A novel stable feature selection algorithm for machine learning based intrusion detection system Sowmya T, Mary Anita E A Procedia Computer Science, 2025 The advent of new technologies like artificial intelligence, and big data has influenced many cyber attackers to launch their attacks on the network. Hence researchers have already proposed Intrusion Detection Systems by incorporating machine learning as well. Building an effective IDS is still a challenging task because of low accuracy. Managing high dimensional data is another major problem that occurs in IDS. Hence in this paper, an efficient Machine Learning based Intrusion Detection System is developed by means of a novel stable feature selection strategy called IV-RFE. The proposed methodology aims to select only the relevant features that contribute to the attack, which is purely based on relative variance, and weight factor in combination with RFE. This methodology increases the performance in terms of accuracy and maintains a stable set of features. Previous studies only focussed on the feature selection strategy and their performance. The feature stability also has to be considered which is an equally important metric, especially in the field of Intrusion Detection Systems. Hence in the current study, an efficient ML based IDS is proposed which selects only the relevant and stable features. Experimental results also revealed that the proposed IV-RFE outperformed well for three attacks with respect to accuracy and stability metrics also. The results show that stability is also an important indicator in selecting the features in the field of Intrusion Detection Systems.
A Novel SHiP Vector Machine for Network Intrusion Detection Sowmya T., E. A. Mary Anita IEEE Access, 2025 In this paper, network intrusion detection is proposed using an improved version of the support vector machine model to detect DoS attacks. Here, the SVM model considers the weight parameter along with the kernel to find the best decision boundary that separates the data into DoS and normal. The proposed model provides a novel kernel trick that reduces the overlapping of data. The intrusion detection system aims to construct an ideal system that can detect attacks with very high performance using a ShiP vector machine(Sophisticated High Performance Vector Machine). The framework comprises three major steps: data collection and preprocessing, Recursive Feature Elimination (RFE) based feature selection, and the ShiP Vector Machine classification strategy. The system is evaluated using the DoS dataset from UNSWNB15 and real time PSD-23 sniffer dataset. DoS data is generated by extracting the normal and DoS attacks from the UNSWNB15 dataset. Experimental results show that the proposed ShiP vector machine shows outstanding performance by achieving 96.44 % accuracy on the DoS dataset and 90.12 % accuracy for real time PSD-23 data.
Research, Development, and Comparative Characterization of Methods for Invisible Embedding Digital Watermarks in Electronic Text Documents Maxim Martemyanov, Maria Lapina, Mary Anita E. A. Navigating Technological Advancement in the Vuca and Bani World, 2025 This article discusses the concept of digital watermark (DWС), its classification, the main file formats for storing textual information electronically, types of attacks on DWC, and methods of embedding digital watermarks in electronic text documents based on formatting changes, using metadata, structural and linguistic changes. The practical development of each method is carried out, their advantages, disadvantages and resistance to various attacks are identified, and their comparative characterization is presented. On the basis of the analysis, we experimentally developed our own methods of introducing DWC into text documents, including the approach using zero-width spaces (ZWSP) (U+200B), which provides high resistance to various attacks and imperceptibility for the user. The results of the work demonstrate the effectiveness of each implementation and provide recommendations for their application depending on the requirements for copyright protection and authentication of text files.
Automated Detection Model (ADM) for Glaucoma, Exudate and Diabetic Retinopathy (DR) Diagnosis Using Fundus Images M P Karthikeyan, E.A. Mary Anita, D. Mohana Geetha 2nd International Conference on Research Methodologies in Knowledge Management Artificial Intelligence and Telecommunication Engineering Rmkmate 2025, 2025 A total of 15 million people in India suffer from blindness yet statistical analysis shows 75% of these cases can be treated. The research shows DR and Glaucoma lead to blindness in India. Long-term diabetes mainly causes diabetic retinopathy which stands as the primary cause of blindness. Glaucoma damages the optic nerve until blindness develops. The digitized format of fundus images provides useful diagnostic information about infected retinas for proper eye disease detection. Eye defect diagnosis at an early stage enables medical care that greatly decreases patient vision loss risk. An ophthalmologist conducted the disease screening process through examination of fundus image abnormalities. Higher rates of DR and glaucoma prevalence do not affect the number of available ophthalmologists for evaluating fundus images so the prevention of diseases has been delayed. An automated analytical system should be developed presently to help ophthalmologists enhance their diagnostic process efficiency. The paper introduces an artificial learning methodology that utilizes concatenate systems to detect input fundus images in three categories namely ND and GI and EI and DRI. No Diseases (ND), ii. Glaucoma (GI) iii. The classification groups include Exudate infected Images (EI) along with two other categories namely Glaucoma (GI) and DR Images (DRI). The proposed model Automated Detection Model (ADM) starts by analyzing input samples with histogram-based model and employs DenseNet121 and Inception-ResNetV2to facilitate further processing. The Convolution Neural Networks (CNN) function gathers and sorts the feature extraction data obtained from both models. The proposed approach demonstrates improved accuracy and recall plus average precision when used instead of a solitary model. The proposed machine-learning approach using fundus images proves successful for Glaucoma, Exudate and DR diagnosis according to this experiment.
Preface Lecture Notes in Networks and Systems, 2025
Preface Lecture Notes in Networks and Systems, 2025
Blockchain-Based Digital Twin for Predictive Maintenance of Machines Using Machine Learning Blockchain Based Digital Twins Research Trends and Challenges, 2025
Preface Lecture Notes in Networks and Systems, 2025
UAV Security Analysis Framework Elena Basan, Evgeny Abramov, Nikita Gladkov, Maria Lapina, Vitalii Lapin, E. A. Mary Anita, Sandeep Kumar Lecture Notes in Networks and Systems, 2024
An Intelligent Hybrid GA-PI Feature Selection Technique for Network Intrusion Detection Systems International Journal of Intelligent Systems and Applications in Engineering, 2023
Message from IEEE InC4 2023 Program Chair Addapalli V N Krishna, E A Mary Anita, Sandeep Kumar Proceedings of IEEE Inc4 2023 2023 IEEE International Conference on Contemporary Computing and Communications, 2023
FEM-hybrid machine learning approach for the detection of sybil attacks in the wireless sensor networks International Journal of Innovative Technology and Exploring Engineering, 2019
Improved centralized base station mechanism to detect replicas or clone nodes in static wireless sensor network International Journal of Innovative Technology and Exploring Engineering, 2019
Detection and prevention of black hole attacks in vehicular ad hoc networks International Journal of Innovative Technology and Exploring Engineering, 2019
