Haptic: the new biometrics-embedded media to recognizing and quantifying human patterns

Mauricio Orozco Trujillo, Ismail Shakra, Abdulmotaleb El-SaddikACM Multimedia 2005 · DOI

TL;DR

This paper explores a novel biometric authentication method by analyzing how users interact with haptic devices while performing everyday tasks. By quantifying subtle patterns in force, speed, and pen position during a virtual phone dialing task, the researchers demonstrate the potential for haptics to provide a unique and secure user identification.

What Problem Does This Paper Address?

As our world becomes increasingly digital, securing access to sensitive information and systems is paramount. Traditional biometrics like fingerprints and facial recognition are well-established, but they often require specialized hardware or are susceptible to spoofing. This paper addresses the challenge of developing new, less intrusive biometric methods that can leverage the growing prevalence of interactive technologies. Specifically, it investigates whether the way a person behaves while interacting with a haptic device—a technology that provides a sense of touch and force feedback—can be used to uniquely identify them, similar to how keystroke dynamics or voice patterns are used.

Methods

The core of this research lies in developing a system that can extract and analyze behavioral patterns from haptic interactions. The authors designed an experiment where participants dialed a virtual phone using a haptic device. The system captures various physical parameters during this task, including:

  • Keystroke Duration: The time taken to press each virtual button.
  • Speed of Dialing: How quickly the user moves between buttons.
  • Force Exerted: The pressure applied to each virtual key.
  • Pen Position: The precise location of the haptic stylus.

These raw data points are then processed through a framework that involves:

  1. Data Acquisition: Recording the haptic interaction data.
  2. Feature Extraction: Identifying and quantifying key behavioral characteristics from the raw data. This involves extending traditional keystroke dynamics by incorporating physical measurements like force and pen angle.
  3. Feature Selection: Choosing the most discriminative features for identification.
  4. Clustering Stage: Grouping similar user patterns using algorithms like K-means to identify distinct user profiles.
  5. Classification/Verification: Comparing new interaction patterns against stored templates to determine authenticity.

The researchers utilized the Reachin® API and hardware for the haptic interaction and developed a virtual mobile phone interface for the experiment.

Datasets

The study used a custom dataset collected from 20 university students (7 female, 13 male) who were introduced to haptic technology. Each participant performed a specific 7-digit dialing code 10 times. This dataset was used for both training and evaluating the proposed authentication system. The paper does not mention any publicly available benchmarks for haptic-based biometrics, suggesting this is a novel area of data collection.

Key Insights and Ideas

  • Haptics as a Biometric Source: The paper strongly suggests that the subtle, often unconscious, ways individuals interact with haptic devices can serve as a unique biometric identifier. This opens up new avenues for authentication beyond traditional methods.
  • Behavioral Patterns are Discriminative: Even in a seemingly simple task like dialing a phone, individual differences in force, speed, and pen control are significant enough to distinguish users.
  • Pen Position is Highly Discriminative: The study found that features related to the pen's position during the haptic interaction were the most effective in recognizing individuals, outperforming other parameters like force.
  • Force Measurement Challenges: While force is a key haptic output, the researchers noted that for most users, force applied to virtual keys fell within a narrow band, leading to higher error rates and making it less discriminative than other features.

Key Results

The proposed system achieved an acceptable performance level, with a probability of verification (PV) of nearly 80% for the users in the experiment. The analysis indicated that features related to the pen's position were the most effective for individual recognition. While dialing speed also proved to be an important parameter, force measurements were found to be less discriminative.

Limitations and Future Work

The authors acknowledge that while the results are promising, there were some difficulties in discriminating between users, particularly when relying solely on force measurements. They suggest several directions for future work:

  • Exploring Other Behavioral Approaches: Investigating new methods like Hidden Markov Models (HMM) and statistical classification to analyze the haptic data.
  • Designing More Spontaneous Scenarios: Creating experimental setups that mimic more natural, spontaneous interactions, similar to handwriting signature capture, to reduce conscious interference.
  • Considering Environmental Factors: Investigating how factors like user tiredness, happiness, or time stress might affect behavioral patterns and authentication accuracy.
  • Expanding the Dataset: Collecting data from a larger and more diverse group of users to improve the robustness of the system.

Summary updated

BibTeX
@inproceedings{trujillo2005haptic,
  author = {Mauricio Orozco Trujillo and Ismail Shakra and Abdulmotaleb El-Saddik},
  title = {{Haptic: the new biometrics-embedded media to recognizing and quantifying human patterns}},
  booktitle = {ACM Multimedia},
  year = {2005},
  doi = {10.1145/1101149.1101232},
}
Figure