Bilevel optimization models attacker-defender co-evolution in malware detection, cutting evasion rates from up to 90% to 0-1.89% on three families while raising attacker query costs by up to 100x.
CoRR abs/2003.03100(2020), https://arxiv.org/abs/2003.03100
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.CR 2years
2026 2verdicts
UNVERDICTED 2representative citing papers
IAT-based poisoning samples can significantly lower recall in a LightGBM malware detector trained on continuous data, while a homogeneous ensemble defense filters up to 95.6% of such attempts.
citing papers explorer
-
Adversarial Co-Evolution of Malware and Detection Models: A Bilevel Optimization Perspective
Bilevel optimization models attacker-defender co-evolution in malware detection, cutting evasion rates from up to 90% to 0-1.89% on three families while raising attacker query costs by up to 100x.
-
Gray-Box Poisoning of Continuous Malware Ingestion Pipelines
IAT-based poisoning samples can significantly lower recall in a LightGBM malware detector trained on continuous data, while a homogeneous ensemble defense filters up to 95.6% of such attempts.