Graph Embedding-Based Sensitive Link Protection in IoT Systems

Algorithm 3

Generate link protection graph embedding.

Input:: the original graph

  : the adjacency matrix of training graph

  : the privacy embedding

Output:: the link protection graph embedding

  for each epoch do

   Generate the adjacency matrix of training graph

3:  Input and to the encoder to generate

   Input to the softmax classifier

   Adding or concatenating and to form

6:  Input to the inner-product decoder

   Input and the Gaussian samples to the discriminator

   Update the generator by minimizing

9:  Update the discriminator by minimizing

  end for

  return