top of page

ivPair: Context-Based Fast Intra-Vehicle Device Pairing for Secure Wireless Connectivity

RESEARCH OVERVIEW

Connecting a mobile device to an in-vehicle infotainment (IVI) system has long been a problem that is much more bothersome than it sounds. To do this, the user has to navigate through multiple steps to discover the device to pair and enter a randomly generated pin to verify the device's authenticity. This pairing process is tedious and lengthy, and sometimes not so user-friendly and unsafe to do while driving. In ivPair, we investigates the use of vibration simultaneously measured by a vehicular computer and a mobile phone in the same vehicle to subsequently establish a secure wireless connection (e.g., Bluetooth or Wi-Fi) between them. We design and implement integral techniques to overcome challenges in realizing ivPair on commercial mobile devices, such as lack of time synchronization and sampling frequency mismatch.

PROTOCOL

The only additional hardware component required is a reference accelerometer embedded in vehicle in the direction parallel to the car. To pair with a mobile device on the go, the user simply holds the device (embedded with an accelerometer) against the moving vehicle's interior door frame. To initiate pairing, Device B (user's mobile device) transmits a pairing request to Device A (host vehicle). Following the conditioning and pin generation phases, the two devices generate identical pins to communicate through a secure encrypted channel.

ALIGNMENT & KEY EXTRACTION

The sampling rate variation between the devices results in of vibration signal misalignment as the number of measurements accumulate. To deal with this, we correct the sampling frequency discrepancy by adopting dynamic time warping (DTW) method. Each device extracts its non-linear warping path which represents the indices with minimum distance with respect to each other. 
Then the devices independently apply their warping path on measured signal, to obtain tightly aligned fingerprints with respect to each device. the figure shows signal measured from two devices and their sample-wise error before and after the proposed signal alignment. The error is drastically reduced (from root mean square error (RMSE) of 0.10 to 0.01), resulting in a significant improvement in correlation. 

Afterwards, the two time-aligned signal obtained by two devices are the main source of randomness to harvest identical bit sequences. Two signals are uniformly segmented into several subsections and if the signal value at each index is greater than the mean of the subsection, a bit 1 is extracted; otherwise a bit 0 is extracted.

EVALUATION RESULTS

Road Conditions

Device Locations

We evaluate the performance with different body types of vehicles driven on various types of roads.
More than 3-hour worth of real-world driving data is collected using triple-axis ADXL345 MEMS accelerometer connected to Arduino Uno boards at a sampling frequency of 800 Hz. As illustrated in the first figure, the high bit agreement rates lead to high success rates above 85% for all freeway and city driving in both sedan and SUV vehicle types. We also show the bit agreement rate and the pairing success rate for four different location pairs. Our experiments suggest all the passengers in the vehicle will experience an acceptable success rate of above 70% with a mean of 85%, regardless of their seat position.

Pairing Time

Adversarial Scenario

In terms of pairing time, a user sitting in the driver side of the car can expect an average pairing time of 11.0 s thanks to the higher success rate as compared to other location pairs. Overall, regardless of their sitting positions, all the users within the same vehicle can expect a reasonably short expected pairing time of less than 14 s. We also consider two adversarial scenarios: the same-lane scenario is where the adversary is actively driving with the target vehicle in the same lane and in side-lane scenario, the adversary is driving the car side-by-side in a multi-lane road.  Under both attack scenarios, none of the attempts successfully pairs with the victim's vehicle with maximum  achieved bit agreement rate of 87%.

REFERENCE

Kyuin Lee, Neil Klingensmith, Dong He, Suman Banerjee, and Younghyun Kim, "ivPair: Context-Based Fast Intra-Vehicle Device Pairing forSecure Wireless Connectivity," in Proceedings of ACM Conference on Security and Privacy in Wireless and Mobile Networks (WiSec), pp.25–30, Linz, Austria, 2020

[PDF]    [Presentation]    [Video]    [ACM link]   

bottom of page