AI Cyber Security ICP 3

https://github.com/Hiresh12/UMKC/tree/master/CSEE5590%20-%20AI%20Cyber%20Security/ICP3/Source

Differential Privacy (DP):

Differential Privacy ensures that the DL algorithms learns only what is is supposed to learn from the data while ignoring what it is not supposed to learn from the data

Create a Simple Database

`num_entries = 5000

db = torch.rand(num_entries) > 0.5`

Creating Parallel DB

Determine the sensitivity of the DB and PDB

Calculate L1 Sensitivity For Threshold

A Basic Differencing Attack

Perform a differencing attack using the sum query on row 10

db,pdbs = get_db_and_parallel(10)

creating parallel DB on row 10

Performing sum on row 10 of simple DB and parallel DB and we can see different results,

Calculating the difference of mean of simple DB and parallel DB

Performing differential attack using threshold on row 10