Fake news is misleading information which can often damage individual’s reputation and can change person’s opinion. Recently, the rise of social media platforms gave a boost to circulation of fake news across the globe. Our paper presents a machine learning model using voting ensemble with equal and different weights of each base model. We used 4 different classification algorithms as base learners- 1- Logistic Regression, 2-Multinomial Naive Bayes, 3- Passive Aggressive Classification Algorithm, 4- Random forest algorithm. All algorithms chosen works best with the long textual in stream of data. We compared accuracy of each base model. After assigning different weights to each base model, voting ensemble gave an accuracy of 93.5
In recent years, there has been a surge in interest in a few features of high-tech commercial drones. Small drones have a higher potential for being used for illegal operations due to their ability to evade ground security while transporting payloads. Security breaches like this may only be avoided with drone tracking and monitoring. Recognizing drones in surveillance footage can be challenging due to the similarity between small drones and birds, especially against complex backdrops. Keeping an eye out for drones and other flying objects manually is a time-consuming and difficult task. Therefore, it is necessary to employ a mechanical means of telling drones apart from birds. In this research, we create a system for drone identification using focus measure operators (FMOs). On every video frame, the five FMO parameters are calculated. Drone identification begins with a feature ranking to determine which features are most important and then continues with a classification using a random forest (RF) classifier. The Workshop on Small-Drone Surveillance, Detection, and Counteraction Techniques (WOSDETC), with funding from the Safe Shore Consortium, provides the data used to assess the suggested method’s efficacy in the Drone-vs-Bird Detection Challenge at IEEE AVSS2021. The suggested method presents to identify drones with drone present (DP) vs neither drone nor bird present (NDNBP) (two class), DP vs both bird and drone present (BBDP) vs NDNBP (three class), DP vs BP vs BBDP vs NDNBP (four class) with average acc 94.15
As the popularity of Transformers in computer vision rises, it is likely that Transformer-based models will become the standard for many types of vision applications. The Transformer-based model’s unmatched scalability and vast quantity of trained data make it tough for literatures with fewer data then less computational skills to apply it. Because of its patch-based representation, Transformer-based models have recently been shown to perform very well, and a study presented ConvMixer to demonstrate this. ConvMixer excels in picture classification, but its wasteful isotropic design makes it unfit for other vision tasks. In this research, we offer HEConvMixer, a patch-based representation network that is both hierarchical and data-efficient. In contrast to the original Transformer-based models, our network features two down sample layers and some simple convolutional blocks in place of the Transformer blocks. Our network was trained from scratch on modest datasets using a single GPU. Our HEConvMixer achieves 98.12
The rampant increase in the spread of misinformation around the globe on the social media in the wake of the pandemic international conflicts has urged the innovation in machine learning paradigms of the fake news detection to tackle the threat. Given that the task belongs to the text classification under the subfield of Natural Language Processing, the benchmark defying performance of the transformers-based models such as BERT inspired to use the pre-trained model for text classification using two approaches, one using the BERT for feature extraction and then classifying the text using standard classifiers, a methodology which has not been widely researched, while other involving fine-tuning the BERT for text classification. The various classifier models achieved remarkable accuracy up to 98
In cloud cryptosystems, a diversity of advancement can be experienced to make strong both the authentication and data storage (compression) in open cloud environment. This paper expands the standing effort of authentication with compression schemes. The earlier drawbacks include normal schemes for client’s metrics, direct authentication verification with risks of information outflow to attackers, uploading of replica content and slow speeding of the network. To overcome these existing flaws, this effort delivers a framework deploying a trusted server. When the clients access to open cloud, they share the identities and then accelerated to open cloud with authentication keyword. For compression, the modified deflate algorithm is incorporated for better performance. This strategy helps in generating better results for authentication and data storage milestones. The main objective of the proposed work is to secure the data in the cloud from the unauthorized users and improve the storage capacity in cloud. The proposed framework is sub divided into three stages: Requisition phase, authentication phase and storage phase. The primary stage involves the requisition phase which uses the basis login and password requisition model. The secondary stage executes authentication and authorization. In the final stage the data is compressed and stored in the network using modified deflate data compression algorithm.
