ICP4

https://github.com/Hiresh12/UMKC/tree/master/CSEE5590%20-%20AI%20Cyber%20Security/ICP4/Source

Local Differential Privacy

LDP is a type of differential privacy by adding some randomness to the data while collecting from the individuals

For example collecting data from the people using 2 coins flip and rules are,

• If 1st coin is head – record honest value
• Else if 2nd coin is head – mean yes to the question else no.

Below is the code snippet to create original DB and augmented DB based on the coin rule discussed above,

Defining the sum and mean function,

Now we will see how the results varies based on the size of the size on the sum and mean operations, Below is the sample query to how to calculate the sum and mean from the DB

Below function is used to calculate sum and mean of DB with different size,

Results for DB size of 100,

Results for DB size of 1000,

Results for DB size of 10000,

Results for DB size of 100000

Conclusion:

From the above results we can conclude that the accuracy of the data increases with increase in DB size and we can conclude that the LDP is good for large amount of data.