Analyzing and Predicting Cyber Hacking with Time Series Models

Abstract

Cyber hacking implies utilizing PCs to fake action, for instance misrepresentation, security intrusion, taking corporate, individual information, and so on. Analyzing and predicting cyber hacking knowledge is a very vital technique for deepening our understanding of the evolution of the threat scenario. The different algorithms utilized for forecasting can be classified into linear models (AR, ARIMA, ARIMAX) and non-linear models (ARCH, GARCH, Neural Network). ARIMAX is one of the notable models dependent on time series which was utilized in the existing framework. In this system, SARIMAX (Seasonal Autoregressive Integrated Moving Average with exogenous) and RNN (Recurrent Neural Network) are proposed for forecasting the cyber hacking breaches from 2014 to 2017(4years) of a cyber-hacking dataset. These models are trained with the human health care services hacking breach dataset to forecasting the next eight years of the breach size and the incident arrival years based on the past values and also analyze the trend of the breach. The result obtained was compared to the SARIMAX and RNN models and it's been discovered that the recurrent Neural Network is outperforming the present linear model. This forecasting is very useful for making a decision about the security of an organization. In conclusion in phrases of these models to evaluate the superiority in forecasting performance.