Crowdsensed WiFi Security Corpus: Anonymised Passive Scan Observations for Graph-Based Threat Detection Research

This dataset contains 53,371 anonymised passive WiFi scan observations collected by mobile auditors in a field crowdsensing campaign. Each row records one access-point sighting with received signal strength (RSSI), frequency and channel metadata, band-specific path-loss distance estimates, vendor information, reduced-precision geolocation, administrative location labels, and heuristic threat-exposure labels. The corpus supports research on WiFi security, sparse wireless measurement topology, and graph-based detection under crowdsensing conditions. It accompanies a study on connectivity-limited graph learning, in which structural sparsity in the induced communication graph limits message-passing performance more than classifier choice.Contents wifi_crowdsensing_corpus.csv — curated scan-level observations (53,371 rows, 26 fields) DATA_DICTIONARY.md — field definitions and heuristic-label protocol LICENSE.txt — CC BY 4.0 Privacy and governance Processing follows the Mexican Federal Law on Protection of Personal Data Held by Private Parties (LFPDPPP). Hardware and device identifiers are replaced by salted SHA-256 hashes generated at release time; the salt is not distributed, so identifiers are irreversible. Raw network names (SSIDs), capability strings, and street-level address fields are excluded. GPS coordinates are reduced to three decimal places (approximately 110 m). Heuristic labels is_anomaly_heuristic = 1 flags scans associated with OPEN or WEP security, or co-located duplicate network identities under the same auditor and timestamp (evil-twin candidates). These labels proxy exposure to weak encryption and duplicate identities; they do not certify active attacks. Licence: Creative Commons Attribution 4.0 International (CC BY 4.0).