Authors - Sanae.Chakir, Adil.Bekraoui, El moukhtar Zemmouri, Hassan.Majdoubi, Mhamed. Mouqallid Abstract - For cuttings to successfully root indoor environmental conditions in a greenhouse are essential. This article examines the efficacy of two predictive models, linear regression and artificial neural networks, in predicting the parameters associated with rooting plant cuttings. For evaluation the analysis uses the RMSE MAPE and R² indices. According to the results artificial neural networks perform better than linear regression in terms of prediction accuracy. By utilizing these insights, farmers can use artificial neural network models to implement optimal control strategies which will al-low them to accurately predict indoor variables and ultimately increase crop productivity.
Authors - Miroslav Stefanov, Stoyan Denchev, Kristiyan Stefanov Abstract - Large Language Models (LLMs) are increasingly used in educational settings, but they are not inherently constrained to the boundaries of specific instructional materials. This can lead to unsupported claims, external knowledge leakage, and reduced instructional precision. This study proposes and evaluates a lesson-bounded LLM architecture for reliable educational AI systems. The architecture combines retrieval-augmented generation, context restriction, structured response control, explicit refusal behavior, and post-hoc confidence calibration. Using a multi-domain instructional dataset and a benchmark of inscope and out-of-scope questions, the proposed system is compared against an unconstrained baseline LLM. Results show strong retrieval discrimination and boundary control, with high Area Under the Receiver Operating Characteristic Curve, high Average Precision, strong refusal recall, low out-of-scope answer rate, reduced verbosity, and improved support-based instructional density. Calibration analysis further shows that raw retrieval scores are not reliable probability estimates, but Platt scaling substantially improves confidence reliability. These findings suggest that lesson-bounded architectural constraints can improve the controllability, auditability, and reliability of intelligent educational systems while highlighting the need for stronger factuality evaluation and confidence interpretation.
Authors - Hatem Yousif Alkhonini, Fethi Fkih Abstract - This study evaluates five post hoc explanation methods using XAI-Bench under controlled settings with ground-truth explanations. Results show significant differences in robustness, with MAPLE outperforming Shapley-based methods under high correlation. Feature correlation impacts explanation quality more than the choice of method, and performance degrades with increasing dimensionality-especially for LIME. Exact methods become infeasible beyond d=10, and robust evaluation requires multi-seed replication.
Authors - Maria Jose Cantos Cedeno, Kevin Michael Mero Ramirez Abstract - The spread of disinformation through social media, messaging apps, and other digital channels is a growing problem in Ecuador, due to the limitations of manual fact-checking processes in the face of the high volume of information. In this context, Transformer-based models are presented as high-potential solutions for detecting fake news across various domains and languages. The objective is to comparatively evaluate Transformer architectures pre-trained using fine-tuning techniques for the automatic classification of fake and real news in the Ecuadorian context. The CRISP-DM methodological framework was applied to guide the development of deep learning models. A balanced dataset of 5,000 news items in Ecuadorian Spanish was constructed, equally distributed between real and fake news. Data processing was carried out through a 12-stage sequential pipeline to reduce noise, prevent data leakage, and preserve relevant semantic features of Spanish. Furthermore, the models were trained using homogeneous hyperparameters and evaluated using various metrics employed in the scientific field. As a result, all models exceeded 89% accuracy; BETO achieved the highest precision, mBERT the best recall, and DistilBERT the highest computational efficiency. It is concluded that Transformer architectures proved to be scalable, effective, and viable solutions for the automatic detection of fake news in Ecuador.
Authors - Arsalan Vahi Abstract - The successful deployment of the Internet of Things (IoT) applications relies heavily on their robust security, and lightweight cryptography is considered an emerging solution in this context. While existing surveys have been examining lightweight cryptographic techniques from the perspective of hardware and software implementations or performance evaluation, there is a significant gap in addressing different security aspects, such as design principles, specific to the IoT environment. This study aims to bridge this gap. This research presents an examination with focusing on the security evaluation of symmetric lightweight ciphers commonly used in IoT systems. The objective of this study is to provide a concise overview of lightweight ciphers with emphasizing on their security challenges which is an essential consideration for real-time and resource-constrained applications.
Authors - Amine El Ameri, Ahmed Drissi Abstract - Quantum computing threatens TLS through the Harvest Now, Decrypt Later (HNDL) attack: adversaries record encrypted traffic today and decrypt it once quantum capabilities mature. Existing approaches integrate post-quantum key encapsulation mechanisms directly into the TLS handshake; while cryptographically sound, they inflate the first handshake message and cause IP fragmentation of the ClientHello record that some middleboxes reject, leaving these solutions undeployable on today’s Internet. This paper proposes PH-PQ-TLS, a post-quantum key establishment framework for TLS 1.3 that adds post-quantum confidentiality while preserving the standard handshake, thereby avoiding fragmentation and middlebox incompatibilities. The framework requires no redesign of TLS, uses only standardized TLS mechanics, and is crypto-agile by design. We provide a full Go implementation and evaluate its overhead against a standard TLS 1.3 baseline. The ClientHello record measures 289 bytes, well below the minimum IPv6 Maximum Transmission Unit of 1280 bytes, whereas a hybrid ML-KEM-768 ClientHello reaches roughly 1473 bytes and exceeds that threshold. The post-handshake upgrade adds 1.18 ms of latency per full connection, a one-time cost amortized over the session lifetime and avoided on Pre-Shared Key resumed connections. These results show that HNDL protection does not require redesigning TLS, replacing the Web PKI, or accepting deployment failures, offering a practical, incremental path toward quantum-resilient TLS on today’s infrastructure.
Authors - Sayuni Dewapriya, Pehan Gunasekara, Shenal Peiris, Charith Herath, Kavinga Yapa Abeywardena, Ayesha Wijesooriya Abstract - DNS is often trusted within modern network environments, making it a common channel for covert communication, malware activity, and infrastructure abuse. This paper presents a near-real-time AI/ML-based DNS threat detection framework designed to identify suspicious DNS behaviour and transform raw network activity into actionable security events. The proposed approach combines machine learning, behavioural analysis, flow-based detection, event aggregation, risk scoring, and contextual threat intelligence to improve visibility across plaintext DNS and DNS-over-HTTPS traffic patterns. The framework supports practical security operations by reducing raw alert noise and producing structured outputs suitable for dashboard monitoring and SIEM-based investigation. Evaluation using public datasets, generated attack traffic, and live DNS traffic demonstrates that the framework can support effective DNS threat monitoring, alert prioritisation, and SOC-level analysis.
