Search

Ciphertext Policy Attribute-Based Encryption (CPABE) has gained popularity in the research area among the many proposed security models for providing fine-grained access control of data. Lightweight ECC-based CP-ABE schemes can provide feasible selective sharing from resource-constrained devices. However, the existing schemes lack support for a complete revocation mechanism at the user and attribute levels. We propose a novel scheme called Ecc Proxy based Scalable Attribute Revocation (EPSAR-CP-ABE) scheme. It extends an existing ECC-based CP-ABE scheme for lightweight IoT and smart-card devices to implement scalable attribute revocation. The scheme does not require re-distribution of secret keys and re-encryption of ciphertext. It uses a proxy server to furnish a proxy component for decryption. The dependency of the proposed scheme is minimal on the proxy server compared to the other related schemes. The storage and computational overhead due to the attribute revocation feature are negligible. Hence, the proposed EPSAR-CP-ABE scheme can be deployed practically for resource-constrained devices.

Kotlin Coroutines are generalised subroutines that can suspend and resume execution freely. They exhibit lower overhead than Java Virtual Machine (JVM) Threads, making them suitable for asynchronous applications in resource-constrained environments. This paper analyses the performance of Kotlin Coroutines framework on Android and measures its benefits over JVM Threads-based concurrency framework RxJava using a novel benchmarking methodology for reliable and reproducible results. It also highlights the performance improvements enabled by Kotlin Coroutines in real-world projects by comparing two Model-View-Intent (MVI) architecture libraries MvRx (by Airbnb), and Vector (by author Chauhan, K) based on RxJava and Coroutines, respectively.

Yogic visual concentration techniques have proved to create a long term effect on mindfulness in adults. However, none of the existing gaze detection platforms actively monitor attention and provide feedback to the users. This paper presents a novel method to correlate and estimate the effect of a short-term candle gaze concentration technique on the sustained attention, by providing feedback to the participants and helping them keep track of their sustained attention. We present the prototype of an Android mobile-based flame gaze application known as mFlameGaze to perform the candle flame test (CFT). The app determines for how long the participants focus on the candle tip in their screens. It also detects the moment in time when the eyes first deflect from the flame. The proposed work uses this app to analyze the impact of candle gaze on sustained attention for a small group of normal, healthy children. The participants’ sustained attention is measured before and after conducting the study using a letter cancellation test (LCT) as the benchmark. The results indicate an increase in the performance of participants in the LCT and also an increase in the focused gaze time, as measured by the mFlameGaze mobile app in a short duration of ten days.

Weeds are a dangerous factor for a good yield of crops. The conventional method of removing weeds was either plucking manually or spraying herbicides uniformly all over the field. Spraying herbicides not only contaminates crops but also gives rise to many health-related issues. The purpose of the paper is to develop a mobile model that can detect weeds in real- time with their position coordinates and scrape them off. The model first scans the specific area for leaf detection and classifies it as weed or crop with a prediction accuracy of 99.5%. If the classified leaf is a weed, the coordinates are found and the robotic arm removes them with the help of a high-speed rotating blade, without harming the crops and environment. The left outs can further be utilized as fertilizer and no harmful chemicals have been used.

With the ever-increasing involvement of people into social media and online gaming, Cyberbullying has become a serious issue affecting almost all parts of the demographic. Cyberbullying can cause severe mental and emotional impacts on a person, especially on minors; hence, there is a requirement of having intelligent automated systems to detect questionable content present on social media platforms and remove it. In this paper, we introduce our novel algorithm Q-Bully which can automatically detect cyberbullying on various social media and online gaming platforms using Reinforcement Learning along with Natural Language Processing techniques. Previously the techniques used to detect cyberbullying have a good accuracy related to the text they have been trained on and do not incorporate new word patterns without complete retraining of model. In this paper, we incorporated the use of Reinforcement Learning and have conducted an experimental study in which we feed the messages and posts of bullies as well as victims to a Reinforcement Learning Agent for classification. We compare our model with the other baseline models on based on F1 scores (0.86 a benchmark dataset of 16K annotated tweets) and are able to infer that our model outperforms other state-of-the-art models when the dataset is highly dynamic and populated with words which are deliberately misspelled to trick the conventional detection systems.

This work aims to investigate the performance of the Long Short-Term Memory (LSTM) Model for EEG-Based Emotion Recognition. For the experimentation, we use the publicly available DEAP dataset, which consists of preprocessed EEG and physiological signals. Our work limits itself to the study of only the EEG signals to have a scope for developing an efficient headgear model for real-time monitoring of emotions. In this study, we extract the band power, a frequency-domain feature, from the EEG signals and compare the classification accuracies for Valence and Arousal domain for different classifiers. The proposed Long Short-Term Memory (LSTM) model achieves the best classification accuracy of 94.69% and 93.13% for Valence and Arousal scales, respectively, illustrating a significant average increment of 16% in valence and 18% in arousal in comparison to other classifiers.

Pneumonia is an infection that causes inflammation of lungs and can be deadly if not detected on time. The commonly used method to detect Pneumonia is using chest X-ray which requires careful examination of chest X-ray images by an expert. The method of detecting pneumonia using chest X-ray images by an expert is time-consuming and less accurate. In this paper, we propose different deep convolution neural network (CNN) architectures to extract features from images of chest X-ray and classify the images to detect if a person has pneumonia. To evaluate the effect of dataset size on the performance of CNN, we train the proposed CNN’s using both the original as well as augmented dataset and the results are reported.

The growth of internet and users has increased exponentially and drastically in this decade. It provides services inheriting various benefits to the users such as online banking, marketing, buying/selling and various facility management services etc. It attracts some people to develop programs that perform various malicious activities intentionally or unintentionally such as stealing sensitive information from computer, displaying advertisement, causing harmful, unwanted activities. The malicious software is referred as malware, viruses, worms, spyware, trojans, adware and botnets are also malwares. Therefore, in this paper, we run a honeypot and capture various zero-day attacks and malwares. The harmful and malicious activities of these malwares are studied by analyzing the network logs of the honeypot. Based on their properties and activities, the malwares are classified. These classifications of malware and their behavior help researchers to develop a security mechanism to prevent an organization from these kinds of harmful and malicious activities triggered by the malwares.

Users can directly access and share information from portable devices such as a smartphone or an Internet of Things (IoT) device. However, to prevent them from becoming victims to launch cyber attacks, they must allow selective sharing based on roles of the users such as with the Ciphertext-Policy Attribute Encryption (CP-ABE) scheme. However, to match the resource constraints, the scheme must be efficient for storage. It must also protect the device from malicious users as well as allow uninterrupted access to valid users. This paper presents the CCA secure PROxy-based Scalable Revocation for Constant Cipher-text (C-PROSRCC) scheme, which provides scalable revocation for a constant ciphertext length CP-ABE scheme. The scheme has a constant number of pairings and computations. It can also revoke any number of users and does not require re-encryption or redistribution of keys. We have successfully implemented the C-PROSRCC scheme. The qualitative and quantitative comparison with related schemes indicates that C-PROSRCC performs better with acceptable overheads. C-PROSRCC is Chosen Ciphertext Attack (CCA) secure. We also present a case study to demonstrate the use of C-PROSRCC for mobile-based selective sharing of a family car.

A high amount of stress is experienced by people of all ages nowadays and is affecting their physical and mental health adversely. Traditionally, Perceived Stress Score (PSS-14) questionnaire is used to calculate the stress score, which requires manual intervention of people. In this work, we look into the usage of Electroencephalography (EEG) to help predict the stress level of people without manual intervention and recording by a Brain-Computer Interface(BCI) instrument. EEG can estimate and record the electrical signals in brains utilizing sensors exhibited over the scalp. We propose the design of a brain mapping based stress identification system using a single electrode EEG device. Here we have taken EEG recording of 63 students using a single electrode EEG device NeuroSky Mindwave Mobile. A novel feature combination of bandpower ratios of alpha, beta, delta and theta bands are extracted and fed to the classifiers. We evaluate the performance of Support Vector Machine and K-Nearest Neighbour machine learning algorithms. Target class of the training set is assigned using the stress score calculated from the response of the PSS-14 questionnaire. We achieve the highest average classification accuracy of 74.43 % using K-NN algorithm. We observe that there exist a correlation of bandpower ratios of different bands obtained from recorded EEG signals from the frontal portion of the brain with the stress level of subjects.

Credit card transactions have become the preferred mode of payments in developed countries and its utility is rapidly growing in developing countries making frauds an increasingly consequential problem leading to financial losses and erosion of consumer confidence. Although, credit card data is highly class imbalanced and this makes training of models to classify fraud data difficult. This study employs the use of multiple adversarial networks to generate pseudo data to enhance model performance. This study uses the vanilla implementation, Least Squares, Wasserstein, Margin Adaptive, Relaxed Wasserstein of GANs. The distribution of the generated data against original fraud data, the classifier accuracy, convergence for each model and an optimal number of data generations is analyzed. The generated data is then augmented and tested using an Artificial Neural Network model and a 12.86 % increase in recall for a dataset with a class imbalance of initial 579 to 1 is recorded.

Portfolio optimisation is a thoroughly research domain, and with developments in asset allocation objective functions, current optimisation methods are proving insufficient. This study compares the performance and convergence time of different swarm-intelligence based methods for 2 types of objective functions, each consisting of 12, 24, 48 and 96 stocks each. The objective functions are Sharpe ratio maximisation and maximization of Value at Risk weighted return of the portfolio. Cuckoo search, Firefly, Micro-Bat Echolocation, Elephant Herd, Harmony Search, Flower Pollination, Differential Evolution and Particle Swarm Optimisation algorithms are used for optimisation and analysis. The performance of each algorithm is measured by the resultant return, variance measured risk, sharpe ratio and time per iteration. The portfolios are sampled from and the return and risk is calibrated from Nifty100 historical price data.

Object Detection in recent times has become less of a monumental task with the help of cameras and other recording devices. The aim of this paper is to come up with a cost effective object detection system with support of ultrasonic sensors. The use of ultrasonic sensors prove to be a cost effective solution when compared traditional cameras for the purpose of object recognition. The readings obtained from the sensor are then used to plot a graph using dynamic graphing. Result evaluation is done by comparing the obtained curve from the ultrasonic sensor readings against the learning curve, trained using self created data set. The user is prompted with the result regarding the shape of the object which classifies the object as spherical in shape or not. The proposed system delivers the required result with higher accuracy for a fraction of the cost when compared to the conventional methods. This finds applications in various fields such as Security, Object Avoidance, and an aid to the visually impaired using text to speech.

An attempt has been made to map the potential avalanche zones of Alaknanda valley of central Himalaya by employing analytical hierarchy process (AHP) using remote sensing data in GIS environment. The AHP has been found to be an effective method for mapping of the avalanche prone areas of highly undulating mountains. Further, a numerical simulation of one of the potential avalanching site has been carried out using Rapid Mass Movement Simulation (RAMMS) model was used to determine the flow dynamics of the avalanche sites. The mapping of potential avalanche zones in combination with numerical simulation can provide promising results for prediction of natural hazards occurring in glacial and peri glacial areas.