Differential Privacy

• Using the Laplace mechanism
• Dataset adult.csv

Definitions

Note:[Column Name] refers to replacing it with the actual column name.

Pairwise Comparison

• ([Column Name]:[Value])

• Example: (age:30) means comparing age with 30

Comparison Operators

• {[Comparison Operator]:[Pairwise Comparison]}

Symbol	Operators
$gt	>
$lt	<
$eq	=

Note: Other operators can use the above comparison operators and the not operator below for inference.

• Example: {$gt:(age:30)} is equivalent to age>30

Logical Operators

• {[Logical Operators]:[[Sub-Operator],..]}

Symbol	Operator	Number of Sub-Operators
$and	and	2
$or	or	2
$not	not	1

• Example 1: ​{$and:[{$gt:(age:30)},{$lt:(age:50)}]} is equivalent to age>30 and age<50

• Example 2: ​{$not:[{$gt:(age:30)}]} is equivalent to age<=30

• Example 3 ​

{
​$or:[
​ {$gt:(age:30)},
​ {$and:[
​         {$lt:(age:20)},
​         {$not:[
             {$eq:(workclass:Private)}
               ]}
        ]
​    ]
}

Equivalent to:

(age>30) or (age<20 and not workclass==Private)

Full Query

• [Column].[Query Type]([Operator])
• Example: age.mean({$eq:(workclass:Private)}) returns the average age of records with wordclass as Private Note: If the Operator is left blank, the result will compute on the entire dataset.

List of Query Types

• min: minimum value
• max: maximum value
• mean: mean
• count: count
• median: median
• sum: sum
• variance: variance
• mode: mode

Note: For numeric queries, the column data must be numeric.

Installation

Use pip to install the necessary packages.

pip install -r requirements.txt

Sử dụng

Use python to run the program.

python main.py

Contribution

For change requests, please contact me and discuss the issue with me.