ABSTRACT
The fortification of networks and databases has become increasingly dynamic in recent years, and intrusion detection systems are essential to this process[1][2]. An IDS can provide real-time statistics on network traffic in the event that an attack is identified, inform users, and/or halt doubtful activity. The indefinite features continue to limit the effectiveness and accuracy of ML-based IDS, despite the fact that they have been shown to be efficient at monitoring real-time traffic.
The challenge with machine learning is how the system becomes accustomed to experiencing accumulation to automatically develop performance and is compatible with the idea of detecting the attacks by making the machine learning against intrusion to increase the rate of detection and to decrease the false positive rate[2].
A standard feature set may not be suitable for identifying different forms of intrusion attacks since some attributes may be unnecessary or unrelated and slow down the functionality of machine learning. In order to increase a detection system’s accuracy, it is important to research the ideal qualities. If an attack is discovered, an IDS can offer real-time statistics on network traffic and immediately alert users or stop questionable activity. The accuracy and effectiveness of ML-based IDS are still hampered by their unspecified features, despite the fact that they have confirmed strength in real-time traffic monitoring. The principal task is to focus on topology and hyperparameter configurations that are near to the ideal configuration, which are used to identify the smallest topologies that require the least computational resources. Keywords Intrusion attack, Security, Machine learning, Neural network
