominent source for information in multiple business domains as it provides a true and real-time reflection of societal inclinations. The tourism domain possesses multiple societal activities in the form of travel blogs, reviews, forums, feedbacks, etc. This information can be leveraged in multiple contexts like creation of travel recommender systems, tourism marketing, targeted advertisements for tourists, etc. Personalized travel recommender systems have become an essential part of the travel and tourism domain and also a key research area as it is still maturing. Conventional travel recommender systems based on collaborative filtering techniques prove to be computationally intensive and lack scalability against the social media big data. Considering the exponential growth and high dimensions of social media data, it is very much required to have efficient mechanisms in place for information retrieval. The adoption of an appropriate clustering algorithm and dimensionality reduction technique is considered to be the solution to deal with such a scenario. This paper proposes a loosely coupled, technology agnostic framework that can be adopted to build a tourism recommender system by taking social media big data as an input data source.

Phishing is a common social engineering attack performed by attackers, wherein they try to benefit from the unaware users by tricking them to reveal sensitive data (like credit card details, email passwords). A common phishing detection method uses comparison with a list of websites stored in some database, which cannot be accurate always, as they get outdated sometimes. Machine learning methods can make a system more reliable. The websites can be identified using different parameters. In this paper, we are trying to investigate the main features that can help to differentiate between phishing websites and genuine websites and develop three classification models using SVM, Random Forest and ANN that will try to predict phishing websites.