Deep Learning in Cybersecurity: A Hybrid BERT–LSTM Network for SQL Injection Attack Detection

Table 7

This is a table, T. Comparison of resources consumed by different model training and test.

Model	Environment	Time of training	Time of test
BERT-LSTM	Python3.5/GPU	204.3 s	51.7 s
Word2Vec-LSTM		198.6 s	32.2 s
GRU		247.1 s	49.9 s
CNN		279.5 s	76.2 s
SVM	Python3.5/CPU	143.8 s	18.2 s
BERT-LSTM	Python2.7/CPU	407.6 s	142.6 s
Word2Vec-LSTM		392.3 s	123.1 s
CNN		430.7 s	158.7 s
GRU		467.2 s	182.9 s