sigo

SIGO

Sigo is a k-anonymity and l-diversity compliant anonymization tool.

Usage

The following flags can be used:

• --k-value,-k <int>, allows to choose the value of k for k-anonymization (default value is 3).
• --l-value,-l <int>, allows to choose the value of l for l-diversity (default value is 1).
• --quasi-identifier,-q <strings>, this flag lists the quasi-identifiers of the dataset.
• --sensitive,-s <strings>, this flag lists the sensitive attributes of the dataset.
• --anonymizer,-a <string>, allows to choose the method used for data anonymization (default value is "NoAnonymizer"). Choose from the following list ["general", "meanAggregation", "medianAggregation", "outlier", "laplaceNoise", "gaussianNoise", "swapping"].
• --cluster-info,-i <string>, allows to display information about cluster.
• --entropy <bool>, allows to choose if entropy model for l-diversity used.

DEMO

The data.json file contains the following data,

    {"x": 5, "y": 6},
    {"x": 3, "y": 7},
    {"x": 4, "y": 4},
    {"x": 2, "y": 10},
    {"x": 8, "y": 4},
    {"x": 8, "y": 10},
    {"x": 3, "y": 16},
    {"x": 7, "y": 19},
    {"x": 6, "y": 18},
    {"x": 4, "y": 19},
    {"x": 7, "y": 14},
    {"x": 10, "y": 14},
    {"x": 15, "y": 5},
    {"x": 15, "y": 7},
    {"x": 11, "y": 9},
    {"x": 12, "y": 3},
    {"x": 18, "y": 6},
    {"x": 14, "y": 6},
    {"x": 20, "y": 20},
    {"x": 18, "y": 19},
    {"x": 20, "y": 18},
    {"x": 18, "y": 18},
    {"x": 14, "y": 18},
    {"x": 19, "y": 15}

Step 1: Generalization

By using the NoAnonymizer method and --cluster-info,i we can see in which cluster the original data is located.

< data.json | sigo -k 6 -q x,y -i id > clusters.json

  {
    "x": 4,
    "y": 4,
    "id": 1
  },
  {
    "x": 8,
    "y": 4,
    "id": 1
  },

With the generalization method (general) we can see the scope of each cluster.

< data.json | sigo -k 6 -q x,y -a general -i id > generalization.json

  {
    "id": 1,
    "x": [2,10],
    "y": [3,10]
  },
  {
    "id": 1,
    "x": [2,10],
    "y": [3,10]
  },

Step 2: Anonymization

• Aggregation

< data.json | sigo -k 6 -q x,y -a meanAggregation -i id > aggregation/meanAggregation.json

Original Data	Anonymized data

< data.json | sigo -k 6 -q x,y -a medianAggregation -i id > aggregation/medianAggregation.json

Original Data	Anonymized data

• Top and Botton Codding

< data.json | sigo -k 6 -q x,y -a outlier -i id > top-bottom-coding/coding.json

Original Data	Anonymized data

• Random Noise

< data.json | sigo -k 6 -q x,y -a laplaceNoise -i id > random-noise/laplace.json

Original Data	Anonymized data

< data.json | sigo -k 6 -q x,y -a gaussianNoise -i id > random-noise/gaussian.json

Original Data	Anonymized data

• Swapping

< data2.json | sigo -k 6 -l 4 -q x,y -s z -a swapping -i id > swapping/swapping.json

Original Data	Anonymized data

l-diversity

In the examples/demo/l-diveristy folder is the data.json file containing the following data:

    {"x":15, "y":18, "z":"c"},
    {"x":10, "y":20, "z":"b"},
    {"x":6, "y":7, "z":"c"},
    {"x":12, "y":20, "z":"b"},
    {"x":2, "y":19, "z":"a"},
    {"x":18, "y":6, "z":"c"},
    {"x":2, "y":16, "z":"b"},
    {"x":4, "y":9, "z":"a"},
    {"x":18, "y":7, "z":"c"},
    {"x":9, "y":7, "z":"a"},
    {"x":13, "y":0, "z":"b"},
    {"x":17, "y":2, "z":"c"},
    {"x":8, "y":13, "z":"c"},
    {"x":14, "y":14, "z":"c"},
    {"x":12, "y":10, "z":"b"},
    {"x":4, "y":9, "z":"b"},
    {"x":7, "y":5, "z":"b"},
    {"x":18, "y":8, "z":"a"},
    {"x":15, "y":20, "z":"b"},
    {"x":16, "y":3, "z":"b"},
    {"x":10, "y":11, "z":"c"},
    {"x":7, "y":15, "z":"a"},
    {"x":19, "y":20, "z":"c"},
    {"x":14, "y":9, "z":"a"}

Assuming attributes x and y are quasi-identifiers and attribute z is sensitive data. We want our dataset to respect 6-anonymity and 3-diversity.

< data.json | sigo -k 6 -l 3 -q x,y -s z -i id > diversity.json

Usage of PIMO

SIGO considers quasi-identifiers as float numbers. Therefore, QIs of the orignal dataset must all be float number. However, we can find categories or dates that SIGO won't understand.

PIMO can be used to transform a string attribute into a sequence of float numbers (it's up to the user to create this sequence).

In the original dataSet, the attribute Year is a quasi identifier, but SIGO cannot process it.

   {
      "Name":"chevrolet chevelle malibu",
      "Miles_per_Gallon":18,
      "Cylinders":8,
      "Displacement":307,
      "Horsepower":130,
      "Weight_in_lbs":3504,
      "Acceleration":12,
      "Year":"1970-01-01",
      "Origin":"USA"
   }

With a simple masking.yml, we transform this attribute into a sequence of float numbers.

version: 1
seed: 42
masking:
  - selector:
      jsonpath: "Year"
    mask:
      dateParser:
        inputFormat: "2006-01-02"
        outputFormat: "2006"
  - selector:
      jsonpath: "Year"
    mask:
      fromjson: "Year"

DataSet after sequencing:

   {
      "Name":"chevrolet chevelle malibu",
      "Miles_per_Gallon":18,
      "Cylinders":8,
      "Displacement":307,
      "Horsepower":130,
      "Weight_in_lbs":3504,
      "Acceleration":12,
      "Year":1970,
      "Origin":"USA"
   }

(After de-identification with SIGO, the operation can be undone with another call to PIMO. Original values will be saved, using caches for example.)

Dates can be easily transformed into a sequence of floats, but one can imagine categories like colors, origin (if not a sensitive value), or even genders.

Contributors

• CGI France ✉Contact support

License

Copyright (C) 2022 CGI France

PIMO is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

PIMO is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with SIGO. If not, see http://www.gnu.org/licenses/.