beacon-gui/BEACON-Dataset

BEACON Full Multimodal Release

BEACON, the Behavioral Engine for Authentication and Continuous Monitoring, is an anonymized multimodal behavioral telemetry dataset collected during VALORANT gameplay. This full release contains HID telemetry, scrubbed hardware and game configuration files, anonymized packet captures, and redacted screen recordings where available.

Dataset Summary

• Participants: 28 anonymized participant IDs (P001-P028)
• Sessions: 124 anonymized session IDs (S001, S002, ...)
• Files: 4,332
• Approximate uncompressed size: 570.2 GB
• Included modalities: mouse CSV, keyboard CSV, hardware JSON, VALORANT config, PCAP, MP4 screen recordings
• Time representation: session-relative elapsed time in released CSVs
• Raw consent files are not included

Repository Structure

participant_P001/
  session_S001/
    mouse_P001_S001.csv
    keyboard_P001_S001.csv
    captured_packets_P001_S001.pcap
    screen_record_P001_S001.mp4
    hardware_info_P001_S001.json
    valorant_config_P001_S001/
metadata/
  beacon_croissant.jsonld
  beacon_croissant_full_template.jsonld
  public_participant_session_manifest.csv
  public_release_manifest.csv
  sha256_checksums.csv
consent_policy.md
README.md

Some sessions may not have every modality. Check metadata/public_release_manifest.csv for per-session availability and release status.

Modalities

• mouse_*.csv: session-relative mouse movement, click, scroll, speed, distance, and acceleration telemetry when present in the source capture
• keyboard_*.csv: session-relative key press/release, key label, duration, polling, and consecutive-key telemetry when present in the source capture
• captured_packets_*.pcap: anonymized packet captures
• screen_record_*.mp4: redacted screen recordings
• hardware_info_*.json: scrubbed hardware metadata
• valorant_config_*: scrubbed VALORANT configuration files

Metadata

• metadata/public_release_manifest.csv: artifact-level manifest with participant ID, session ID, release path, modality, status, size, and SHA-256 checksum when available
• metadata/public_participant_session_manifest.csv: participant/session path index
• metadata/sha256_checksums.csv: checksum list for released files
• metadata/beacon_croissant.jsonld: Croissant metadata for the release
• metadata/beacon_croissant_full_template.jsonld: full-release Croissant template
• consent_policy.md: public consent and release policy summary

Intended Uses

This release supports research on multimodal behavioral telemetry, continuous authentication, dataset quality analysis, human-computer interaction during gameplay, and reproducible benchmark construction.

Out-of-Scope Uses

Do not use this dataset to identify real people, infer protected attributes, train systems for covert surveillance, or make employment, education, legal, financial, or safety-critical decisions about individuals. Do not redistribute derived versions that reintroduce direct identifiers or raw timestamps.

Limitations and Biases

The dataset comes from a specific first-person shooter, capture environment, participant pool, and hardware/software context. Behavioral telemetry may vary with player skill, role, match state, keybinds, device sensitivity, accessibility settings, fatigue, and network conditions. Results should be interpreted as evidence about this collected setting unless externally validated elsewhere.

PCAPs and videos are high-context modalities. Even after anonymization and redaction, users should handle them as sensitive research data and avoid publishing examples that expose unintended personal or third-party information.

Privacy and Responsible Release

Raw participant names, raw consent files, hostnames, MAC addresses, direct device IDs, direct account identifiers, and raw absolute timestamps are not included. Participant and session identifiers are opaque. Config and hardware files are scrubbed before release. Behavioral telemetry can still be re-identifying in some settings, so downstream users should minimize access, avoid linking with outside identity data, and report aggregate results whenever possible.

Citation

If you use this dataset, please cite it:

Singh, I., Kaur, G., Atwal, U. P. S., Singh, G., Singh, G., & Singh, M. (2026). BEACON: A Multimodal Dataset for Learning Behavioral Fingerprints from Gameplay Data [Data set]. Zenodo. https://doi.org/10.5281/zenodo.20034625

License

This release is distributed under cc-by-nc-4.0.