RFID Dataset

Title:

123

RFID Dataset

Creator:

Bin Hu

Tianming Zhao

Yan Wang

YIngying Chen

Format:

.h5py File

Description:

These datasets were used for the paper "BioTag: robust RFID-based continuous user verification using physiological features from respiration" published in MOBIHOC 2022. (https://dl.acm.org/doi/10.1145/3492866.3549718)

Abstract:

Our proposed user verification approach “BioTag” captures unique physiological characteristics rooted in the users’ respiration motions by using two RFID tags attached to a user’s chest and abdomen. “BioTag” follows two approaches: 1) it adopts respiratory feature extraction methods based on waveform morphology analysis and fuzzy wavelet transformation (FWPT) to derive unique biometric information from the user’s respiration signals. 2) it trains an adaptive classifier using the gradient boosting decision tree (GBDT) to identify legitimate users and attackers accurately. “BioTag” achieves over 95.2% and 94.8% verification accuracy on random attack and imitation attack scenarios, respectively. To evaluate the performance of “BioTag”, we have created a standard dataset involving 41 participants. The dataset can be used by fellow researchers to reproduce the original work or to further explore other machine-learning problems in the domain of RFID Signals.

Experimental Setup:

We attach two RFID tags to a participant’s clothes in the chest and abdomen areas as shown in Figure 3(a). During each experiment, a participant sits on a chair that is 1m in front of the antenna. We keep the antenna at the same height as the participant’s chest as shown in Figure 3(b). Participants are asked to breathe normally during the experiments. To evaluate our system’s robustness in different environments, we conduct experiments in four types of indoor spaces including a bedroom with a twin-size bed (4.2m×4.3m), a typical lab with office furniture and (3.0m×5.0m), a corridor with no obstacle (2.8m×2.8m), and a home office with office furniture (7.0m×4.0m).

Data Collection:

We conduct extensive experiments with 41 participants (i.e., 33 males and 8 females, aged from 12 to 70) for 3 days at different times across 5 months. Each participant takes part in 10 experiments, each of which lasts 60s. We also collect about 200 − 300 respiration segments for imitation attacks, treating 1 participant as a legitimate user and 15 participants as the attackers. Unless mentioned otherwise, we use 70% respiration segments of each legitimate participant for training and the rest of the segments for testing.

Dataset Description:

We are releasing one dataset: recordings of RFID signal data collected from a commodity RFID reader Impinj R420, equipped with a directional antenna Laird S9028PCL. The RFID signals hop among 50 channels within a spectrum from 902.5MHz to 927.5MHz.

The dataset has 8 columns: