Authors: G. Veera Shekar, Associate Professor N.S.C. Mohana Rao
Abstract: Due to the growing complexity and frequency of cyber attacks, it is important to shift the para-digm towards predictive security solutions. This paper presents a new anomaly-based framework for smart cyber security using Artificial Intelligence techniques. It utilizes a hybrid Deep Learning (DL) model with the ability to combine CNNs(CNNs) for spatial anomaly detection and Long Short-Term Memory (LSTM) networks for temporal anomaly detection. The proposed frame-work is tested using the CIC-IDS2017 and CSE-CIC-IDS2018 data sets. It shows improved accuracy and low false positive rates compared to machine learning-based security solutions. It is capable of achieving 99.82% accuracy and 0.15% false positive rates, making
