Skip to content

Latest commit

 

History

History
70 lines (48 loc) · 5.27 KB

PRIVACY.md

File metadata and controls

70 lines (48 loc) · 5.27 KB

Privacy

This document describes how player data is used and stored in SkyEye.

How SkyEye Works

SkyEye works by connecting to TacView Exporter and SimpleRadio-Standalone Server to read gameplay, audio and radio data and transmit audio data.

flowchart TD
    dcs[DCS World Server] -->|gameplay data| tacview[TacView Exporter]
    tacview -->|gameplay data| skyeye[SkyEye]
    srs[SRS Server] <-->|audio and radio data| skyeye
    skyeye <-.->|audio data| openai["OpenAI Platform (optional)"]
    player[You] <-->|gameplay| dcs
    player <-->|voice| src[SRS Client]
    src <-->|audio and radio data| srs
Loading

Chat Integration

The server administrator can optionally enable SkyEye to directly read in-game chat messages from DCS World. These are used for players who are not able to use their voice to trigger commands.

flowchart TD
    dcs[DCS World Server] -->|chat messages| grpc[DCS-gRPC server]
    grpc -->|chat events| skyeye
    player[You] <-->|chat messages| dcs
Loading

Discord Integration

The server administrator can optionally enable SkyEye to publish trace messages to a Discord text channel. These messages include text transcriptions of voice audio, and are useful for players to self-troubleshoot speech recognition problems.

flowchart TD
    player[You] <-->|voice| src[SRS Client]
    srs[SRS Server] <-->|audio| skyeye[SkyEye]
    src <-->|audio| srs
    skyeye -->|transcriptions| discord[Discord]
Loading

Your Data in SkyEye

SkyEye connects to the SimpleRadio Standalone server (just like the official SRS-Client software) and a TacView real-time telemetry service (just like the official TacView client software). SkyEye has access to the same information these applications have, including:

  • Your name inside both DCS and SRS
  • Unencrypted voice audio in SimpleRadio
  • Gameplay activity inside DCS and SRS including UnitID, vehicle name, in-game position data and your selected radio channels

Note: The "encryption" feature inside SRS is not an actual encryption system. It does not actually protect your voice audio from being read by other SRS clients.

In order to function, SkyEye temporarily buffers audio data from SRS broadcast on the configured SkyEye frequencies. SkyEye then uses either local speech recognition or cloud-based speech recognition to transcibe the audio into text. The server administrator chooses which form of speech recognition to use.

If local speech recognition is used, the audio data is used as input to an AI model within SkyEye's internal memory. This audio data is discarded immediately after it is transcripted into text, usually within seconds. The audio is never saved to disk or to a database.

If cloud-based speech recognition is used, the audio data sent over the Internet to the OpenAI Audio Transcription API. The data is sent over an encrypted connection. OpenAI's servers then perform the audio transcription and send the transcripted text back to SkyEye over the same encrypted connection. Please read OpenAI's Enterprise Privacy Policy for more information on how OpenAI handles data within their API.

SkyEye outputs logs so server operators and developers can troubleshoot and improve SkyEye. The logs may be stored as long as the operator of the SkyEye server chooses to do so. These logs may include:

  • Your name inside both DCS and SRS
  • Automated transcriptions of voice audio in SimpleRadio on the configured SkyEye frequencies
  • Gameplay activity inside DCS and SRS including UnitID, vehicle name, in-game position data and your selected radio channels
  • If the in-game chat integration is enabled, in-game chat messages

SkyEye prioritizes speed over accuracy, so the transcriptions are not accurate and may contain things that were not actually said. For example, the codeword "Fox" is commonly mis-transcribed as "F**ks". The speech recognition technology used is infamous for its inaccuracy. SkyEye performs a great deal of post-processing of the transcriptions to best-guess at the player's intended selection from a limited set of commands; it is not a general-purpose dictation service. This software is not intended to be an accessibility tool for the hearing-impaired or a moderation tool.

The server operator may change the content of these logs using the enable-transcription-logging option. If this option is disabled, full text transcriptions are not logged. Small tokens of the text may still be logged if they are interpreted as part of a bot command. Transcriptions which do not seem to be directed at the bot will not be logged at all, and the full text of each message will be excluded from the logs.

If the Discord integration is enabled, the transcriptions are also sent as messages to a Discord text channel along with some debugging data. The Discord server administrator may choose to make these messages visible to members of the Discord server. Please read the Discord Terms of Service, Discord Privacy Policy, and this article on data retention practices for more information on how Discord handles data.