Deep Analysis of Risks and Recent Trends Towards Network Intrusion Detection System

Abstract

In the modern world, information security and communications concerns are growing due to increasing attacks and abnormalities. The presence of attacks and intrusion in the network may affect various fields such as social welfare, economic issues and data storage. Thus intrusion detection (ID) is a broad research area, and various methods have emerged over the years. Hence, detecting and classifying new attacks from several attacks are complicated tasks in the network. This review categorizes the security threats and challenges in the network by accessing present ID techniques. The major objective of this study is to review conventional tools and datasets for implementing network intrusion detection systems (NIDS) with open source malware scanning software. Furthermore, it examines and compares state-of-art NIDS approaches in regard to construction, deployment, detection, attack and validation parameters. This review deals with machine learning (ML) based and deep learning (DL) based NIDS techniques and then deliberates future research on unknown and known attacks.

RFSA: A Ransomware Feature Selection Algorithm for Multivariate Analysis of Malware Behavior in Cryptocurrency

“…Another area of concern emerging in the current literature is the differentiation or classification accuracy between ransomware and other malware types [29], [30], which can affect the reliability and precision of the detection process. Moreover, many methodologies rely heavily on the quality and comprehensiveness of their training data [31], [32]. This leads to potential biases or inadequacies in their predictive capabilities.…”

“…In 2021, Nkongolo et al [7] introduced the UGRansome dataset (Figure 3). UGRansome has demonstrated its inestimable value in identifying and combating ransomware threats, even those deemed zero-day vulnerabilities [32], [51]. What differentiates UGRansome from other datasets in the domain of Intrusion Detection Systems (IDS) is its all-encompassing coverage of previously unexplored ransomware attack types [52].…”

“…This metric assesses the algorithm's precision in positive predictions, where a high precision denotes fewer false positive errors. Recall, synonymous with sensitivity or true positive rate, gauges the ratio of true positive predictions to the overall number of actual positives (comprising true positives and false negatives) [31], [32]. It evaluates the algorithm's capacity to correctly identify all positive instances.…”

“…Furthermore, categorical variables were proficiently converted into numerical equivalents, making them suitable for a wide range of modeling and analytical https:// journal.uob.edu.bh The enriched dataset will enhance the model's understanding of underlying patterns, leading to improved accuracy and performance in predictive tasks. Overall, the numerical enrichment of the dataset empowers ML models to extract meaningful insights and make more accurate https:// journal.uob.edu.bh predictions [31], [32], [48].…”

Lightweight Hybrid CAE-ELM and Enhanced Smote Based Intrusion Detection for Networks

“…The detection of malicious traffic in the network is considered a binary and multiclassification issue. The AI (artificial intelligence) based models detect the intrusion in the network through various criteria like network based model, host based model, anomaly based model and signature based model [9,10,28]. Among these, signature-based network intrusion detection is widely utilized by creating a set of rules for identifying the network pattern.…”