DNN-BASED INTELLIGENT INTRUSION DETECTION SYSTEM

Abstract

In order to automatically and quickly identify and categorize cyber-attacks at the
host and network levels, machine learning techniques are being utilized extensively in the
development of intrusion detection systems (IDS). But no prior research has demonstrated a
comprehensive evaluation of the efficacy of different ML algorithms on a variety of open-source
datasets. This research delves into the exploration of deep neural networks (DNNs), a subset of
deep learning models, with the goal of creating adaptable and efficient intrusion detection
systems (IDS) capable of detecting and categorizing previously unseen cyber-attacks. Due to the
ever-evolving nature of both network activity and assaults, it is essential to assess different
datasets produced throughout time using both static and dynamic methods. Research of this kind
helps to find the most effective algorithm for spotting future cyberattacks. On several publicly
accessible benchmark malware datasets, a thorough evaluation of trials including DNNs and
other traditional machine learning classifiers is demonstrated. Incorporating the IDS data into
our DNN model's numerous hidden layers allows it to learn the features' abstract and highdimensional
representation. It has been proven via extensive experimental testing that DNNs
outperform standard machine learning classifiers. Last but not least, we provide Scale-Hybrid-
IDS-AlertNet (SHIA), a framework for hybrid DNNs that can be utilized in real-time to
successfully monitor host-level events and network traffic in order to proactively notify potential
cyber-attacks.