The rapid growth of Internet of Things (IoT) networks has increased the need for intrusion detection mechanisms that provide accurate threat detection while preserving privacy, resisting malicious clients, and limiting communication overhead. Existing Federated Learning (FL)-based intrusion detection systems (IDS) remain challenged by heterogeneous non-IID client data, exposure of model updates, adversarial client participation, and frequent model transmission. To address these limitations, this paper proposes FLZT-IDS (Federated Learning with Zero-Trust for Intrusion Detection in IoT). The framework employs gradient-deviation monitoring to mitigate the effects of heterogeneous client updates, while Graph Neural Networks (GNN) and an Attention-based BiLSTM capture relational and temporal attack patterns, respectively. Secure Aggregation with Local Differential Privacy (SALDP) protects transmitted model updates, whereas dynamic trust-based client filtering limits the contribution of potentially malicious participants. An adaptive communication mechanism transmits local updates only when the model change exceeds a predefined threshold, thereby reducing unnecessary communication. Experiments on CICIDS2017 and Edge-IIoTset show that FLZT-IDS achieves 97.8% and 96.2% accuracy, respectively. Compared with FL-IDS, the adaptive communication and trust-filtering mechanisms reduce cumulative communication overhead by more than 30% after 100 communication rounds. Under model-poisoning experiments with 30% malicious clients, FLZT-IDS maintains accuracies of 89.3% on CICIDS2017 and 87.1% on Edge-IIoTset, demonstrating improved robustness under the evaluated adversarial setting.

Sentiment analysis is one of the most important tasks in natural language processing, as it allows opinions expressed in text to be automatically categorized. This study presents a systematically controlled comparative analysis of classical machine-learning, deep-learning, feature-fusion, and fully fine-tuned transformer models for binary sentiment classification on the Internet Movie Database (IMDb) Movie Reviews dataset. While the individual components of the framework, including Term Frequency-Inverse Document Frequency (TF-IDF), Bidirectional Encoder Representations from Transformers (BERT) embeddings, feature fusion, and Multi-Layer Perceptron (MLP) classification, are established techniques, this study contributes a comprehensive empirical evaluation of their combined effectiveness for the target classification task, providing insights into the practical benefits of integrating lexical and contextual representations. Under a single experimental framework, all models are evaluated on a commonly held-out test set with 95% bootstrap confidence intervals, paired McNemar significance testing, confusion-matrix and error analysis, and a computational-complexity comparison. The fully fine-tuned transformers achieve the highest performance Robustly Optimized BERT Pretraining Approach (RoBERTa), F1-Score = 0.9427. The TF-IDF and BERT feature-fusion model is significantly more accurate than the classical, deep-learning, and BERT-only baselines (all p < 0.001), but significantly less accurate than the fine-tuned transformers (all p < 0.01); furthermore, concatenating the [CLS] token and mean-pooled representations does not significantly improve upon mean pooling alone ( p = 0.48). The study provides a transparent, evidence-based account of the accuracy and cost trade-offs among these model families.