TIF applies temporal invariant learning to Android malware detectors, improving F1 by up to 8 points in the first year after deployment across multiple feature spaces.
Detecting Android Malware: From Neural Embeddings to Hands-On Validation with BERTroid
1 Pith paper cite this work. Polarity classification is still indexing.
abstract
As cyber threats and malware attacks increasingly alarm both individuals and businesses, the urgency for proactive malware countermeasures intensifies. This has driven a rising interest in automated machine learning solutions. Transformers, a cutting-edge category of attention-based deep learning methods, have demonstrated remarkable success. In this paper, we present BERTroid, an innovative malware detection model built on the BERT architecture. Overall, BERTroid emerged as a promising solution for combating Android malware. Its ability to outperform state-of-the-art solutions demonstrates its potential as a proactive defense mechanism against malicious software attacks. Additionally, we evaluate BERTroid on multiple datasets to assess its performance across diverse scenarios. In the dynamic landscape of cybersecurity, our approach has demonstrated promising resilience against the rapid evolution of malware on Android systems. While the machine learning model captures broad patterns, we emphasize the role of manual validation for deeper comprehension and insight into these behaviors. This human intervention is critical for discerning intricate and context-specific behaviors, thereby validating and reinforcing the model's findings.
citation-role summary
citation-polarity summary
fields
cs.CR 1years
2025 1verdicts
CONDITIONAL 1roles
method 1polarities
use method 1representative citing papers
citing papers explorer
-
TIF: Learning Temporal Invariance in Android Malware Detectors
TIF applies temporal invariant learning to Android malware detectors, improving F1 by up to 8 points in the first year after deployment across multiple feature spaces.