Terrorism And Fake News Detection

Fake news dissemination is a critical issue in today’s fast -changing network environment. The issues of online fake news have attained an increasing eminence in the diffusion of shaping news stories online. This paper deals with the categorical cyber terrorism threats on social media and preventive approach to minimize their issues. Misleading or unreliable information in form of videos, posts, articles, URLs are extensively disseminated through popular social media platforms such as Facebook, Twitter, etc. As a result, editors and journalists are in need of new tools that can help them to pace up the verification process for the content that has been originated from social media. existing classification models for fake news detection have not completely stopped the spread because of their inability to accurately classify news, thus leading to a high false alarm rate. This study proposed a model that can accurately identify and classify deceptive news articles content infused on social media by malicious users. The news content, social-context features and the respective classification of reported news was extracted from the PHEME dataset using entropy-based feature selection. The selected features were normalized using Min-Max Normalization techniques. The model was simulated and its performance was evaluated by benchmarking with an existing model using detection accuracy, sensitivity, and precision as metrics. The result of the evaluation showed a higher 17.25% detection accuracy, 15.78% sensitivity, but lesser 0.2% precision than the existing model, Thus, the proposed model detects more fake news instances accurately based on news content and social content perspectives. This indicates that the proposed classification model has a better detection rate, reduces the false alarm rate of news instances and thus detects fake news more accurately.


I. INTRODUCTION
The research focuses on terrorism and fake news detection on social media. A growing interest related to fake news detection has attracted many researchers as fake information is circulated through online social media platforms such as Facebook, Twitter, etc. The fake content is spreading at a faster pace to gain popularity over social media, to distract people from the current critical issues. Most of the people believe that the information they receive from various social media sites is reliable and true, i.e., people are inherently truth-biased. Also, people easily trust and want to believe in what they actually interpret in them minds, i.e., confirmation-biased. In general, it has been analysed that human are unable to recognize deception effectively. Due to which a serious and negative impact of fake articles can be seen on society and individuals leading to an imbalance of the news ecosystem. It was observed that during US president election, most of the widely spread articles on social media were fake. Recently a fake video related to Therefore, an improved fake news detection classification model that would incorporate both reinforcement and supervised approach is highly needed. Thus, in this paper, an attempt was made to develop an improved model with the capability of accurately detecting fake news on the social media platform by classifying news content as either real news or fake news using a hybrid approach of supervised and reinforcement learning approaches. fake news on social media.

II. LITERATURE REVIEW
The automated systems can play an important tool for identifying fake stories, articles, blogs, clicks which manipulate public opinion on social networking sites. several hybrid algorithms have been developed for fake news detection using data mining methods. In This study showed that the only supervised learning approach is not adequate for accurate fake news detection. In, a hybrid deep model was proposed to address the timely problem of fake news detection by incorporating the text, the response an article receives, and the users who source it. The model was able to address the timely problem of fake news detection. meanwhile, it was only based on user behaviour and does not incorporate reinforcement learning and humans in the learning process and thus affected the accuracy of the detection. In, the identification of fake news on Twitter was formulated as a binary machine learning problem. the result showed that instead of creating a small, but accurate hand-labelled dataset, using a large-scale dataset with inaccurate labels yields very good results. In, a tensor modelling of how to accurately distinguish different categories of fake news based mainly on the content, where latent relations between articles and terms were captured as well as spatial or contextual relations between terms, towards unlocking the full potential of the content was proposed.
The proposed algorithm, a supervised learning approach was able to identify different news categories with corpus with a percentage of 80%, but with high false positives rate. In, the tensor decomposition ensembles model was proposed by clustering the news content of fake news articles into different categories. They were able to do this by capturing spatial relations between terms for each article. This was able to identify all different categories of fake news within the dataset. Therefore, it was noted in and that they are mainly supervised learning approaches. Though they all yield a good detection rate, but they have not completely stopped the spread because of their inability to accurately classify news. In, a stacked ensemble of 5 independent classifiers in the context of the natural language processing was developed to detect fake news. The stack ensemble was a two-layer classifier architecture that leverages predictions from weaker slave classifiers as features to a stronger master classifier.
The classification experienced a difficulty in which the test set was far greater than that of the development data split which impacted results. Also in some cases, multiple layers of neural networks are required for effective training of models. This revealed that a stack ensemble method of classifiers is appropriate to detect fake news accurately. In, a framework that detects and classifies fake news   Fig. 7. The main reason for using SVM is because the problem of fake news detection is not linearly separable. SVM was used to find the best separating hyperplane also called decision boundary. Also, to design a hyperplane that classifies all dataset into two classes i.e., training and testing dataset. SVM helps to label the datasets appropriately since mislabeling can decrease model performance. Random Forest which was used as the reinforcement learning method of the ensemble model operates by constructing a multitude decision at training time and also used to reduce the variances used for classification of news. The algorithm is presented in Fig. 8. By incorporating the reinforcement learning into the modelling process will enhance the detection rate of the model. The   Table 1 showed that the proposed model was able to correctly classify 12,544 (87% of real news instances) as real news (i.e. True Positives (TP)), while 4329 (70% of fake news instances) was correctly classified as fake news instances (i.e. True Negatives (TN)). The result also showed that 1,885 (13% of real news instances) were misclassified as fake news instances (i.e., False Positives (FP)), while 1,885 (30% fake news instances) were misclassified as real news instances (i.e., False Negatives (FN)). This shows that the proposed model detects fake news with a reduced false alarm rate.
The testing dataset was also fed into the existing model, the result from the confusion matrix as shown in  The evaluation results indicate that the proposed model is more sensitive and accurate to classify news as either fake or real than the existing model.    and sensitivity rates, thus has a better performance in detecting fake news instances accurately. Therefore, the proposed model can be adapted by the social media administrators to combat the spread of fake news. The future works is directed to having a model that will the address the timely prediction of the fake news in social media. The proposed model was simulated and it was obvious in the results that the proposed model has a high reduction in theinstances of false positives states and high detection accuracy and sensitivity rates, thus has a better performance in detecting fake news instances accurately many it companies has been working on stop hate spread but this is one step further which focuses to stops terrorist activities and fake news spreading through social media.