LAB ASSIGNMENT 1

AI Cyber Security Programming

GitHub: https://github.com/Hiresh12/UMKC/tree/master/CSEE5590%20-%20AI%20Cyber%20Security/LAB1/Source

Name: Hiresh Jakkala Bhaskar

Class ID : 6

Introduction:

Local Differential Privacy(LDP) is the randomness applying at the data level(individual data), while collecting the data we are going to apply some randomness so that no one can find which data is true and this way we can provide privacy to the user’s data.

Objectives :

• To apply various range of noise to the data and calculate the original and augmented data’s result. • Applying different noise value to the various db size of 10,100,1000,10000,100000.

Code:

Build the Original and Argumented Database, Create DB will create random values for the database of size parameter and augmented DB created using the following logic

1. If First coin flip = head then original value
2. Else , second coin flip = head then yes else no

Compute the average of the data in the Augmented and original DB

Truth values calculated based on the following equation,

Observe the mean values for different range of noise and DB size values:

Observation Looks like,

Conclusion:

From the above observation we can conclude that, • If noise increases ,privacy of the data increases ,which will reduce the accuracy of the result we are computing from a aggregate function. • We can observe that when DB increase the accuracy of the data increases Therefore the local differential privacy works well for large database.