.
Plasmonic biosensors, particularly Surface Plasmon Resonance and Surface-Enhanced Raman Spectroscopy, have gained significant attention for real-time, label-free biochemical detection. However, optimizing these sensors for maximum sensitivity and selectivity remains a challenge due to their complex plasmonic interactions with different biomolecules. This work proposes SERA, an AI driven framework that integrates machine learning algorithms with experimental Surface-Enhanced Raman Spectroscopy (SERS) data for the predictive modeling and optimization of plasmonic sensing performance. Using supervised learning techniques, the ML models are trained on a spectral dataset - SERS-DB obtained from various plasmonic nanostructures. The model predicts key parameters such as resonance shift, intensity variations, and molecular binding efficiency, allowing for rapid optimization of biosensor designs without extensive trial-and-error experimentation. This approach accelerates plasmonic biosensor development and enables real-time adaptive sensing based on live data. The results through evaluation on the SERS-DB database with 420 samples for training and 180 for the testing phase, 6 classes like Thiacloprid, Imidacloprid, Thiamethoxam, Nitenpyram, Tetrahydrofolate, and Dihydrofolate, an accuracy of 92%, precision & recall of 90%, and F1-score of 92% were attained. The SERA model excelled with an overall score of around 0.90 in all 6 classes, proving additional superiority in biosensing applications. Further comparative analysis of the proposed approach with conventional methods underscores the best performance in accuracy with 92%, sensitivity, 1000 nm/RIU, and 95% in optimization efficiency. Overall, this research highlights a scalable and cost-effective strategy for advancing biosensor technology in medical diagnostics, environmental monitoring, and bio photonics.
Active noise control systems often degrade in environments with impulsive disturbances that follow heavy-tailed statistics. To address this challenge, we introduce Alpha Score Matched Active Impulsive Noise Control (α-SMANC), a normalized filtered-x strategy that employs a surrogate score function to represent the behaviour of alpha-stable processes. Stability is reinforced through a scale tracker designed with an exponentially weighted median and median absolute deviation, eliminating the need for variance-based measures. A potential function is developed to establish theoretical bounds on step size and ensure convergence. In near-Gaussian scenarios the method behaves like normalized LMS, while under impulsive noise it maintains robustness even in the presence of secondary-path mismatch. With linear computational cost in filter length, the approach is suitable for real-time execution. Simulation studies demonstrate reduced residual error and effective suppression of large impulsive bursts.
The rapid growth of digital multimedia sharing platforms increases the demand for secure video copyright protection and unauthorized content tracking. Conventional video watermarking approaches often suffer from low robustness, limited embedding capacity, and poor resistance against geometric and signal-processing attacks. Existing methods also exhibit inadequate detection accuracy under compressed and noisy transmission environments. These limitations create significant challenges in multimedia authentication, copyright verification, and secure video communication applications. This study presents a Hybrid Deep Neural Video Watermarking Framework that integrates attention-driven watermark embedding with intelligent tamper detection mechanisms for robust multimedia security. The proposed method combines Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BiLSTM), and Transformer-based attention modules to achieve adaptive watermark insertion and reliable extraction. Initially, video frames are decomposed into spatial and temporal components through adaptive feature learning. The embedding model identifies perceptually significant regions using an attention-guided encoder, where encrypted watermark information is inserted with minimal visual distortion. Subsequently, a dual-stage decoder network performs watermark recovery and tamper localization through deep residual feature analysis. The framework also incorporates adversarial training and adaptive noise filtering to improve resilience against compression, frame dropping, Gaussian noise, rotation, and scaling attacks. Experimental evaluation demonstrates that the proposed framework achieves a PSNR of 48.7 dB, SSIM of 0.986, NC value of 0.994, tamper detection accuracy of 98.4%, and watermark recovery accuracy of 96.7% under diverse multimedia attack environments. The framework also maintains stable extraction performance under compression, scaling, Gaussian noise, rotation, and frame-dropping attacks. Comparative analysis indicates that the proposed model improves robustness by 19.8%, perceptual quality by 12.4 dB, and watermark reconstruction accuracy by 18.5% compared with conventional DWT-SVD and CNN-based watermarking approaches.
Zoonotic diseases are an increasing and considerable menace in the contemporary livestock ecosystem because of late detection, hostile monitoring and absence of integrated foretelling intelligence. The current surveillance methods are concentrated more on individual parameters like animal health or environmental status and do not offer integrated, real time and context-sensitive risk assessment framework. This paper responds to this limitation by suggesting ZoonoGuard-AI, a multi-layer edge-cloud smart farming intelligence framework to predict zoonotic risk model. The system combines multi-mod data, such as physiological and behavioral patterns and environmental parameters, which allows constant monitoring of livestock. Real-time detection of anomalies and data filtering are done using edge computing modules, and cloud-based components can support large-scale data fusion and predictive analytics. A new Zoonotic Risk Index (ZRI) is presented to measure the total risk with weighted normalized indicators. The model is tested with a simulated dataset that depicts the real variability in the farm. Experimental findings indicate 91 percent classification accuracy, which is a good predictor of risk patterns at an early stage. The behavioral differences add about 60-70 percent of the overall risk and thus their significance in early diagnosis. The suggested system will facilitate scalable implementation and facilitate proactive intervention, which will improve the farm biosecurity and the One Health paradigm.
In this paper, a new decision-generation model is proposed, which is named Cipher-Adaptive Cognitive Inference Engine (CACIE) and is intended to be applied in high-stakes numerical ecosystems where trustworthiness, robustness, and interpretability are paramount. The proposed approach does not follow a standard approach of utilizing data-driven optimization and instead performs the entire process in cipher-domain representations, without directly accessing original inputs. CACIE uses many parallel encrypted number streams researching processors who effectively maintain a position of separation among their results. The framework does not demand the existence of agreement but measures non-agreement as an indicator of informational stress and handicrafts it into an operator of morphogenesis which recursively restructures the internal geometry of decision. Another major accomplishment of CACIE is its ambiguity-minimization goal which is aimed at suppressing instability of decisions in the presence of adversarial interference rather than maximizing predictive performance. This allows the system to withstand attempts to probe, imitate and manipulate without the use of stored histories or hiding variable abstractions. The method also adds verification entropy as a governance signal, using which the inference pathways are re-configured depending on the situation, that preserves the cipher integrity. The use of experimental reasoning proves that the use of CACIE is able to support consistent decision behavior even when evaluators themselves have been compromised partially because resilience occurs due to structural adjustments and not due to pre-existing defenses. The suggested solution introduces a novel twist to cipher-domain inference systems and focuses on self-modifying decision structures controlled by verification dynamics as opposed to traditional optimization principles. The suggested CACIE technique attains an overall accuracy of 96.3%, indicating exceptional decision-making reliability under safeguarded numerical conditions.
