The growth of Internet of Things (IoT) gadgets has ushered in a new era of connectedness and convenience, but it has also sparked worries about security flaws. Long Short-Term Memory (LSTM) networks are used in this research's use of intrusion detection as a novel strategy to strengthen IoT security. The proposed LSTM-based model excels in detecting both known and evolving cyber-attack patterns with an accuracy rate of 98.75% and an F1 score of 98.59% in extensive experimental evaluations using the vast CIC-IoT2023 dataset, representing a varied array of IoT network traffic scenarios. This research contributes significantly to IoT security while addressing the urgent need for adaptable intrusion detection systems to defend against changing cyber threats. It is an essential step toward ensuring IoT technology's long-term development and dependability in a world that is becoming more interconnected.

The role of Intrusion Detection Systems (IDS) in the protection against the increasing variety of cybersecurity threats in complex environments, including the Internet of Things (IoT), cloud computing, and industrial networks. This study evaluates the existing state-of-the-art IDS methodologies using Deep Learning (DL) approaches, and advanced feature engineering techniques. This research also highlights the success of models such as Genetic Algorithms (GA), Particle Swarm Optimization (PSO), and Explainable AI (XAI) in improving detection accuracy as well as computational efficiency and interoperability. Blockchain and quantum computing technologies are explored to improve data privacy, resilience, and scalability in decentralized and resource-constrained environments. This work primarily identifies key challenges, including real-time anomaly detection, adversarial robustness, and imbalance datasets, to assist researchers in investigating further research opportunities. Focusing on future research in filling these gaps, proceeds toward developing lightweight, adaptive, and ethical IDS frameworks that can operate in real-time across dynamic and heterogeneous environments. In this paper, existing IDS approaches, research opportunities, and advanced cybersecurity strategies are critically synthesized to create a useful resource for academics, researchers, and industry practitioners.

An increase in cyber threats directed at interconnected devices has resulted from the proliferation of the Internet of Things (IoT), which necessitates the implementation of comprehensive defenses against evolving attack vectors. This research investigates the utilization of machine learning (ML) prediction models to identify and defend against cyber-attacks targeting IoT networks. Central emphasis is placed on the thorough examination of the CIC-IoT2023 dataset, an extensive collection comprising a wide range of Distributed Denial of Service (DDoS) assaults on diverse IoT devices. This ensures the utilization of a practical and comprehensive benchmark for assessment. This study develops and compares four distinct machine learning models Logistic Regression (LR), K-Nearest Neighbors (KNN), Decision Tree (DT), and Random Forest (RF) to determine their effectiveness in detecting and preventing cyber threats to the Internet of Things (IoT). The comprehensive assessment incorporates a wide range of performance indicators, such as F1-score, accuracy, precision, and recall. Significantly, the results emphasize the superior performance of DT and RF, demonstrating exceptional accuracy rates of 0.9919 and 0.9916, correspondingly. The models demonstrate an outstanding capability to differentiate between benign and malicious packets, as supported by their high precision, recall, and F1 scores. The precision-recall curves and confusion matrices provide additional evidence that DT and RF are strong contenders in the field of IoT intrusion detection. Additionally, KNN demonstrates a noteworthy accuracy of 0.9380. On the other hand, LR demonstrates the least accuracy with a value of 0.8275, underscoring its inherent incapability to classify threats. In conjunction with the realistic and diverse characteristics of the CIC-IoT2023 dataset, the study's empirical assessments provide invaluable knowledge for determining the most effective machine learning algorithms and fortification strategies to protect IoT infrastructures. Furthermore, this study establishes ground-breaking suggestions for subsequent inquiries, urging the examination of unsupervised learning approaches and the incorporation of deep learning models to decipher complex patterns within IoT networks. These developments have the potential to strengthen cybersecurity protocols for Internet of Things (IoT) ecosystems, reduce the impact of emergent risks, and promote robust defense systems against ever-changing cyber challenges.

One of the widening perils in network security is the Distributed Denial of Service (DDoS) attacks on the Internet of Things (IoT) ecosystem. This paper presents an enhanced Intrusion Detection System (IDS) through the proposal of an enhanced version of the long short-term memory (LSTM) model to detect DDoS attacks using honeypot-generated data. The proposed model aggregates the Conv1D, Bidirectional Long Short-Term Memory (Bi-LSTM), Bidirectional Gated Recurrent Unit (Bi-GRU), and dropout layers to extract temporal and spatial features from IoT traffic effectively. We tested the efficacy of the proposed system on a real-world IoT-DH dataset, which showed a remarkable accuracy of 99.41%, with an AUC score of 0.9999. A comparative analysis with other baseline models, such as LSTM, Bidirectional LSTM (Bi-LSTM), Gated Recurrent Unit (GRU), Recurrent Neural Network (RNN), Feedforward Neural Network (FNN), and Temporal Convolutional Network (TCN), proved that enhanced LSTM outperformed the other models. This indicates the robustness of the proposed model in correctly detecting DDoS attacks with high generalization capability for unseen traffic data. The contribution of this paper will be an addition to the deep learning techniques applied for the solution of intrusion detection systems (IDS), which will also allow the building and implementation of more efficient security mechanisms in IoT environments.

Timely, balanced, and transparent detection of retinal diseases is essential to avert irreversible vision loss; however, current deep learning screeners are hampered by class imbalance, large models, and opaque reasoning. This paper presents a lightweight attention-augmented convolutional neural network (CNN) that addresses all three barriers. The network combines depthwise separable convolutions, squeeze-and-excitation, and global-context attention, and it incorporates gradient-based class activation mapping (Grad-CAM) and Grad-CAM++ to ensure that every decision is accompanied by pixel-level evidence. A 5335-image ten-class color-fundus dataset from Bangladeshi clinics, which was severely skewed (17–1509 images per class), was equalized using a synthetic minority oversampling technique (SMOTE) and task-specific augmentations. Images were resized to $150\times 150$ px and split 70:15:15. The training used the adaptive moment estimation (Adam) optimizer (initial learning rate of $1\times {10}^{-4}$, reduce-on-plateau, early stopping), ${\ell }_2$ regularization, and dual dropout. The 16.6 M parameter network converged in fewer than 50 epochs on a mid-range graphics processing unit (GPU) and reached 87.9% test accuracy, a macro-precision of 0.882, a macro-recall of 0.879, and a macro-F1-score of 0.880, reducing the error by 58% relative to the best ImageNet backbone (Inception-V3, 40.4% accuracy). Eight disorders recorded true-positive rates above 95%; macular scar and central serous chorioretinopathy attained F1-scores of 0.77 and 0.89, respectively. Saliency maps consistently highlighted optic disc margins, subretinal fluid, and other hallmarks. Targeted class re-balancing, lightweight attention, and integrated explainability, therefore, deliver accurate, transparent, and deployable retinal screening suitable for point-of-care ophthalmic triage on resource-limited hardware.

Breast cancer is a highly common and life-threatening disease that affects people worldwide. Early and accurate diagnosis of breast cancer can enhance patients' prognosis and survival rate. This paper conducts a comparative examination of the Wisconsin Breast Cancer Diagnostic (WBCD) dataset by employing four distinct deep learning models: Feedforward Neural Network (FNN), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU). The collection consists of 569 examples of Fine Needle Aspirate (FNA) photographs of breast cancers, with each case containing thirty parameters that define the features of the cell nuclei. By doing a comparative analysis of the advantages and disadvantages of the models, we will evaluate them based on their accuracy, precision, recall, and F1-score. Based on our research, CNN achieves the best level of accuracy at 98.25%, which is followed by GRU at 97.37%, FNN at 96.49%, and LSTM at 95.61%. It is determined that CNN is the most suitable model for this task and that deep learning models are valuable and encouraging tools for diagnosing breast cancer.