Skip to main content

Table 7 Detection rate of CEAD and DAS at different workload thresholds

From: Detecting compromised email accounts via login behavior characterization

Dataset

Total pairs

Top 10%

Top 20%

Top 30%

Pairs to be analyzed

CEAD

DAS

Pairs to be analyzed

CEAD

DAS

Pairs to be analyzed

CEAD

DAS

Dataset-1

8632

863

132/136

20/136

1726

134/136

93/136

2589

135/136

136/136

Dataset-2

17580

1758

–

–

3516

–

–

5274

–

–

Dataset-3

46695

4669

1/3

0/3

9339

3/3

3/3

14008

3/3

3/3

Dataset-4

21886

2188

–

–

4377

–

–

6565

–

–

Dataset-5

40035

4003

3/6

3/6

8007

4/6

6/6

12010

5/6

6/6

Dataset-6

23246

2324

7/55

19/55

4649

13/55

55/55

6973

20/55

55/55

Dataset-7

45442

4544

2/2

0/2

9088

2/2

1/2

13632

2/2

1/2

Dataset-8

47939

4793

–

–

9587

–

–

14381

–

–

Dataset-9

31563

3156

–

–

6312

–

–

9468

–

–

  1. The bold in the table is to highlight the system that perform better