Combat Mobile Evasive Malware via Skip-Gram-Based Malware Detection

Table 1

Summary of related work.

Research	Acc. (%)	Dataset size	Methods	Feature types
[8]	94.4	26,093	Boosted DT	Opcode n-gram
[9]	69	48,000	CNN	Opcode word embedding
[9]	98	2,123	CNN	Opcode word embedding
[10]	96	70,693	CNN	Embedding
[11]	98	10,868	KNN	API call sequence opcode skip-gram CBOW
[13]	97.7	110,438	MLP + LSTM	APK XML opcode embedding
[14]	98	19,979	XGBoost	Skip-gram-based opcode n-gram
[18]	97	59,749	SVM	HIN metagraph2vec
[19]	98	—	Multiview NN	Skip-gram
[20]	99	190,696	HG learning	API call graph heterogeneous graph
[21]	91	14,416	Clustering + DNN	Skip-gram CBOW
[22]	97.49	2520	SVM	Skip-gram
[23]	98.86	58,139	DNN	API call graph skip-gram