Food safety remains a critical public health conceated food causing significant morbidity and mortality worldwide. Accurate and transparent risk assessment is essential to safeguard consumer health and support evidence-based regulatory decisions. This study introduces a comprehensive and novel AI framework that leverages machine learning, deep learning, and transformer-based models to classify food safety risks using the Rapid Alert System for Food and Feed (RASFF) dataset, enhanced through advanced data enrichment techniques. The enriched dataset addressed challenges such as short explanations and class imbalances, resulting in significant performance improvements across all model categories. Transformer-based models, including BERT with an accuracy of 0.978 and RoBERTa with an accuracy of 0.979, outperformed traditional machine learning methods such as logistic regression, which achieved an accuracy of 0.954, and SVM, which reached 0.959, as well as deep learning models like LSTM, which obtained 0.971, and BiLSTM, which achieved 0.97 3. Explainable Artificial Intelligence (XAI) techniques were applied to uncover critical insights into the models’ decision-making processes. Influential features such as salmonella, aflatoxins, and listeria were identified, enhancing model transparency and interpretability. The framework’s combination of data augmentation and XAI improves predictive accuracy while ensuring interpretability, making it suitable for real-time risk prioritization in regulatory surveillance. This study demonstrates the potential of AI-powered approaches for food safety risk assessment, bridging predictive accuracy with transparency to support more reliable and actionable food safety management systems. Graphic

Current commercial antivirus detection engines still rely on signature-based methods. However, with the huge increase in the number of new malware, current detection methods become not suitable. In this paper, we introduce a malware detection model based on ensemble learning. The model is trained using the minimum number of signification features that are extracted from the file header. Evaluations show that the ensemble models slightly outperform individual classification models. Experimental evaluations show that our model can predict unseen malware with an accuracy rate of 0.998 and with a false positive rate of 0.002. The paper also includes a comparison between the performance of the proposed model and with different machine learning techniques. We are emphasizing the use of machine learning based approaches to replace conventional signature-based methods.

The proper interpretation of the malware API call sequence plays a crucial role in identifying its malicious intent. Moreover, there is a necessity to characterize smart malware mimicry activities that resemble goodware programs. Those types of malware imply further challenges in recognizing their malicious activities. In this paper, we propose a standard and straightforward contextual behavioral models that characterize Windows malware and goodware. We relied on the word embedding to realize the contextual association that may occur between API functions in malware sequences. Our empirical results proved that there is a considerable distinction between malware and goodware call sequences. Based on that distinction, we propose a new method to detect malware that relies on the Markov chain. We also propose a heuristic method that identifies malware’s mimicry activities by tracking the likelihood behavior of a given API call sequence. Experimental results showed that our proposed model outperforms other peer models that rely on API call sequences. Our model returns an average malware detection accuracy of 0.990, with a false positive rate of 0.010. Regarding malware mimicry, our model shows an average noteworthy accuracy of 0.993 in detecting false positives.

Online misogyny is a significant societal challenge that reinforces gender inequalities and discourages women from engaging fully in digital spaces. Traditional moderation methods often fail to address the dynamic and context-dependent nature of misogynistic language, making adaptive solutions essential. This study presents a framework that integrates advanced natural-language processing techniques with strategic data augmentation to improve the detection of misogynistic content. Key contributions include emoji decoding to interpret symbolic communication, contextual expansion using Sentence-Transformer models, and LDA-based topic modeling to enhance data richness and contextual understanding. The framework incorporates machine-learning, deep-learning, and Transformer-based models to handle complex and nuanced language. Performance analysis highlights the effectiveness of the selected models, and comparative results emphasize the transformative role of data augmentation. This augmentation significantly enhanced model robustness, improved generalization, and strengthened the detection of misogynistic content.

Text classification remains a challenging task in natural language processing (NLP) due to linguistic complexity and data imbalance. This study proposes a hybrid approach that integrates grammar-based feature engineering with deep learning and transformer models to enhance classification performance. A dataset of factoid and non-factoid questions, further categorized into causal, choice, confirmation, hypothetical, and list types, is used to evaluate several models, including CNNs, BiLSTMs, MLPs, BERT, DistilBERT, Electra, and GPT-2. Grammatical and domain-specific features are explicitly extracted and leveraged to improve multi-class classification. To address class imbalance, the SMOTE algorithm is applied, significantly boosting the recall and F1-score for minority classes. Experimental results show that DistilBERT achieves the highest binary classification accuracy, equal to 94%, while BiLSTM and CNN outperform transformers in multi-class settings, reaching up to 92% accuracy. These findings confirm that grammar-based features provide critical syntactic and semantic insights, enhancing model robustness and interpretability beyond conventional embeddings.