Forgery-Resistant Touch-based Authentication on Mobile Devices

Neil Zhenqiang Gong, Mathias Payer, Reza Moazzezi, Mario Frank

TL;DR

This paper introduces a novel touch-based biometric authentication system for mobile devices that is resistant to forgery attacks. By subtly altering screen settings, the system leverages users' subconscious adaptation of their touch behavior, making it difficult for attackers to replicate legitimate user interactions even if they possess recorded touch data.

What Problem Does This Paper Address?

Mobile devices are increasingly central to our lives, storing sensitive personal data and controlling a growing ecosystem of Internet-of-Things (IoT) devices. Securing access to these devices is paramount. While passwords are a common solution, they are often inconvenient, leading users to disable them. Existing biometric methods like fingerprint or facial recognition are also vulnerable to sophisticated forgery attacks. Even touch-based biometrics, which offer passive authentication by analyzing how a user interacts with the screen, have been shown to be susceptible to replay attacks where an attacker records and replays a user's touch patterns. The core challenge is to create a biometric system that is both secure against forgery and user-friendly.

Methods

The paper proposes an "adaptive touch-based continuous authentication" system. The core innovation lies in dynamically altering screen settings, which subtly influences how touch inputs are interpreted by the device's operating system. These settings act as a "random secret" that the user unconsciously adapts to.

Here's how it works:

  • Screen Settings as Secrets: The system manipulates parameters like horizontal or vertical distortion of touch coordinates. For example, a setting of "0.8 Y-distortion" means a 1cm vertical touch movement is registered as 0.8cm by the application.
  • User Adaptation: Users, without consciously realizing it, adjust their touch behavior (e.g., stroke length, speed) to achieve their intended actions on the device, effectively compensating for the screen setting changes. This adaptation is subconscious, ensuring a smooth user experience.
  • Stability and Sensitivity: The researchers found that users' touch behaviors exhibit two key properties under these varying settings:
  • Stability: A user's touch patterns in one setting are more similar to their own patterns in another setting than to another user's patterns in either setting. This ensures individual users can be distinguished.
  • Sensitivity: A user's touch patterns in different settings are sufficiently distinct to be separable in a feature space. This is crucial for detecting deviations.
  • Adaptive Authentication:
    • Registration Phase: The system samples a set of screen settings where the user's behavior is both stable and sensitive. It then trains classifiers for each setting to distinguish the user's touch behavior from others.
  • Authentication Phase: During normal use, the system randomly selects one of these screen settings at random time intervals. When a user interacts with the device, the system uses the classifier corresponding to the current random setting to authenticate them.

This dynamic selection of settings is the key to forgery resistance. An attacker who has recorded touch data from a user in one or more settings cannot know which setting the system is currently using, making it difficult to craft a successful replay attack.

Datasets

The study collected data from 25 subjects. They implemented an Android application that allowed them to configure five different screen settings (0.8, 0.9, 1.0, 1.1, 1.2 distortions along either the X or Y axis). For each subject and each setting, they collected both horizontal and vertical touch strokes while the subjects performed two tasks: one requiring horizontal swipes and another requiring vertical swipes. This dataset was used for training classifiers and evaluating the authentication system's performance against various attack scenarios.

Key Insights and Ideas

  • Subconscious Adaptation is Key: The most significant insight is that users can subconsciously adapt their touch behavior to subtle screen setting changes without noticing or being inconvenienced. This is the foundation of the system's security.
  • Screen Settings as Dynamic Secrets: The paper cleverly repurposes screen settings, typically used for accessibility or display adjustments, as a dynamic "secret" for authentication. This is a novel approach to biometric security.
  • Stability and Sensitivity are Crucial: The findings on stability and sensitivity provide a theoretical basis for why this approach works. Stability ensures individual users are distinguishable, while sensitivity allows for the detection of deviations when the wrong "secret" (screen setting) is used.
  • Forgery Resistance Through Uncertainty: The core security mechanism relies on introducing uncertainty for the attacker. By randomly switching screen settings, the system ensures the attacker cannot reliably know which touch data to replay.

Key Results

  • Improved EERs: The proposed adaptive touch-based authentication system (S-ATCA) significantly outperforms previous touch-based authentication methods. For targeted forgery attacks, it achieved 0.18 smaller Equal Error Rates (EERs) compared to previous approaches.
  • User Experience Unaffected: Experiments showed that users did not notice the transitions between screen settings, indicating no negative impact on user experience.
  • Fast Learning: The system's classifiers can be trained effectively with just a few minutes (around 30 strokes) of user data, making the registration process quick.
  • More Settings, Better Security: The EERs consistently decreased as the number of screen settings used by the system increased, demonstrating that a larger set of "secrets" enhances security.

Limitations and Future Work

  • Advanced Robots: The authors acknowledge that highly sophisticated robots equipped with advanced sensors and AI might be able to infer screen settings. Their current work focuses on commercialized robots.
  • Human Attackers: While the paper suggests human attackers would find it harder to mimic touch behavior than keystroke dynamics, it's an area for future exploration.
  • Other Screen Adjustments: The study focused on distortion. Future work could explore other screen adjustments like "sloppiness" and "jitter" for authentication.
  • Real-World Deployment: The authors suggest a large-scale study in real-world conditions and exploring more advanced attacks would be valuable.
  • Combining with PINs: For enhanced security, they suggest combining this system with PINs to block attacks when suspicious activity is detected.

Summary updated

Abstract

Mobile devices store a diverse set of private user data and have gradually become a hub to control users' other personal Internet-of-Things devices. Access control on mobile devices is therefore highly important. The widely accepted solution is to protect access by asking for a password. However, password authentication is tedious, e.g., a user needs to input a password every time she wants to use the device. Moreover, existing biometrics such as face, fingerprint, and touch behaviors are vulnerable to forgery attacks. We propose a new touch-based biometric authentication system that is passive and secure against forgery attacks. In our touch-based authentication, a user's touch behaviors are a function of some random "secret". The user can subconsciously know the secret while touching the device's screen. However, an attacker cannot know the secret at the time of attack, which makes it challenging to perform forgery attacks even if the attacker has already obtained the user's touch behaviors. We evaluate our touch-based authentication system by collecting data from 25 subjects. Results are promising: the random secrets do not influence user experience and, for targeted forgery attacks, our system achieves 0.18 smaller Equal Error Rates (EERs) than previous touch-based authentication.
BibTeX
@inproceedings{gong2016forgery,
  author = {Neil Zhenqiang Gong and Mathias Payer and Reza Moazzezi and Mario Frank},
  title = {{Forgery-Resistant Touch-based Authentication on Mobile Devices}},
  booktitle = {AsiaCCS},
  year = {2016},
  url = {https://openreview.net/forum?id=r-vFhvkIW5},
}
Figure