An OFLN by addition of PSO to perform automated hyperparameter optimization and MAML to enable high-speed adaptation to new and previously seen attacks is proposed, which is more effective than baseline ensemble models in detection accuracy, resistance to class imbalance and training efficiency.
Abstract
Traditional Intrusion Detection Systems (IDSs) tend to have a high false positive detection rate, are time-consuming to train, and lack the flexibility to react quickly to changing cyber threats, such as Distributed Denial of Service (DDoS) attacks and phishing. Such restrictions decrease the responsiveness and reliability of the IDS services implemented in the university network, both wired and wireless, at the Uganda Christian University (UCU), where network and traffic analysis have become a necessity. This study analyzes the current threat environment of networking at UCU in terms of the type, origin, and target of various attacks to identify gaps in existing detection tools. To address data imbalance and high dimensionality, a range of machine learning algorithms was implemented and paired: Random Forest (RF), XGBoost, AdaBoost, Decision Trees, Convolutional Neural Networks (CNNs), and SMOTE. The highest performance was obtained by the combination of RF and SMOTE, with an accuracy of 81.88%, a precision of 82.17%, a recall of 81.88% and an F1-score of 80.19%. These ensemble learning methods can be effectively combined with oversampling to better detect minority attacks and minimize false alarms. This paper proposes an OFLN by addition of PSO to perform automated hyperparameter optimization and MAML to enable high-speed adaptation to new and previously seen attacks. The OFLN incorporates swarm-based optimization with meta-learning, attaining faster convergence, better generalization and efficiency. Empirical results show that OFLN is more effective than baseline ensemble models in detection accuracy, resistance to class imbalance and training efficiency, making it an innovative, scalable and practical IDS.
The rapid growth of digital communication technologies, cloud computing, and Internet of Things (IoT) devices has
increased both the frequency and sophistication of cyber-attacks, making effective intrusion detection an essential component of
modern cybersecurity systems. Traditional signature-based intrusion detection systems (IDS) are effective against known attacks
but fail to detect previously unseen or evolving threats. This study investigates the application of deep learning models for binary
network intrusion detection using the NSL-KDD benchmark dataset. Three standalone architectures, Convolutional Neural
Networks (CNN), Long Short-Term Memory (LSTM) networks, and Deep Neural Networks (DNN), are implemented and
evaluated, alongside a CNN-LSTM Hybrid model that integrates spatial and sequential learning, and a DNN-LSTM Ensemble
model that combines independently trained DNN and LSTM predictions through weighted averaging. Following data cleaning,
categorical encoding, normalization, and Random Forest-based feature selection (41 features reduced to 20), all models were
trained and evaluated under identical conditions using Accuracy, Precision, Recall, F1-Score, ROC-AUC, training time, and
inference time. The standalone DNN model achieved the best overall performance, with 80.98% accuracy, 97.08% precision,
68.66% recall, 80.43% F1-score, and 96.11% ROC-AUC, while also requiring the shortest training time (39.69 s). The CNNLSTM Hybrid model attained the highest precision (97.23%) but did not outperform the standalone architectures overall, and
the DNN-LSTM Ensemble produced balanced but not superior results. These findings indicate that carefully designed
standalone architectures can match or exceed the performance of more complex hybrid and ensemble models for binary
intrusion detection, while incurring substantially lower computational cost. The study contributes a controlled, commonframework comparison of five deep learning architectures and provides practical guidance for selecting computationally
efficient models for anomaly-based intrusion detection.
Ketki Naik, Sanjeev Ghosh· International Journal for Re...· 0 citations
The findings indicate that hybrid deep learning techniques can improve network security by enhancing intrusion detection capability while reducing false alarms.
A. O. Jimoh-Mahmud, Abubakar Dayyabu, Abubakar Sadiq Idris et al.· FUDMA Journal of Sciences· 0 citations
As network security issues become increasingly severe, the complexity and concealment of network attacks continue to increase, and the need for a reliable network intrusion detection system (NIDS) is imminent. This study is dedicated to solving this problem and proposes a unique NIDS based on neural networks. The Transformer encoder is integrated into the traditional neural network architecture, and a convolutional neural network (CNN) is applied to extract features, building a model that can process complex network traffic data more efficiently. After a series of rigorous experimental verifications, the model shines in performance. In terms of key indicators such as accuracy, recall, and precision, it significantly surpasses traditional models and other common neural network models. Regardless of the network bandwidth, attack frequency, or data set size, it shows excellent adaptability and stability. Especially when dealing with class imbalanced data sets, the model's detection ability for minority attacks has been effectively improved, providing a more solid guarantee for network security protection and injecting new vitality into the development of network intrusion detection technology.
An Enhanced Multi-Model Ensemble Network Intrusion Detection System (EME-NIDS), a deep meta-learning system that combines five different heterogeneous learning paradigms, including Convolutional Neural Networks, Dense Neural Networks, Transformers, XGBoost, and Random Forests is introduced.
This study examines a one-dimensional Convolutional Neural Network and a hybrid model, investigating how both architectures can detect network attacks in binary and multiclass classification settings, and provides actionable insights for practitioners choosing between deep learning and classical approaches under real-world NIDS deployment constraints.
Applying Artificial Intelligence (AI) to the context of networking systems has greatly enhanced the ability to detect, protect, and manage the kind of flows in networks in real time. This research explores the relationship between AI and networking, and more particularly, the extent to which AI technology can be used to enhance, protect, and scale the NIDS using the NSL-KDD dataset. This research focuses on the following objectives: (a) enhancing the performance of network intrusion detection models, and (b) strengthening the defense mechanisms against dangerous new forms of cyber threats, and (c) making flexible technical solutions for effectively managing large and dynamically changing networks. To address these objectives, different classification algorithms such as SVM of different kernels, KNN, Random Forest, and Decision trees are applied on the NSL-KDD dataset. Testing accuracy, precision, recall, F1 score, and cross-validation score are used to evaluate them. In addition, feature extraction algorithms, including both fuzzy and correlation-based methods, are applied to reduce the model’s computational complexity. Targets of the investigation include the evaluation of each model and feature extraction, and its impact on the model’s ability to predict. The results affirm that, compared with other methods, SVM-based models with Gaussian and Sigmoid kernels are particularly more accurate (up to 99%) and less sensitive to fluctuations.
Anish Antony, S. Thaseen, Ashvini Alashetty et al.· Discover Artificial Intellig...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
