Flow Graph Anomaly Detection Based on Unsupervised Learning

Algorithm 1

Flow graph anomaly detection.

Input: Graph stream G

Output:abnormal object

1:function PREPROCESSOR(G)

2:      if called by Controller then

3:    3 types of Anomaly Samples A←NEGATIVESAMPLING(G)

4:    A←NODE2VEC(A)

5:    return A

6     else

7:     X←NODE2VEC(G)

8:     X←SEMODEL(X)

9:     return X

10:     end if

1l:end function

12:

13:function CONTROLLER(G)

14:   A ← PREPROCESSOR(G)

15:   for a in A do

16:      Gaussian distribution←VARIATIONAL AUTO-ENCODER(a)

17:      Save Model Gaussian distribution

18:   end for

19:end function

20:

2l:function OPTIMIZER(X-test)

22:   for x in X-test do

23:   xReconst←AUTO-ENCODER(x)

24:   xSample←GAUSSIAN DISTRIBUTION

25:   Deviation←LOSSFUNCTION(x,xReconst, xSample)

26:   end for

27:end function