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

February 09, 2026

Reading time: 2 minute

...

📝 Original Info

Title: Learning Robust Representations for Malicious Content Detection via Contrastive Sampling and Uncertainty Estimation
ArXiv ID: 2512.08969
Date: 2025-12-01
Authors: Elias Hossain, Umesh Biswas, Charan Gudla, Sai Phani Parsa

📝 Abstract

We propose 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, stabilizing training using positive anchors, and adapts temperature parameters to batch-level variability. Applied to malicious content classifications, UCF-generated embeddings enabled multiple traditional classifiers to achieve over 93.38% accuracy, precision above 0.93, and near-perfect recall, with minimal false negatives and competitive ROC-AUC scores. Visual analyses ...

📄 Full Content

As digital ecosystems grow increasingly complex and interconnected, detecting malicious content, such as adversarial malware [1], fraudulent URLs [2], and coordinated misinformation, has emerged as a critical challenge. For traditional Machine Learning (ML) algorithms [3], it is challenging to detect malicious content using only a few key ingredients (i.e., privacy, scalability, and labeling costs). Consequently, PU [4] [5] [6] learning has become a promising paradigm, allowing classifiers to learn from a limited set of positively labeled data while treating the r

…(Content truncated for length.)

📄 Read Full PDF on ArXiv