International Journal of Innovations in Science & Technology
Not a member yet
813 research outputs found
Sort by
Recurrent Neural Network and Multi-Factor Feature Filtering for Ransomware Detection in Android Apps
Get PDFThe market is flooded with Android Software (apps), and at the same time that number is growing quickly, and so are the many security exploits that take advantage of such apps. The effectiveness of traditional defensive systems is at risk due to the growing diversity of Android malware. This situation has sparked significant interest in improving malware detection accuracy and scalability for smart devices. By examining the Long Short-Term Memory (LSTM) method, we have developed an effective deep learning-based malware detection model for enhanced Android ransomware detection. For feature selection, eight different methods were applied. By comparing the outcomes of all feature selection procedures, we used a simple majority vote process to choose the 19 crucial characteristics. The Android Malware dataset (CI-CAndMal2017) and common performance metrics were used to assess the proposed technique. With a detection accuracy of 97.08%, our model surpasses existing approaches. We advocate our proposed method as effective in malware and forensic analysis based on its remarkable performance
Leveraging CIELAB Segmentation and CNN for Wheat Fungi Disease Classification
Get PDFWheat is the third most harvested and consumed grain globally, but a significant portion of its production is wasted due to diseases. Fungal infections caused by pathogenic fungi are particularly harmful, greatly reducing crop yields. Manual visual inspection of large fields is slow, exhausting, and requires specialized expertise. This research introduces a novel combination of image augmentation, CIELAB segmentation, and a fine-tuned pre-trained CNN, achieving an unprecedented 98.43% accuracy in wheat fungal disease classification, addressing gaps in current detection methods and promoting sustainable agriculture. To conduct this research, datasets from Kaggle were merged and meticulously validated to create a comprehensive set with five classes: healthy wheat and four fungal diseases. Preprocessing steps included resizing, contrast enhancement and noise removal to ensure uniform and high-quality images followed by rigorous image augmentation techniques to expand and diversify the dataset ultimately enhancing the deep learning model\u27s robustness and accuracy. The CNN model, trained over 80 epochs achieved an impressive 98.43% accuracy in classifying wheat fungal diseases. With a precision of 98.47% and an F1 score of 98.43% the model demonstrated strong positive classification accuracy. Additionally, a recall of 98.43% and specificity of 98.47% indicated its effectiveness in identifying true positive cases and accurately detecting disease presence or absence
A Demographic Fuzzy Similarity Computation Method for Recommender Systems
Get PDF In this allegedly never-ending stream of e-commerce, it is crucial to offer high-quality suggestions so that consumers can choose wisely from a wide range of picks. Recommender Systems (RS) has proven to be an essential instrument for improving sales of online vendors and enabling consumers with personalized product recommendations. Collaborative Filtering (CF), an extensively preferred approach for Recommender Systems, provides suggestions based on the ratings of users with similar interests. The primary operating component in CF is to measure similarity among items or users. Recommender Systems use a user-item matrix, which is often highly sparse and suffers from cold start, ultimately leading to imprecise recommendations. Instead of relying merely on ratings that are uncertain and can be fake, we integrated the demographic information of users with CF to attain more precise predictions and recommendations in our work. To cope with the uncertainty factor in the recommendation process and to depict the physical world more realistically, we applied the Fuzzy set theory to users’ ratings and demographic features. ML-100K and ML-latest-small datasets are used to evaluate the accuracy of the proposed similarity measure. Compared to the most advanced methods, our proposed demographic fuzzy similarity computation method exhibits considerable achievements in terms of MAE, RMSE, and coverage metrics on standard recommendation datasets, which we used for experimentation
Predictive Maintenance Using Deep Learning: Enhancing Reliability and Reducing Electrical System Downtime
Get PDFPredictive Maintenance (PM) is crucial for enhancing the reliability of electrical systems and minimizing unscheduled outages. However, the conventional methodology lacks the ability to address the escalating and diverse problems of the modern complex environment. The current study provides an alternative approach to carrying out the predictive maintenance activity based on the use of deep learning models that enhance conventional procedures. For the given analysis, we used an artificial data set consisting of 10,000 samples and 14 variables, such as air temperature, process temperature, flipping rate, and tool wear level. We conducted a self-assessment using the specified models to confirm their effectiveness in predicting various failure modes and forms, such as tool wear and heat dissipation conking. This research demonstrated that deep learning models, specifically LSTMs, outperform the established statistical methods in predicting equipment failures. LSTMs provided high accuracy and predicted the failing system before it happened. Furthermore, integrating deep learning with the statistical method that is normally used for anomaly detection improves the model\u27s stability and reliability. The evaluation\u27s findings emphasize the potential of deep learning algorithms for expanding the range of PM applications to achieve better and faster failure predictions. The beneficial thing about this approach is that it presents a means for addressing the inherent problems of large electrical systems’ predictive maintenance that are beyond the scope of traditional practice
A Review on Cloud Computing Threats, Security and Possible Solutions
Get PDFCloud computing is increasingly popular, with major companies like Microsoft, Google, and Amazon creating expansive cloud environments to support vast user bases. Despite its benefits, security remains a significant concern, complicating full trust in cloud solutions due to potential hazards and the consequences of security breaches. This study introduces a novel approach to address the gaps in existing frameworks for summarizing and analyzing cloud security issues and requirements. We explored various cloud computing security challenges, assessed the impact of different cloud models, and discussed risk mitigation techniques and policies for both cloud providers and users. Our analysis offers a comprehensive examination of security risks affecting cloud computing, alongside the latest security solutions. Rather than focusing solely on specific issues, we presented a broader perspective on advanced, high-level security frameworks. We outlined multiple strategies for developing a secure, reliable, and cost-effective cloud infrastructure
AI Based Predictive Tool-Life Computation in Manufacturing Industry
Get PDFFor maximum productivity and optimal utilization of tools, predictive maintenance serves as a standard operation procedure in the manufacturing industry. However, unnecessary or delayed maintenance both causes increased downtime and loss of revenue which should be optimized. Accordingly, this paper presents a method for predicting the maintenance requirement to ensure the optimal utilization of the tools. The experimental data for this research has been collected from a CNC lathe machine in a manufacturing plant for multiple days. The CNC machine equipped with three sensors leads to a detailed log for parameters related to tool wear including current, voltage, acceleration in 3D, motor rpm, and tool temperature respectively. Detailed experimentation has been performed to investigate the importance of different parameters. A direct relationship between current and tool temperature was observed leading to an immediate halt of machine operations. In the subsequent step, maintenance prediction was performed using Logistic regression and Random Forest technique respectively to validate the machine behavior. The retrospective data validated the performance with precise accuracy equal to 98% and 95% for both of methods respectively. The promising results predicting the maintenance schedule of the Lathe machine signify the effectiveness of Machine Learning towards advance scheduling for maintenance. The proactive maintenance strategy helps in potential benefits such as avoiding further costs, avoidance of disruptions, and increased efficiency productivity, thereby enhancing tool life cycles
Deep Learning-Based Automated Classroom Slide Extraction
Get PDFAutomated extraction of valuable content from real-time classroom lectures holds significant potential for enhancing educational accessibility and efficiency. However, capturing the spontaneous insights of live lectures often proves challenging due to rapid visual transitions, instructor movement, and diverse learning styles. This paper presents a novel approach that combines the strengths of YOLO and Scale-Invariant Feature Transform (SIFT) techniques to automatically extract slides from live classroom lectures. YOLO, a real-time object detection algorithm, is employed to identify board area, teacher, and other objects within the video stream. While SIFT, a robust feature-based method, was used to accurately merge key points from multiple pictures of the same region. The proposed method involves a multi-stage process: first, YOLO detects the potential place of the teacher, which occluded the board within the video frames. Subsequently, the teacher was removed from the image. The board was divided into multiple segments, to remove and merge redundant content Scale-invariant feature Transform (SIFT) was employed. Experimental results on a diverse dataset of classroom lecture videos demonstrated the effectiveness of the proposed method in extracting slides across different environments, lecture styles, and recording conditions. The potential benefits include improved note-taking, reduced manual effort in content curation, and enhanced accessibility to lecture materials. The presented approach contributes to the broader goal of leveraging computer vision and machine learning techniques to transform traditional classroom settings into modern, interactive, and adaptive learning environments
Drinking Water Monitoring: Computer Vision Kit for Early E.coli Detection
Get PDFThis work presents an easy-to-use and accurate method to find up to 1 coliform unit (CFU) of a pathogenic bacterium i.e., Escherichia coli (E. coli) in 100ml of drinking water in 6-8 Hours of the incubation period. A larger number of CFUs is easy to detect and incubation time is reduced to 5-7 Hours for the testing samples containing more than 20 CFUs. Normally in laboratories up to 1 ml of a water sample is spread on an endo agar medium and incubated for about 24 Hours, and the E. coli coliform in metallic green color becomes visible through the naked eye. Which has a limitation of finding 1 CFU in just 1 ml of water and a limitation of a large amount of time. In the proposed work Membrane filtration method is used for experiments and a microscopic camera with deep learning algorithms i.e., yolov5 and yolov8 is used for the early detection and counting of E. coli colonies. This system is generalized on the field data of 8k images taken from different cities\u27 water samples in Pakistan. Yolov5s model achieved a mean average precession ([email protected]) of .949, while the latest release version yolov8 achieved [email protected] of 0.950. An automatic imagery system is developed that takes the images just by placing a petri dish in it processes those images through Raspberry Pi, and shows the detected colonies on the screen, while remote users can use a low-cost microscopic camera manually with a developed mobile application
Particle Filter Based Multi-sensor Fusion for Remaining Service Life Estimation of Energized LV-Aerial Bundled Cables
Get PDFAerial Bundled Cables (ABC) consist of several wires that contain numerous layers of thermal insulation, which reduces the risk of theft. Nonetheless, there have been regular reports of rapid degeneration of such cables in coastal areas, resulting in multiple unplanned breakdowns. This study employs the data, collected from field-based nondestructive assessment techniques such as ultrasonic listening and thermal imaging. There is a pressing need for advanced tools to estimate the remaining lifespan of ABCs deployed along coastlines. This paper presents a novel approach using a particle filter-based fusion of multiple sensors framework for estimating the Remaining Useful Life (RUL) of in-service ABCs in a severe coastal atmosphere. The use of multi-sensor measurement data improves the accuracy and reliability of the RUL estimation. This will allow electric power distribution companies to plan maintenance and replacement activities well in time. In the reported research work, the f-step prediction scheme under the framework of the Particle filter algorithm is implemented to predict the posterior density function of degradation growth in the cable insulation. The Particle Filter (PF) method performs effectively with nonlinear state transitions and measurement functions, even when addressing non-Gaussian or multidimensional noise variations. The technique also contains a step error calculation approach for determining forecast accuracy when measurement data is missing. The encouraging outcomes of this strategy illustrate its efficacy
Capturing CO2 and Recovering NH3 by Producing Ammonium Bicarbonate Through Stripping Batch Process at Lab Scale
Get PDFA sustainable Environment is a crucial need for today’s world. To save energy, reduce pollution, and save the economy, researchers are coming up with various sustainable waste management practices to reduce air pollution and water pollution. The production of cement, chemical processing, and power plants are among the industries that release the most CO2. These emissions can be greatly decreased via ammonia-based absorption, which helps to make industrial processes cleaner. One of the methods among all the technologies and solutions is the stripping process where CO2 can be captured by removing ammonia from the water. Not only this but also the chemical is produced NH4HCO3 which can be used in industries or as fertilizer. In this study, A lab-scale stripping process is studied for the recovery of ammonia and capture of CO2 at different experimental conditions which were not studied by other researchers in previous studies. Apart from that, the precipitated product is studied by various characterization techniques including SEM-EDS and XRD. Results show that varying absorption times and flow rates of CO2 and concentrations of solutions affect product quantity. The research concludes the optimum conditions to achieve maximum product i.e., NH4HCO3 was 110 min, 0.5 CO2 gas flow rate, and 15 % NH4OH solution