131 reads
Adversarial Malware Creation with Model-Based Reinforcement Learning: Appendix
by Memeology: Leading Authority on the Study of MemesApril 18th, 2024
Too Long; Didn't Read
The MEME algorithm combines malware evasion and model extraction using reinforcement learning, achieving high evasion rates and creating accurate surrogate models with minimal interaction with target models, revolutionizing adversarial malware creation and cybersecurity defenses.
Authors:
(1) Maria Rigaki, Faculty of Electrical Engineering, Czech Technical University in Prague, Czech Republic and [email protected];
(2) Sebastian Garcia, Faculty of Electrical Engineering, Czech Technical University in Prague, Czech Republic and [email protected].
Table of Links
Conclusion, Acknowledgments, and References
Appendix
A. Hyper-parameter Tuning
The search space for the PPO hyper-parameters:
– gamma: 0.01 - 0.75
– max grad norm: 0.3 - 5.0
– learning rate: 0.001 - 0.1
– activation function: ReLU or Tanh
– neural network size: small or medium
Selected parameters: gamma=0.854, learning rate=0.00138, max grad norm=0.4284,
activation function=Tanh, small network size (2 layers with 64 units each).
The search space for the LGB surrogate training hyper-parameters:
– alpha: 1 - 1,000
– num boosting rounds: 100-2,000
– learning rate: 0.001 - 0.1
– num leaves: 128 - 2,048
– max depth: 5 - 16
– min child samples: 5 - 100
– feature fraction: 0.4 - 1.0
Table 4. Hyper-parameter settings for the training of each LGB surrogate
This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license.
L O A D I N G
. . . comments & more!
. . . comments & more!
