Advance Security: Anomaly Detection in Mobile Crowd Sensing Using Machine Learning Techniques

Abstract

New technologies increasing with new requirements made the public available with many mobile comfortable and social network savvy to collect scaling data for various purposes as crowdsourcing with heterogeneous and cross space data. In order to secure sensing, mobile crowdsensing applications have to address a security threat, such as jamming, fake sensing attacks, and other threats during transmissions. Previous study on a particular part of the issue. But we need an innovative protection solution to achieve the advantages of the new sensing paradigm. This thesis will explore advanced mobile crowd-sensing protection through machine-learning techniques such as SVM (Support Vector Machine) help and the ANN (Artificial Neural Network). We demonstrated using full-blown implementation techniques and experimental evaluation, using precision and false alarm rate as metrics. In this work, the Artificial Neural Network's accuracy and false alarm rate as compared with SVM. Specifically, the proposed ANN achieved an average of 96.4% and more than 7% of the usual incorrect positive rate utilizing 10-fold cross-validation. We strongly recommend the deployment of trustworthy mobile crowdsensing applications using Artificial Neural Network.