Journal of Cybersecurity and Information Management

In recent years, most of the current intrusion detection methods run for critical information infrastructure are tested for IDS datasets, but does not provide desired protection against emerging cyber- threats. Most machine and deep learning-based intrusion detection methods are inefficient on networks due to their high imbalanced or noisy IDS datasets. Therefore, in this paper, our proposed work implements a comprehensive framework, using multiple models of machine learning and deep learning by taking advantage of advanced feature engineering approaches. Our research explores the impacts of a variety of feature engineering approaches on dimensionality reduction methods used to train and test model performance with execution time taken on the CICIDS2017 dataset to reduce the time complexity and enhance performance to detect intrusion by experiment and leveraging feature engineering techniques like PCA (Principal Component Analysis), LDA (Linear Discriminant Analysis), t_SNE (t-Distributed Stochastic Neighbor Embedding), and Autoencoders. This framework also resolves the class imbalance issues by using SMOTE (Synthetic Minority Oversampling Technique), generates synthetic samples of those classes, which have a very low number of samples to balance the class for a better model performance. Our comparative analysis is performed on metrics like accuracy, training time and memory usage for machine learning models like Gradient Boosting, Logistic Regression, XGBoost and deep learning models. DL with LDA feature engineering approach achieved the highest test accuracy of 95.99% and Gradient Boosting shows strong performance by attaining a high-test accuracy of 90.8%. Illustrated DL model had higher memory usage, but LR and XG- Boost models performed computationally efficient. Further, it is observed that LDA performed better with ML and DL models in comparison to other feature engineering techniques to enhance the intrusion detection efficiency.

The Rivest–Shamir–Adleman (RSA) cryptosystem is one of the most prevalently utilized public-key cryptographic systems in current practice. Prior investigations into vulnerabilities of this cryptosystem have concentrated on diminishing the complexity associated with the integer factorization challenge, which is integral to the RSA modulus, expressed as 𝑁=𝑝𝑞. Possessing partial knowledge about the least significant digits (LSDs) of both p and q is a common assumption attacker’s advantage to enable the polynomial-time factorization of N, ultimately undermining the security of RSA. This paper presents a novel heuristic algorithm predicated on the Constraint Satisfaction Problem (CSP) principles, which estimates k-LSD pairs of the RSA prime factors,  and . The proposed Generate and Test (GT) and Backtracking with Heuristic Variable Ordering (BHVO) solver guarantees polynomial-time factorization of known bits by iteratively refining candidate pairs and eliminating invalid combinations through effective constraint propagation. The proposed approach obviates the requirement for specialized hardware for side-channel attacks to reveal a portion of  and . In our results, we have successfully estimated up to 5-LSDs of  and  with a reduced number of iterations and factored 2048 bits, N based on the known 4-LSDs of the prime in polynomial time. Our research lays the groundwork for factorization algorithms that require partial knowledge of the prime factors. We have highlighted the possible vulnerabilities linked to existing RSA key generation techniques. These may make RSA moduli susceptible to the attacks discussed in this study and proposed countermeasures to ensure secure prime generation.

Numerous criteria are in place for social network applications. They require identification of network's core nodes. Traditional centrality measurements focus on specific node's direct connections or reachability. Often this disregards inherent ambiguity and complexity in real-world social networks. To address these constraints, we have introduced new method called Node Pack Fuzzy Information Centrality based on Pythagorean Neutrosophic Fuzzy Theory. Three essential values truth, falsity and indeterminacy have been added to this approach. This new approach provides a thorough depiction of social networks and it also offers a more sophisticated comprehension of connections between nodes. Complex and ambiguous interactions between entities can be effectively expressed using Pythagorean Neutrosophic values. Unlike traditional values, Pythagorean Neutrosophic values consider several uncertainty dimensions; this is a major improvement over traditional fuzzy value. Our approach handles relational complexity well and it includes self-weight for every node too. It represents each node's unique value, significance, or impact on the network. The network assessment is now more precise and contextual so we can assess centrality with greater precision. We applied this approach to a small academic network called university faculty/researchers. The application of Node Pack Fuzzy Information Centrality yielded promising results. It can enhance various activities associated with social network analysis. It can also offer valuable insights into the network architecture.

A firewall is one of the devices that supports network security, especially at the organizational level. A Firewall's effectiveness in supporting network security is highly dependent on the capabilities and abilities of the Network Administrator. Unfortunately, the high complexity of creating rules and the process of configuring Firewall rules carried out statically by the Network Administrator weakens the effectiveness of the Firewall, and it cannot adapt to increasingly dynamic network pattern changes. Machine Learning is one of the potentials that can be used so that the Firewall can work adaptively. Adaptive Firewall configuration in recognizing various attacks in the network will undoubtedly increase the effectiveness of the Firewall in ensuring network security. The success of the machine learning model performance cannot be separated from the dataset used during the learning process. The dataset used in learning often has a large dimension, but various noises and attributes are irrelevant in representing one class of data. Therefore, it is necessary to support the feature selection technique, which will show the presence of relevant characteristics in the dataset and maximize the machine learning model's performance. This study will be conducted on adding feature selection techniques to develop machine learning models on the Benchmark dataset related to network security. Various popular feature selection techniques will be evaluated, and their performance will be compared based on scenarios between feature selection techniques or scenarios that only use a single classification.

In this research, we investigate sophisticated methods for Intrusion Detection Systems (IDS) and Intrusion Prevention Systems (IPS), leveraging AI-based feature optimization and diverse machine learning strategies to bolster network intrusion detection and prevention. The study primarily utilizes the NSL-KDD dataset, an enhanced version of the KDD Cup 1999 dataset, chosen for its realistic portrayal of various attack types and for addressing the shortcomings of the original dataset. The methodology includes AI-based feature optimization using Particle Swarm Optimization and Genetic Algorithm, focusing on maximizing information gain and entropy. This is integrated with the use of Random Forest (RF) to reduce class overlapping, further enhanced by boosting techniques. Grey Wolves Optimization (GWO) alongside Random Forest. This innovative approach, inspired by grey wolf hunting strategies, is employed for classification tasks on the NSL-KDD dataset. The performance metrics for each intrusion class are meticulously evaluated, revealing that the GWO-RF combination achieves an accuracy of 0.94, precision of 0.95, recall of 0.93, and an F1 score of 0.94.