Cloud Data Loss Prevention (DLP) can detect and classify sensitive data within structured content such as CSV or JSON. By inspecting or de-identifying as a table, the structure and columns provide Cloud DLP with additional clues that may enable it to provide better results for some use cases.
Inspecting a table
The code samples below demonstrate how to check a table of data for sensitive content. Tables support a variety of types.
See the JSON quickstart for more information about using the Cloud DLP API with JSON.
JSON Input:
POST https://dlp.googleapis.com/v2/projects/[PROJECT_ID]/content:inspect?key={YOUR_API_KEY}
{
"item":{
"table":{
"headers": [{"name":"name"}, {"name":"phone"}],
"rows": [{
"values":[
{"string_value": "John Doe"},
{"string_value": "(206) 555-0123"}
]}
],
}
},
"inspectConfig":{
"infoTypes":[
{
"name":"PHONE_NUMBER"
}
],
"includeQuote":true
}
}
JSON Output:
{
"result": {
"findings": [
{
"quote": "(206) 555-0123",
"infoType": {
"name": "PHONE_NUMBER"
},
"likelihood": "VERY_LIKELY",
"location": {
"byteRange": {
"end": "14"
},
"codepointRange": {
"end": "14"
},
"contentLocations": [
{
"recordLocation": {
"fieldId": {
"name": "phone"
},
"tableLocation": {
}
}
}
]
},
"createTime": "2019-03-08T23:55:10.980Z"
}
]
}
}
Text versus structured text
Structuring text can help provide context. The same request as the one in the previous example, if inspected as a string—that is, as just "John Doe, (206) 555-0123"—would provide less accurate findings. That's because Cloud DLP has fewer contextual clues about what the purpose of the number might be. When possible, consider parsing your strings into a table object for the most accurate scan results.
