Table 10 Effect of similarity threshold range on performance, balance, and runtime.

From: Optimizing IoT intrusion detection with cosine similarity based dataset balancing and hybrid deep learning

Similarity range

Accuracy

Precision

Sensitivity

F1 Score

Normal

Attack

Time (s)

0.0–0.2

0.9996

0.9996

1.0000

0.9998

2

12,507

46.87

0.2–0.4

0.9753

0.9892

0.9825

0.9858

1,839

12,507

34.31

0.4–0.6

0.9604

0.9820

0.9646

0.9732

4,151

12,507

62.49

0.6–0.8

0.9667

0.9769

0.9009

0.9374

32,423

12,507

162.38

0.8–1.0

0.9940

0.9684

0.9781

0.9732

99,984

12,507

303.57

Back to article page