Lane detection plays a vital role in making the idea of the autonomous car a reality. Traditional lane detection methods need extensive hand-crafted features and post-processing techniques, which make the models specific feature-oriented, and susceptible to instability for the variations on road scenes. In recent years, Deep Learning (DL) models, especially Convolutional Neural Network (CNN) models have been proposed and utilized to perform pixel-level lane segmentation. However, most of the methods focus on achieving high accuracy while considering structured roads and good weather conditions and do not put emphasis on testing their models on defected roads, especially ones with blurry lane lines, no lane lines, and cracked pavements, which are predominant in the real world. Moreover, many of these CNN-based models have complex structures and require high-end systems to operate, which makes them quite unsuitable for being implemented in embedded devices. Considering these shortcomings, in this paper, we have introduced a novel CNN model named LLDNet based on an encoder–decoder architecture that is lightweight and has been tested in adverse weather as well as road conditions. A channel attention and spatial attention module are integrated into the designed architecture to refine the feature maps for achieving outstanding results with a lower number of parameters. We have used a hybrid dataset to train our model, which was created by combining two separate datasets, and have compared the model with a few state-of-the-art encoder–decoder architectures. Numerical results on the utilized dataset show that our model surpasses the compared methods in terms of dice coefficient, IoU, and the size of the models. Moreover, we carried out extensive experiments on the videos of different roads in Bangladesh. The visualization results exhibit that our model can detect the lanes accurately in both structured and defected roads and adverse weather conditions. Experimental results elicit that our designed method is capable of detecting lanes accurately and is ready for practical implementation.

Filtered Non-Orthogonal Multiple Access (F-NOMA) is a multi-carrier wave form and is considered a suitable contender for 5G radio. Peak to average power ratio (PAPR) is regarded as a major hurdle in the NOMA wave form because it hampers the efficiency of the power amplifier of the NOMA transmitter. In this study, a novel selective mapping (SLM) algorithm is used to minimize the PAPR of the NOMA. The conventional SLM increases the intricacy of the structure, and the projected SLM algorithm is applied to the transmitter part of the F-NOMA. Furthermore, we evaluate the performance of SLM on F-NOMA for 16, 64, and 256-Quadrature Amplitude Modulation (QAM) transmission methods. The parameters such as Bit Error Rate (BER), PAPR, power spectral density (PSD), and complexity are estimated and compared with different transmission patterns. The simulation outcomes of the work reveal that the optimal PAPR can be achieved by selecting the sub-block (S) and phase rotation elements (Ps). PAPR in F-NOMA achieves 1 dB gain in different QAM transmissions and its saving performance is 70.07%; however, complexity increases with an increase in modulation order.

The scheme of water resources management is necessary for reducing water scarcity in arid areas and improving water availability in general. However, water leak detection and irrigation scheduling traditional AI models are often computationally intensive and require complex hyperparameter tuning, making them less scalable. This study presents an artificial intelligence-based optimization framework that improves forecasting accuracy, computational speed, and real-time adaptability. The architecture combines the ensemble-learning algorithms (XGBoost, LightGBM), hybrid AIs (XGBoost + Autoencoder), and metaheuristic feature selection (GA, PSO, SA) for making intelligent decisions. Moreover, ontology-based feature structuring enhances interpretability, while hyperparameter tuning (GridSearchCV, Bayesian Optimization) and model compression techniques (pruning, quantization, knowledge distillation) ensure computational efficiency. A large number of experiments on real-world IoT sensor data testify to the effectiveness of the framework. It achieves 0.992 AUC-ROC scores for leak detection, an RMSE of 0.227 hours for irrigation scheduling, and an overall accuracy of 94.8%. Additional performance measures comprise precision (89.0%), recall (95.2%), F1-score (0.92), and inference speed (0.003 ms/sample). Although quantization has reduced the computational overhead, we still see a 13.02% increase in the model size as seen in Experiment 6, leading to a trade-off that needs to be optimized further. This study offers a deployable AI-based model for sustainable water management by tackling the issues of scalability, computational cost, and limitations in benchmark evaluation. By virtue of the empirical validation and comparative analysis of the framework, it has been shown to perform better than the regular methods, proving that the methodology can act as a step forward in the field of real-time, AI-assisted irrigation and leak detection systems.

A considerable concern arises with the precise identification of brute-force threats within a networked environment. It emphasizes the need for new methods, as existing ones often lead to many false alarms, as well as delays in real-time threat detection. To tackle these issues, this study proposes a novel intrusion detection framework that utilizes deep learning models for more accurate and efficient detection of brute-force attacks. The framework’s structure includes data collection and preprocessing components performed at the outset of the study using the CSE-CICIDS2018 dataset. The design architecture includes data collection and preprocessing steps. Feature extraction and selection techniques are employed to optimize data for model training. Further, after building the model, various attributes are extracted from the data from feature selection to be used in the training. Then, the construction of multiple architectures of deep learning algorithms, which include Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Long Short-Term Memory (LSTM) models. Evaluation results show CNN and LSTM achieved the highest accuracy of 99.995 Parsant and 99.99 Parsant respectively. It showcases its ability to detect complex attack patterns in network traffic. It indicates that the CNN network got the best optimum results with a test time of 9.94 seconds. This establishes CNN as an effective method, achieving high accuracy quickly. In comparison, we have surpassed the accuracy of current methods while addressing their weaknesses. The findings are consistent with the effectiveness of CNN in brute-force attack detection frameworks as a more accurate and faster alternative, increasing the capability of detecting intrusions on a network in real-time.

This paper presents an extensive comparative analysis of several image steganography methods, examining their efficiency in terms of capacity, media quality, resistance to detection, and computational efficiency. Steganography, the science of hiding sensitive data within a larger image, is vital for secure data transmission. The research compares five of the most used techniques: Least Significant Bit (LSB), RGB, Pixel Value Differencing (PVD), Feature-Based Optimised Steganography, and the VisionStego AI System. The introduction emphasises the importance of cryptography and steganography in concealing data. Cryptography encrypts the data, whereas steganography hides data within multimedia content. Image steganography conceals information by distorting cover images in a way that makes the information difficult to retrieve. Steganalysis is crucial in the retrieval of concealed information. LSB, RGB, and PVD are traditional steganography methods that suffer from drawbacks such as low capacity, media degradation, and detectability. This study will contrast and compare the performance of LSB, RGB, PVD, Feature-Based Optimized approaches, and the VisionStego AI System. A literature review explains various methods in steganography and their evolution. Scientists have discovered that combining encryption and steganography provides enhanced data security. Advanced techniques, such as CNN-based and GAN-based methods, have replaced traditional approaches. Hashing techniques and robust linked list steganography have been proposed to offer greater protection. Generative adversarial networks (GANs) and invertible steganography networks (ISNs) have emerged as methods to enhance capacity, security, and robustness. The methodology section compares five steganography techniques: LSB, RGB, PVD, Feature-Based Optimized Steganography, and VisionStego AI System. The performance of each method is based on capacity, media quality, detection resistance, and computational complexity. The comparison metrics are Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), Mean Squared Error (MSE), embedding time, and extraction time. The LSB technique substitutes the least significant bit of the cover image with secret information. It is simple and fast, but has low capacity and is detectable. The test results indicate that LSB offers high-speed embedding and good quality, but has low capacity and security. The RGB technique disperses the secret data in the red, green, and blue components of the cover image. It offers better data hiding quality and fewer visual distortions compared to LSB, but with a reduced extraction rate; hence, it is not suitable for applications that require rapid data retrieval. The PVD technique conceals information by looking for pixel value differences in the cover image. It is less detectable to manipulation and reduces visual noise. But it is more complicated to operate and can be identified when image enhancement techniques are applied. Feature-Based Optimized Steganography uses information embedding based on some features of multimedia, i.e., image corners and edges. It incorporates data encryption and compression before information embedding, lowering PSNR, SSIM, and MSE. The scheme offers satisfactory security and minimum storage space, but increases extraction and embedding times due to compression and encryption. The VisionStego AI System utilises artificial intelligence to automatically calculate effective places within the image where data must be embedded. It embeds text data in a digital format and calculates areas of importance in the image through saliency detection. It offers good security, optimised storage size, does not distort the image, and provides simple retrieval, along with outstanding security. The outcomes of the performance test reveal that the VisionStego AI System achieves a satisfactory compromise between data concealment quality, storage efficiency, and processing efficiency. The discussion and results section demonstrate that the VisionStego AI System outperforms other systems in preserving image quality after embedding. It achieves high PSNR and SSIM, which are metrics for zero and imperceptible distortion. Feature-Based Steganography maximizes the spreading of data, thereby enhancing concealment performance with no impact on visual coherence. Low measures of MSE confirm that the embedding procedure ensures a subtle and almost imperceptible modification of the image. The VisionStego AI System can offer hiding capacity preservation while maintaining image quality, providing large-scale data embedding in a lossless, imperceptible form. VisionStego AI is furthermore less vulnerable to steganalysis attacks compared to conventional LSB-based methods. With the addition of encryption and compression processes, this further enhances its security. The conclusion and future work focus on how the projected VisionStego AI System exhaustively surpasses LSB-based techniques in terms of concealment capability and confidentiality, while retaining higher image fidelity and resilience against steganalysis attacks. The system's ability to perform optimal data embedding without distorting the structural integrity of the image is evident from the high values of PSNR and SSIM. Future research could focus on optimising the embedding process according to other image characteristics, conducting additional experiments on different types of images, and enhancing detection resistance and extraction resistance using higher-level encryption and obfuscation techniques.

Brute force attacks remain one of the most prevalent and effective methods cybercriminals use to gain unauthorized access to networks and systems. These attacks involve systematically attempting various password or key combinations until the correct one is identified, often targeting critical services such as FTP (File Transfer Protocol) and SSH (Secure Shell). The consequences of these attacks can be severe, including data breaches, financial losses, and reputational damage. Intrusion Detection Systems (IDS) play a crucial role in mitigating these threats by monitoring network traffic and identifying malicious activities. However, traditional IDS methods - such as signature based detection and anomaly detection - struggle to detect emerging and evolving threats. To address these challenges, this study presents an advanced detection model utilizing deep learning techniques, specifically a Probabilistic Neural Network (PNN), to identify brute force attacks on FTP and SSH protocols. The model is trained and evaluated using the CICIDS2018 dataset, with the Bat Optimization Algorithm employed to fine-tune parameters and enhance performance. The proposed model achieves remarkable results, with an accuracy of 99.968%, precision of 99.949%, recall of 99.986%, and an F1-score of 99.968%. These findings highlight the model's potential as a highly effective tool for strengthening network security and preventing unauthorized access.