Deepfake Detection

A Deep Fake Detection API is specifically designed to identify counterfeit or fraudulent profiles during the onboarding process, primarily by detecting fake faces generated using deep fake technologies. As of the current state, this API exclusively supports video input requests.

This API is for demo purposes only and currently in active development. For production, the feature or contract may change in the future.

Deepfake Detection Object

Name
status
Type
string
Description
A short string indicating the status of the result.
1. success
2. failed
Name
result
Type
object
Description
The result of deepfake detection.
- Name
  deepfake_status
  Type
  string
  Description
  Enum code indicating the deepfake status. See Responses below for details.
  REAL
  
  FAKE
- Name
  confidence_score
  Type
  number
  Description
  Confidence score of the model producing the output (in percentage).

POST/face/v1/deepfake

Deepfake Detection

Detect fake faces generated using deep fake technologies from a video or list of frame. Use form-data if you want to request with video file. Use raw json body if you want to request with list of base 64 encoded frame.

Parameter

Name
video
Type
file (.mp4)
Description
The video file to be detected. The maximum video size is 6MB.
Name
frames
Type
list of string base64 encoded
Description
The list of frames to be detected. The maximum total size is 6MB.

Sample Request

POST

/face/v1/deepfake


    1
2
3
4

    curl -v -L -X POST 'https://api.vision.glair.ai/face/v1/deepfake' \
  -H "Authorization: Basic $(printf "%s" "USERNAME:PASSWORD" | base64)" \
  -H 'x-api-key: API_KEY' \
  -F 'video=@"/path/to/video.mp4"'

Sample Response


    1
2
3
4
5
6
7

    {
  "status": "success",
  "result": {
      "deepfake_status": "FAKE",
      "confidence_score": 0.99768
  }
}

Request ID

An associated request identifier is generated for every request made to this endpoint. This value can be found in the response headers under Request-Id

Responses

Various responses for this endpoint, in addition to general responses specified in Errors.

`200 - OK`

Success, detect deepfake in the video input.

Response


    1
2
3
4
5
6
7

    {
  "status": "success",
  "result": {
      "deepfake_status": "FAKE",
      "confidence_score": 0.99768
  }
}

`200 - OK`

Success, no deepfake in the video input.

Response


    1
2
3
4
5
6
7

    {
  "status": "success",
  "result": {
      "deepfake_status": "REAL",
      "confidence_score": 0.06772
  }
}

`400 - Bad Request`

Input video not found.

Response


    1
2
3
4

    {
  "status": "failed",
  "error_message": "Video Not found"
}

Content-Type not found in request header.

Response


    1
2
3
4

    {
  "status": "failed",
  "error_message": "Headers[Content-Type] Not Found."
}