Studies for detecting of fake & real news

Abstract

Fake information has been spreading in extra numbers and has generated increasingly more misinformation, one of the clearest examples being the USA presidential elections of 2016, for which lots of fake facts become circulated earlier than the votes that progressed the photograph of Donald Trump overs Hilary's Clinton. Because faux information is just too much, it will become essential to apply computational gear to locate them; that is why using algorithms of Machine Learning like undefined a Naive Bayes Model and herbal language processing for the identity of fake information in public facts units is proposed. Spreading fake news over popular social media like Facebook, Twitter, Whats App, etc. creates serious social problems. It is observed that the administration had to stop internet service in a large area to control the spreading of such fake news. Bar in the internet service creates another serious problem for the citizens as we are heavily dependent on the internet nowadays. To detect fake news, this work suggests a machine learning method using the Naïve Bayes classifier. The system can be plugged in with any social media, as it predicts the probability of news to be fake the same can be stopped from spreading farther with the software control in the social media. This model was trained and tested in 3 datasets (2 of the English language and 1 of Bengali language). We have successfully achieved a high accuracy which is 81% for the English language and 93% for the Bengali language. This is an acceptable result comparing the same with the recent works. Using Deep Learning or a combination of classifiers the result may be improved further. With the increasing usage of wearable smart devices for caption monitoring, providing caption facilities in remote areas are being developed. For an organization managing all these data from Caption devices and other news information becomes difficult in a local database. The situation is similar for large News with lots of departments as well. The cloud computing technologies can provide a low-cost and scalable solution to this huge volume of data. However, with the advantages of a cloud database, come various security issues. Firstly, secure data transfer through a net- work is needed so that data cannot be stolen or tampered. Also data stored in the cloud database needs to be protected because if the cloud provider is entrusted or is attacked by an intruder, data confidentiality and integrity can be lost. Encrypting all the data stored in the cloud database provides the required level of security. However It produces several challenges in searching an encrypted data base. This thesis describes the challenges and presents an approach to secure a Cassandra database used in an entrusted cloud environment. Specifically, the proposed model provides a secure inter face to the data that is stored in a distributed data base in cloud, by tackling threats of both internal and external attackers. It stores news information in encrypted form in the database and also modification-sensitive information is hashed and stored in block chain. Thus, it provides seamless access to the encrypted data for permitted users while limiting the access for other users in the system, internal and external attackers; along with preventing unauthorized data modification. Storage of hashes allows the system to validate any information at any point of time, while access to the hash values is provided with a secure protocol similar to a block chain.