Learning Robust Representations for Malicious Content Detection via Contrastive Sampling and Uncertainty Estimation

2025-12-01

AI Summary

Addressing the challenges of high noise, scarce positive samples, and severe class imbalance in malicious content detection, this paper proposes an uncertainty-aware Positive-Unlabeled (PU) representation learning framework. Methodologically, it introduces a novel uncertainty-aware contrastive loss coupled with an adaptive temperature scaling mechanism, integrated with a self-attention-guided LSTM encoder to enable dynamic contrastive weight assignment and robust positive-sample anchor construction. The framework significantly enhances discriminability and robustness of the learned embedding space: downstream classifiers achieve 93.38% accuracy, precision above 0.93, and near-perfect recall—substantially reducing false negatives. Moreover, it attains superior ROC-AUC performance, demonstrating strong effectiveness and generalizability in high-noise, low-resource scenarios.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Safety and Robustness
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

We propose the Uncertainty Contrastive Framework (UCF), a Positive-Unlabeled (PU) representation learning framework that integrates uncertainty-aware contrastive loss, adaptive temperature scaling, and a self-attention-guided LSTM encoder to improve classification under noisy and imbalanced conditions. UCF dynamically adjusts contrastive weighting based on sample confidence, stabilizes training using positive anchors, and adapts temperature parameters to batch-level variability. Applied to malicious content classification, UCF-generated embeddings enable multiple traditional classifiers to achieve more than 93.38% accuracy, precision above 0.93, and near-perfect recall, with minimal false negatives and competitive ROC-AUC scores. Visual analyses confirm clear separation between positive and unlabeled instances, highlighting the framework's ability to produce calibrated, discriminative embeddings. These results position UCF as a robust and scalable solution for PU learning in high-stakes domains such as cybersecurity and biomedical text mining.

Problem

• Improves malicious content detection under noisy, imbalanced conditions
• Enhances classification accuracy and recall with robust embeddings
• Addresses Positive-Unlabeled learning challenges in cybersecurity and text mining

Innovation

• Uncertainty-aware contrastive loss for noisy data
• Adaptive temperature scaling for batch variability
• Self-attention-guided LSTM encoder for discriminative embeddings