Driver identification using automobile sensor data from a single turn

Hallac, David; Sharang, Abhijit; Stahlmann, R.; Lamprecht, Andreas; Huber, Markus; Roehder, Martin; Sosič, Rok; Leskovec, Jure

doi:10.1109/itsc.2016.7795670

Cited by 99 publications

(84 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Similar observations have been identified by other authors [8,16]. In [8], Hallac et al mention that as the driver pool increases, the alternative approaches using Multinomial Logistic Regression and Support Vector Machines "dropped off significantly" as compared to Random Forest. In [16] that Random Forest outperforms others for majority of assemblies.…”

Section: Classification For Driversupporting

confidence: 78%

“…Enev et al also hypothesized that given an availability of enough longitudinal data, "everyone can be distinguished." In another related work on driver identification, Hallac et al [8] demonstrated that there are unique patterns in individual driving styles which can be detected even for a short drive. Hallac et al experimentally demonstrated that the vehicle turn signature is often well suited for detecting individual style.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Investigations on Driver Unique Identification from Smartphone’s GPS Data Alone

Chowdhury

Chakravarty

Ghose

et al. 2018

Journal of Advanced Transportation

View full text Add to dashboard Cite

Driver identification is an emerging area of interest in vehicle telematics, automobile control, and insurance. Recent body of works indicates that it may be possible to uniquely identify a driver using multiple dedicated sensors. In this paper, we present an approach for driver identification using smartphone GPS data alone. For our experiments, we collected data from 38 drivers for two months. We quantified the driver's natural style by extracting a set of 137 statistical features from data generated for each completed trip. The analysis shows that, for the "driver identification" problem, an average accuracy of 82.3% is achieved for driver groups of 4-5 drivers. This is comparable to the state of the arts where mostly a multisensor approach has been taken. Further, it is shown that certain behavioral attributes like high driving skill impact identification accuracy. We observe that Random Forest classifier offers the best results. These results have great implications for various stakeholders since the proposed method can identify a driver based on his/her naturalistic driving style which is quantified in terms of statistical parameters extracted from only GPS data.

show abstract

Section: Classification For Driversupporting

confidence: 78%

Section: Introductionmentioning

confidence: 99%

Investigations on Driver Unique Identification from Smartphone’s GPS Data Alone

Chowdhury

Chakravarty

Ghose

et al. 2018

Journal of Advanced Transportation

View full text Add to dashboard Cite

show abstract

“…We achieve a mean accuracy of driver classification between 75-85% for CAN traces with a length of less than 2 minutes. By contrast, most prior works [3], [2] achieved comparable result only with fewer drivers (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15) and when the exact signals could be extracted and are readily available for classification. • We propose a scalable technique to classify a large set of time series even if many of the individual time series are significantly noisy and hence lack sufficient predictive power individually.…”

Section: Introductionmentioning

confidence: 90%

Automatic Driver Identification from In-Vehicle Network Logs

Remeli¹,

Lestyán²,

Ács³

et al. 2019

2019 IEEE Intelligent Transportation Systems Conference (ITSC)

View full text Add to dashboard Cite

Data generated by cars is growing at an unprecedented scale. As cars gradually become part of the Internet of Things (IoT) ecosystem, several stakeholders discover the value of in-vehicle network logs containing the measurements of the multitude of sensors deployed within the car. This wealth of data is also expected to be exploitable by third parties for the purpose of profiling drivers in order to provide personalized, valueadded services. Although several prior works have successfully demonstrated the feasibility of driver re-identification using the in-vehicle network data captured on the vehicle's CAN (Controller Area Network) bus, they inferred the identity of the driver only from known sensor signals (such as the vehicle's speed, brake pedal position, steering wheel angle, etc.) extracted from the CAN messages. However, car manufacturers intentionally do not reveal exact signal location and semantics within CAN logs. We show that the inference of driver identity is possible even with off-the-shelf machine learning techniques without reverse-engineering the CAN protocol. We demonstrate our approach on a dataset of 33 drivers and show that a driver can be re-identified and distinguished from other drivers with an accuracy of 75-85%. 2 bit.ly/2V14YkK 3 www.wsj.com/articles/what-your-car-knows-about-you-1534564861

show abstract

“…Furthermore, these methods are not well suited for IoT applications due to storage and transmission constraints [1], [2], whereas our lowdimensional embedding is compact, runs in real-time, and can aggregate data from many sensors and across many vehicles to continually improve its state estimator. Similarly, there has been work on driver identification [13], [30] and autolabeling of data [26], but these models are typically built for only one specific purpose. They are unable to transfer across different prediction types and thus struggle to extend into more general knowledge-based tasks.…”

Section: Introductionmentioning

confidence: 99%

Drive2Vec: Multiscale State-Space Embedding of Vehicular Sensor Data

Hallac

Bhooshan

Chen

et al. 2018

2018 21st International Conference on Intelligent Transportation Systems (ITSC)

Self Cite

View full text Add to dashboard Cite

With automobiles becoming increasingly reliant on sensors to perform various driving tasks, it is important to encode the relevant CAN bus sensor data in a way that captures the general state of the vehicle in a compact form. In this paper, we develop a deep learning-based method, called Drive2Vec, for embedding such sensor data in a low-dimensional yet actionable form. Our method is based on stacked gated recurrent units (GRUs). It accepts a short interval of automobile sensor data as input and computes a low-dimensional representation of that data, which can then be used to accurately solve a range of tasks. With this representation, we (1) predict the exact values of the sensors in the short term (up to three seconds in the future), (2) forecast the long-term average values of these same sensors, (3) infer additional contextual information that is not encoded in the data, including the identity of the driver behind the wheel, and (4) build a knowledge base that can be used to auto-label data and identify risky states. We evaluate our approach on a dataset collected by Audi, which equipped a fleet of test vehicles with data loggers to store all sensor readings on 2,098 hours of driving on real roads. We show in several experiments that our method outperforms other baselines by up to 90%, and we further demonstrate how these embeddings of sensor data can be used to solve a variety of real-world automotive applications.

show abstract

Driver identification using automobile sensor data from a single turn

Cited by 99 publications

References 18 publications

Investigations on Driver Unique Identification from Smartphone’s GPS Data Alone

Investigations on Driver Unique Identification from Smartphone’s GPS Data Alone

Automatic Driver Identification from In-Vehicle Network Logs

Drive2Vec: Multiscale State-Space Embedding of Vehicular Sensor Data

Contact Info

Product

Resources

About