What-App? App Usage Detection Using Encrypted LTE/5G Traffic

LTE PDCP Dataset The dataset contains LTE Packet Data Convergence Protocol (PDCP) header traces extracted from traffic generated by different mobile applications. Each file represents a single LTE data stream corresponding to one user equipment (UE). The filename indicates the application from which the traffic was captured. Dataset Description The dataset provides PDCP-layer packet header metadata without including any payload content. All files are provided in CSV format and follow a unified schema. This dataset was produced as part of the research work accepted for publication at the 26th Privacy Enhancing Technologies Symposium (PETS 2026), titled “What-App? App usage detection using encrypted LTE/5G traffic”, and is released to support reproducibility and further exploration of mobile traffic analysis methods. Fields CaptureTime: Timestamp when the PDCP PDU was captured. Direction: Transmission direction. `0` = Downlink, `1` = Uplink. SequenceNumber: PDCP sequence number of the PDU. Length: Size (in bytes) of the PDCP PDU. Example CaptureTime,Direction,SequenceNumber,Length2022-12-23 10:16:11.022517+00:00,0,0,602022-12-23 10:16:11.022574+00:00,0,1,602022-12-23 10:16:11.024118+00:00,1,0,622022-12-23 10:16:11.024340+00:00,1,1,622022-12-23 10:16:11.031566+00:00,0,2,12282022-12-23 10:16:11.031589+00:00,0,3,602022-12-23 10:16:11.031626+00:00,0,4,752022-12-23 10:16:11.032510+00:00,1,2,782022-12-23 10:16:11.035039+00:00,1,3,932022-12-23 10:16:11.039723+00:00,0,5,522022-12-23 10:16:11.039748+00:00,0,6,2822022-12-23 10:16:11.045482+00:00,1,4,542022-12-23 10:16:11.047603+00:00,0,7,522022-12-23 10:16:11.047639+00:00,0,8,282 Data Collection and Ethics This dataset were collected by the authors in controlled experimental environments. Traffic was generated using our own testbed, where mobile applications were exercised through scripts with built-in randomness to ensure variability and unpredictability in usage patterns. Traffic related to interactive applications (e.g., games) was manually generated by the authors. All experiments were conducted using anonymous accounts created solely for data collection purposes. No real user accounts were involved. The dataset contains only PDCP-layer packet header metadata. No payload content, user identifiers, personal data, or application-level content is included. Consequently, the dataset does not contain any real user data and does not expose sensitive or private information.