Table 1 Algorithm Symbols List.
Symbols | Details |
|---|---|
M | The original malware sample |
\(G=(V, E)\) | Control flow graph of M |
\(\{F\}\) | Feature vector set of M |
\(M'\) | The modified malware sample |
\(M_a\) | The adversarial malware |
\(G_a=(V_a, E_a)\) | Control flow graph of \(M_a\) |
\(\{F_a\}\) | Feature vector set of \(M_a\) |
f | Function to modify M to \(M_a\) |
L | Objective function |
\(\theta\) | Parameter optimisation by reinforcement learning |
\(\lambda _i\) | Coefficients for weighing objective i |
\(g(M')\) | Functions to determine whether \(M'\) can succeed against |
\(d(M, M')\) | Function to calculate the additional load introduced by \(M'\) compared to M |
\(opcode_i\) | The opcode to be modified in M |
\(addr_i\) | The address of the next opcode of \(opcode_i\) |
\(sequence_b\) | Benign opcode sequence |
\(sequence'_b\) | Preprocessed benign opcode sequences |
\(addr_b\) | Insertion address for \(sequence'_b\) |
