Exploring Security Vulnerabilities of Deep Learning Models by Adversarial Attacks

Table 6

Accuracy changes when the num of replaced words increases (Amazon review dataset).

Num of replaced words	0	10	20	30	40	50
LSTM	0.8848	0.7891	0.7500	0.6875	0.6406	0.6484
BiLSTM	0.8855	0.7656	0.7266	0.6953	0.6328	0.5781