Automatic Detection of Android Malware via Hybrid Graph Neural Network

Table 3

Experimental results compared with benchmarks.

Method	Year	Binary-classify				Five-classify
		Acc	P	R	F1	Acc	P	R	F1
Bi-LSTM+GCNs [43]	2017	99.23	99.54	99.11	99.32	98.36	98.10	98.42	98.25
AndrEnsemble [45]	2019	96.40	96.35	97.12	96.73	95.30	95.12	96.01	95.56
IndRNN+GCNs [11]	2020	99.15	98.82	99.21	99.01	98.50	98.42	98.30	98.36
Bi-LSTM+CNN [44]	2021	98.68	99.43	99.25	99.33	98.02	97.65	98.44	98.04
Bi-GRU+Deep_TNN+Self-attention	—	99.62	99.67	99.50	99.58	99.20	99.15	98.89	99.01