Continuous proliferation of hate speech in different languages on social media has drawn significant attention from researchers in the past decade. Detecting hate speech is indispensable irrespective of the scale of use of language, as it inflicts huge harm on society. This work presents a first resource for classifying the severity of hate speech in addition to classifying offensive and hate speech content. Current research mostly limits hate speech classification to its primary categories, such as racism, sexism, and hatred of religions. However, hate speech targeted at different protected characteristics also manifests in different forms and intensities. It is important to understand varying severity levels of hate speech so that the most harmful cases of hate speech may be identified and dealt with earlier than the less harmful ones. In this work, we focus on detecting offensive speech, hate speech, and multiple levels of hate speech in the Urdu language. We investigate three primary target categories of hate speech: religion, racism, and national origin. We further divide these categories into levels based on the severity of hate conveyed. The severity levels are referred to as symbolization , insult , and attribution . A corpus comprising more than 20,000 tweets against the corresponding hate speech categories and severity levels is collected and annotated. A comprehensive experimentation scheme is applied using traditional as well as deep learning–based models to examine their impact on hate speech detection. The highest macro-averaged F-score yielded for detecting offensive speech is 86% while the highest F-scores for detecting hate speech with respect to ethnicity, national origin, and religious affiliation are 80%, 81%, and 72%, respectively. This shows that results are very encouraging and would provide a lead towards further investigation in this domain.

With the advent of Deep Learning-based Artificial Neural Network models, Natural Language Processing (NLP) has witnessed significant improvements in textual data processing in terms of its efficiency and accuracy. However, the research is mostly restricted to high-resource languages such as English, and low-resource languages still suffer from a lack of available resources in terms of training datasets as well as models with even baseline evaluation results. Considering the limited availability of resources for low-resource languages, we propose a methodology for adapting self-attentive transformer-based architecture models (mBERT, mT5) for low-resource summarization, supplemented by the construction of a new baseline dataset (76.5k article, summary pairs) in a low-resource language, Urdu. Choosing news (a publicly available source) as the application domain has the potential to make the proposed methodology useful for reproducing in other languages with limited resources. Our adapted summarization model urT5 with up to 44.78% reduction in size as compared to mT5 can capture contextual information of the low-resource language effectively with an evaluation score (up to 46.35 ROUGE-1, 77 BERTScore) on par with state-of-the-art models in the high-resource language of English (PEGASUS: 47.21, BART: 45.14 on XSUM Dataset) . The proposed method provided a baseline approach toward extractive as well as abstractive summarization with competitive evaluation results in a limited resource setup.

Citrus canker is among the major plant diseases caused by Xanthomonas citri which affects the quality and quantity of citrus fruit. This results in reduction of citrus production which causes a huge financial loss and livelihood of the farming community. Thus, it is critically important to build a robust, accurate, and time efficient detection method for real time identification of the disease. Due to their powerful learning capabilities and improved feature extraction, deep learning approaches have made it feasible to carry out a number of tasks related to the identification of citrus canker in citrus leaves. Previous research has primarily focused on detecting citrus canker on fruits, early detection on leaves can facilitate the adoption of preventive measures before the disease reaches a critical stage. This paper proposes a novel deep learning-based approach for determining the growth rate of citrus canker by classifying it into six distinct stages: water soaking, yellow chlorosis/initiation, chlorosis, blister formation, canker development start, canker infection (50% of the inoculated area), and canker infection (100% of the inoculated area). The proposed approach involves image conversion, size reduction, image augmentation, and the utilization of DenseNet-121. Experimental results demonstrate a classification accuracy of 98.97% using the suggested approach. The Accuracy was 98.97% with macro precession 97%, weighted precision 99%, Macro recall 98%, weighted recall 98%, macro F1_Score 97% and weighted F1_Score 98%. This study presents a unique technique for detecting and classifying the growth rate of citrus canker based on six different stages, while also calculating the temporal change in affected area of the disease in inoculated citrus leaves. Furthermore, a mathematical model is proposed to predict the disease’s growth rate at any given time, offering valuable insights for disease management and prevention.

The rapid proliferation of Internet of Things (IoT) devices has intensified the need for scalable, privacy-preserving, and interpretable intrusion detection systems (IDS). In this work, we propose a novel framework that combines Federated Learning (FL) with Explainable Artificial Intelligence (XAI) to detect malicious traffic across distributed IoT networks without centralizing sensitive data. Our IDS model employs a lightweight deep neural network trained collaboratively using FedAvg. SHAP provides global and local attributions; LIME supplies instance-level attributions complementary to SHAP. Experiments conducted on three benchmark datasets: Edge-IIoT, CIC-IoT2023, and TII-SSRC-23, demonstrated that our approach achieves 99.3%, 99.5% and 99.0% accuracy in binary classification, respectively and 97.2%, 98.0% and 96.5% accuracy in multi-class scenarios, respectively, closely matching centralized models while preserving data locality and data privacy. Moreover, the integrated XAI methods enhance the model’s transparency by identifying key traffic features contributing to each alert. These results establish FL-XAI as a practical, interpretable, and privacy-respecting IDS for real-world IoT deployments.

The inevitable revolution of the Internet of Things (IoT) and its benefits can be witnessed everywhere. Two major issues related to IoT are the interoperability and the identification of trustworthy things. The proposed Context-Aware Trustworthy Social Web of Things System (CATSWoTS) addresses the interoperability issue by incorporating web technologies including Service Oriented Architecture where each thing plays the role of a service provider as well as a role of service consumer. The aspect of social web helps in getting recommendations from social relations. It was identified that the context dependency of trust along with Quality of Service (QoS) criteria, for identifying and recommending trustworthy Web of Things (WoT), require more attention. For this purpose, the parameters of context awareness and the constraints of QoS are considered. The research focuses on the idea of a user-centric system where the profiles of each thing (level of trustworthiness) are being maintained at a centralized level and at a distributed level as well. The CATSWoTS evaluates service providers based on the mentioned parameters and the constraints and then identifies a suitable service provider. For this, a rule-based collaborative filtering approach is used. The efficacy of CATSWoTS is evaluated with a specifically designed environment using a real QoS data set. The results showed that the proposed novel technique fills the gap present in the state of the art. It performed well by dynamically identifying and recommending trustworthy services as per the requirements of a service seeker.