Exploiting Insurance Telematics for Fun and Profit

Le Guan, Jun Xu, Shuai Wang, Xinyu Xing, Lin Lin, Heqing Huang, Peng Liu, Wenke Lee

Abstract

Nowadays, auto insurance companies set personalized insurance rate based on data gathered directly from their customers' cars. In this paper, we show such a personalized insurance mechanism -- wildly adopted by many auto insurance companies -- is vulnerable to exploit. In particular, we demonstrate that an adversary can leverage off-the-shelf hardware to manipulate the data to the device that collects drivers' habits for insurance rate customization and obtain a fraudulent insurance discount. In response to this type of attack, we also propose a defense mechanism that escalates the protection for insurers' data collection. The main idea of this mechanism is to augment the insurer's data collection device with the ability to gather unforgeable data acquired from the physical world, and then leverage these data to identify manipulated data points. Our defense mechanism leveraged a statistical model built on unmanipulated data and is robust to manipulation methods that are not foreseen previously. We have implemented this defense mechanism as a proof-of-concept prototype and tested its effectiveness in the real world. Our evaluation shows that our defense mechanism exhibits a false positive rate of 0.032 and a false negative rate of 0.013.

Introduction

Usage-Based Insurance (UBI) programs, such as Progressive’s Snapshot, have become a staple of the modern automotive landscape. By plugging a telematics device into a vehicle's On-Board Diagnostic (OBD-2) port, insurers can monitor driving behaviors—specifically "hard brakes"—to offer personalized discounts. However, this shift in the trust boundary introduces a significant security risk: once the device is in the customer's hands, the insurer loses control over the data collection environment.

The Vulnerability: Man-in-the-Middle Attacks

The core problem is that UBI devices trust the digital signals coming from the car's computer. Researchers have demonstrated that an adversary can easily insert a low-cost "Man-in-the-Middle" (MITM) hardware setup between the car and the telematics device. This setup intercepts real-time vehicle data and "flattens" dangerous driving behaviors before they reach the insurer.

The experimental setup for the man-in-the-middle attack, utilizing an ELM327 adapter and a manipulation program to falsify vehicle data.

By using off-the-shelf components like an ELM327 adapter, an attacker can monitor the OBD-2 bus for speed-reduction signals. When a hard brake occurs, the MITM box overrides the signal, reporting a gradual slowdown instead. This "semantic" attack bypasses traditional software defenses like encryption because the data itself is falsified at the source.

Proposed Defense: The Physical Anchor

To counter these data manipulation attacks, we propose an "Enhanced Telematics Device." The fundamental idea is to use physical laws as a ground truth that cannot be easily forged.

Architecture of the Enhanced Telematics Device, featuring a tamper-resistant enclosure and internal sensors to provide a physical anchor for data verification.

The defense mechanism augments the telematics device with a 3D accelerometer housed within a tamper-resistant enclosure. This physical sensor provides an independent stream of data that correlates with the vehicle's movement. Because acceleration and velocity are physically linked, the device can verify if the digital speed reported by the car matches the physical forces experienced by the sensor.

Statistical Modeling

Verifying speed via acceleration is non-trivial due to "integration drift" and environmental noise (e.g., road bumps, engine vibrations). To address this, we employed a Mixture of Gaussian Regression Models. This approach treats the relationship between acceleration and speed as a statistical probability. By using a Dirichlet Process (DP) prior, the model automatically adapts to different vehicle types and driving environments, effectively filtering out "unobserved heterogeneity."

Evaluation and Results

The system was tested using over 1,000 miles of real-world driving data across vehicles from six different manufacturers. The dataset included 40 trips with manipulated speed data generated by the MITM prototype.

The results demonstrate high efficacy:

  • False Positive Rate (FPR): 0.032 (3.2%), indicating the system rarely flags legitimate driving as fraudulent.
  • False Negative Rate (FNR): 0.013 (1.3%), showing extreme sensitivity to data tampering.

Even if a single data point is missed, the model consistently identifies the overall "hard brake event," which is the primary metric used for insurance rate calculation.

Conclusion

As cyber-physical systems become more integrated into daily life, securing the "semantic" integrity of data is as crucial as securing the software. By anchoring digital reports to physical sensors, insurers can reclaim the trust boundary and ensure that personalized discounts are earned through safe driving, not clever hardware manipulation.

Summary updated

Abstract

Nowadays, auto insurance companies set personalized insurance rate based on data gathered directly from their customers' cars. In this paper, we show such a personalized insurance mechanism -- wildly adopted by many auto insurance companies -- is vulnerable to exploit. In particular, we demonstrate that an adversary can leverage off-the-shelf hardware to manipulate the data to the device that collects drivers' habits for insurance rate customization and obtain a fraudulent insurance discount. In response to this type of attack, we also propose a defense mechanism that escalates the protection for insurers' data collection. The main idea of this mechanism is to augment the insurer's data collection device with the ability to gather unforgeable data acquired from the physical world, and then leverage these data to identify manipulated data points. Our defense mechanism leveraged a statistical model built on unmanipulated data and is robust to manipulation methods that are not foreseen previously. We have implemented this defense mechanism as a proof-of-concept prototype and tested its effectiveness in the real world. Our evaluation shows that our defense mechanism exhibits a false positive rate of 0.032 and a false negative rate of 0.013.
BibTeX
@inproceedings{guan2016exploiting,
  author = {Le Guan and Jun Xu and Shuai Wang and Xinyu Xing and Lin Lin and Heqing Huang and Peng Liu and Wenke Lee},
  title = {{Exploiting Insurance Telematics for Fun and Profit}},
  booktitle = {SenSys},
  year = {2016},
  url = {https://openreview.net/forum?id=diSoHJCCwE},
}
Figure