Artificial Intelligence, Computer Networks and Communications
9
Scopus Publications
Scopus Publications
Enhancing Security and Reliability in IIoT-SCADA Networks Through Integrated PRU and Decision Tree Models M. Sharmila Devi, K. Dharmesh, A. Joy Pranahitha Proceedings of the 5th International Conference on Sentiment Analysis and Deep Learning Icsadl 2026, 2026 Stakeholders in the Industrial Web of Things (IIoT) rely on its reliability and longevity to keep people safe as they carry out vital tasks. Integrity, privacy, security, dependability, resilience, and safety are essential components of a reliable IIoT-enabled network. Protocol variations, restricted update choices, and older modifications of security processes render typical security procedures and mechanisms inadequate for protecting these networks. Consequently, these networks need fresh ways to boost security and privacy while also increasing trust level. So, to make IIoT-enabled networks more trustworthy, we provide a new method in this article. We provide a trustworthy method for detecting cyberattacks in SCADA networks that is both accurate and dependable. The proposed approach integrates HoT networks based on SCADA with deep learning-based pyramid recurring units (PRU) & decision trees (DT). To identify cyberattacks in HoT networks that rely on SCADA, we also use an integrated learning approach. High detection rates are achieved by addressing the sensitivity of irrelevant characteristics using the ensemble DT and PRU's nonlinear learning capacity. Fifteen datasets derived from SCADA-based networks are used to assess the proposed method. The suggested strategy outperforms both conventional & machine learning-based detection methods, according to the experimental findings. Networks enabled by the Industrial Internet of Things (IIoT) are made more secure and trustworthy with the help of the suggested scheme.
A Hybrid Deep Learning and Optimization-based Framework for Real-Time Cyber Threat Detection in Network and IoT Environments M. Sharmila Devi, Deevi Radha Rani, M. Sharmila Devi Proceedings of the 2026 International Conference on AI Driven Smart Systems and Ubiquitous Computing Icauc 2026, 2026 The growing complexity and prevalence of cyberattacks in contemporary connected and Internet of Things (IoT) settings require intrusion detection frameworks that are corrective, receptive, and strong. Conventional security measures tend to fail in identifying non-traditional and novel attacks and have low false alarms. In this paper, the hybrid deep learning-based cybersecurity system called TriNet is introduced to detect cyber threats in diverse and heterogeneous network and IoT systems in real-time. The suggested framework combines Chien physics in formed neural networks, Pt-Net, and Generalized LaGrange and Neural Networks to simulate spatial, temporal, and dynamic behavioral patterns of cyber intrusion. Mean imputation, label encoding, and Min–Max normalization are used to preprocess network traffic and IoT telemetry data in the UNSW-NB15 and TON-IoT benchmark datasets. To improve the detection efficiency and complexity computations, a hybrid approach to feature selection that combines the use of the PUMA optimizer and Hippopotamus Optimization Algorithm is used. The predictions of the individual deep learning models are put together through an ensemble-based majority voting system to give the ultimate classification decision. It is shown that TriNet has a high accuracy, better detection rates, and lower false alarm rates than currently used methods of intrusion detection, making it suitable to be used in the next-generation cybersecurity.
Deepfake Image Forensics Using Transfer Learning A. Sreepradha, Murikoti Madhusudana Reddy, B.Swarajya Lakshmi, M.Sharmila Devi, Boreddy Vidya Sagar Reddy Proceedings of the 5th International Conference on Sentiment Analysis and Deep Learning Icsadl 2026, 2026 Over the past few years, deepfakes have emerged as a serious threat to digital security, privacy, and information integrity due to rapid advances in generative models that produce highly realistic facial images. Manual verification has become increasingly unreliable, necessitating robust automated detection techniques. This paper presents a deep learning-based framework for distinguishing real and AI-generated face images using a transfer learning approach built on DenseNet-121. Rather than introducing a complex architecture, the proposed method emphasizes an effective two-stage, curriculum-based training strategy to enhance accuracy and robustness under realworld distortions. Experiments conducted on a large-scale dataset of approximately 140 K images demonstrate strong performance across standard classification metrics, achieving an accuracy of 96.8% and an ROC-AUC of 98.2%. The trained model is further integrated into a lightweight web-based application that provides real-time image authenticity prediction with confidence scores, highlighting its practical applicability.
Smart Botnet Detection in IoT using Stacked Neural Networks V. Lakshmi Chaitanya, P. SomaSekhar Reddy, M. Sharmila Devi, M. Satesh Kumar Reddy, P. Venkata Vardhan, J. Lohith Rao Proceedings of the 2026 International Conference on AI Driven Smart Systems and Ubiquitous Computing Icauc 2026, 2026 The increasing range of IoT (Internet of Things) devices has rendered them very susceptible to attacks by botnets that can destroy the security of networks as well as interfere with services. The proposed project is geared towards creating an effective hybrid machine learning model to identify botnet attacks within the IoT networks. The proposed methodology is a stacked combination of deep learning models, i.e., LSTM, RNN, and CNN, and one of the classic machine learning classifiers. The training dataset is composed of network traffic data, which is pre-processed with such techniques as feature selection (Select Best) and data balancing (SMOTE Tomek). Sequential data characteristics of deep learning models are combined with the classification capabilities of tree-based models in the hybrid model to enhance accuracy of detection and decrease false positives. These findings indicate that LSTM + Random Forest recorded an accuracy of 88% which is better than the individual model. This is a hybrid solution that is scalable and robust to detect botnets and guarantee the security of IoT networks.
CNN2D-Based Deep Learning Framework for IoT Cyberattack Detection with CICAPT-IIOT M. Sharmila Devi, G.Sai Guna Varsheta, B. Sreedhar, B. Swarajya Lakshmi, D. Sravanthi, S. Rubiya Parveen Proceedings of 2nd International Conference on Visual Analytics and Data Visualization Icvadv 2026, 2026 SCADA (supervisory control and data acquisition) systems are particularly vulnerable to the new cybersecurity threats brought forth by the IIoT's fast growth, which has revolutionized industrial automation. When it comes to advanced cyber threats like denial-of-service (DoS) as well as distributed denial-of-service (DDoS), traditional malware detection systems (IDSs) often struggle to identify and mitigate them correctly. In order to improve cybersecurity in IIoT settings, this research suggests a state-of-the-art intrusion detection system (IDS) architecture that uses a CNN2D. The accuracy, recall, precision, F1-score, matrix of confusion, & AUC-ROC metrics was used to evaluate the performance of the proposed model that was trained on the CICAPT-IIoT dataset and tested on several intrusion categories. This study employs CNN2D exclusively because of its capacity to automatically extract geographical and temporal information from network traffic data, in contrast to traditional IDS methods that depend on a number of algorithms. The CNN2D model outperformed the competition when it came to recognizing malicious and benign traffic patterns, with a detection rate of 99.78% on the data used for testing. Optimal hyperparameter setups for IIoTspecific security tasks are also investigated in the research to guarantee resilience and flexibility. Based on these results, CNN2D is a great neural networks solution for current IIoT network security since it is both effective and scalable.
Developing a Robust Intrusion Detection System Using SMOTE and Hybrid SVNN Model B. Kiranmayee, M. Sharmila Devi, Kalva Susheela, Raghu Dhumpati, Kiran Kumar Reddy Penubaka, U Ganesh Naidu 4th International Conference on Sentiment Analysis and Deep Learning Icsadl 2025 Proceedings, 2025 This paper proposes an integrated approach to build up IDS with proper effectiveness toward the rising need for strong network security. Network traffic anomaly detection and classification are one of the major aims and enhance the security layer against various types of cyber threats. This study is a methodical approach in which a diverse set of data is first extracted from Kaggle. The collected dataset is a comprehensive one that includes various kinds of network traffic data. The first step includes preprocessing the data, i.e., handling missing values, removing erroneous entries, and dealing with outliers using the Interquartile Range (IQR) method. To counter class imbalance, Synthetic Minority Oversampling Technique (SMOTE) is utilized in generating synthetic samples in underrepresented classes for generalization of models. The feature selection algorithm relies on the Chi-Square Lasso Feature Selection method. This method unites Chi-Square tests to select statistically significant features with Lasso regularization for feature dimension reduction with the goal of overfitting minimization. For building models, a hybrid of Support Vector Machine and K-Nearest Neighbors, which is termed as SVNN, is used. By using soft voting to aggregate the outputs of SVM and KNN, the proposed SVNN model improves the prediction accuracy, the degree of robustness, and adaptability. The presented methodology has achieved high classification accuracy 93.28%, precision 93.49%, recall 93.28%, F1 score 93.23%, and low RMSE of 0.2592. These metrics demonstrate that the model is reliable and of practical use for intrusion detection in dynamic, high-dimensional environments of networks, providing a proper solution to modern security network challenges.
Rainfall prediction using machine learning Sharmila Devi Mandalapu, Farooq Sunar Mahammad, V. Lakshmi Chaitanya, P. Suguna Reddy, C. Sai Deepika, G. Pravalika, J. Jyoshna Devi, B. Kavya Aip Conference Proceedings, 2024
