Biometric Authentication and Identification Using Keystroke Dynamics: A Survey

Salil Banerjee, Damon L. WoodardJournal of Pattern Recognition Research 2012 · DOI

TL;DR

This paper surveys the field of keystroke dynamics as a behavioral biometric for authentication and identification. It highlights the psychological underpinnings, data acquisition methods, and various algorithmic approaches, concluding that standardization of databases, features, and evaluation protocols is crucial for wider adoption and improved performance.

What Problem Does This Paper Address?

In an increasingly digital world, securing sensitive information stored on computers is paramount. Traditional security measures like passwords are often vulnerable to breaches. This paper addresses the need for robust, cost-effective, and unobtrusive user authentication and identification methods. Keystroke dynamics, which analyzes the unique rhythm and timing of an individual's typing, is presented as a promising behavioral biometric that can augment existing security systems without requiring additional hardware.

Methods

The paper provides a comprehensive overview of keystroke dynamics research, categorizing methodologies into several key areas:

  • Psychological Basis: It delves into the psychological and physiological aspects of typing, explaining how the "ballistic" nature of motor movements makes keystroke patterns difficult to impersonate. Models of typing stages (input, parsing, translation, execution) are discussed.
  • Data Acquisition: Methods for collecting keystroke data are detailed, distinguishing between:
  • Static/Structured Text Entry: Analyzing typing of specific phrases or passwords (e.g., during login).
  • Dynamic/Free Text Entry: Continuously monitoring typing behavior during general computer use. The influence of environmental factors (keyboard type, room conditions) on data quality is also considered.

  • Feature Extraction: Key features extracted from raw keystroke data are described, including:

  • Latencies: Press-to-press (PP), release-to-release (RR), and release-to-press (RP) times.
  • N-graphs: Analyzing sequences of keystrokes (digraphs, trigraphs).
  • Hold/Dwell Time: The duration a key is pressed.
  • Flight Time: The time between key release and the next key press.
  • Key Pressure: Measured by specialized keyboards.
  • Classification Algorithms: The paper categorizes algorithms used for authentication and identification:
  • Statistical Algorithms: Based on mean, standard deviation, and distance measures.
  • Neural Networks: Including backpropagation, Hopfield networks, and others.
  • Pattern Recognition and Learning-Based Algorithms: Such as Support Vector Machines (SVM), Naive Bayes, and Decision Trees.
  • Search Heuristics and Combination Algorithms: Genetic algorithms, ant colony optimization, and hybrid approaches.
  • Error Metrics: Standard metrics like False Acceptance Rate (FAR) and False Rejection Rate (FRR) are explained, along with the Equal Error Rate (EER).

Datasets

The paper lists several publicly available keystroke dynamics databases used in research. Notable examples include:

  • Jugurta and Freire [84]: Offers both static and dynamic text data.
  • Giot et al. [58]: One of the largest publicly available databases.
  • Allen [5]: Unique in including keystroke pressure information.

The authors note that while these databases exist, they are often small compared to real-world scenarios, and the lack of standardized databases hinders direct comparison of research.

Key Insights and Ideas

  • Psychological Foundation: Keystroke dynamics isn't just about timing; it's rooted in the complex motor control and cognitive processes of typing, making it inherently difficult to fake.
  • Cost-Effectiveness: As a behavioral biometric, it leverages existing hardware (standard keyboards), making it significantly cheaper than other biometrics requiring specialized sensors.
  • Non-Intrusiveness: It can be applied passively and covertly, augmenting security without user awareness or explicit action.
  • Feature Importance: While many features can be extracted, research suggests that hold times and inter-key times are particularly important, and the impact of different features and their combinations warrants further investigation.
  • The Need for Standardization: A significant barrier to progress is the lack of standardized databases, feature nomenclature, and evaluation protocols, making it hard to compare results across studies.

Key Results

The paper doesn't present new experimental results but rather synthesizes findings from numerous studies. Key qualitative and quantitative observations include:

  • Identification Rates: Various studies report high identification rates, often exceeding 90% with different algorithms and datasets. For example, Obaidat and Macchiarolo [113] achieved 97.8% identification using a hybrid neural network.
  • Error Rates: Studies report varying error rates. Joyce and Gupta [83] achieved a FAR of 0.25% and FRR of 16.36% using absolute distances.
  • Algorithm Performance: Different algorithms show varying strengths. SVMs are highlighted as a promising approach for both authentication and identification.
  • Language Impact: The paper notes that language (e.g., English vs. non-English) can significantly affect algorithm performance due to keyboard layouts and character frequencies.
  • Password Characteristics: Passwords with a mix of alphanumeric characters and longer passwords generally lead to better identification accuracy.

Limitations and Future Work

The authors identify several limitations and suggest future research directions:

  • Database Size and Standardization: The need for larger, more representative databases and standardized formats for data interchange and benchmarking is repeatedly emphasized.
  • Feature Engineering: Further research into the impact of specific features and their combinations is needed to improve accuracy. Developing a common nomenclature for features would also be beneficial.
  • Adaptive Templates: User typing behavior can change over time due to age, health, or practice. Developing adaptive templates is crucial to reduce false rejections.
  • Longitudinal Studies: Current research often uses short-term data. Longitudinal studies are needed to understand how typing behavior evolves and to address issues related to template variability.
  • Mobile and Virtual Keyboards: While the paper focuses on physical keyboards, the growing popularity of mobile devices necessitates research into keystroke dynamics on virtual keyboards.
  • Language Independence: Developing algorithms that perform uniformly across different languages is an important area for future work.
  • Privacy Concerns: Dynamic monitoring of keystrokes can raise privacy issues, and methods to mitigate these concerns (e.g., data anonymization, focusing on specific patterns) are important.
  • Synthetic Attacks: While some work exists, further investigation into synthetic or bot attacks and developing robust defenses is needed.

Summary updated

Abstract

Dependence on computers to store and process sensitive information has made it necessary to secure them from intruders. A behavioral biometric such as keystroke dynamics which makes use of the typing cadence of an individual can be used to strengthen existing security techniques effectively and cheaply. Due to the ballistic (semi-autonomous) nature of the typing behavior it is difficult to impersonate, making it useful as abiometric. Therefore in this paper, we provide a basic background of the psychological basis behind the use of keystroke dynamics. We also discuss the data acquisition methods, approaches and the performance of the methods used by researchers on standard computer keyboards. In this survey, we find that the use and acceptance of this biometric could be increased by development of standardized databases, assignment of nomenclature for features, development of common data interchange formats, establishment of protocols for evaluating methods, and resolution of privacy issues.
BibTeX
@article{banerjee2012biometric,
  author = {Salil Banerjee and Damon L. Woodard},
  title = {{Biometric Authentication and Identification Using Keystroke Dynamics: A Survey}},
  journal = {Journal of Pattern Recognition Research},
  year = {2012},
  doi = {10.13176/11.427},
}
Figure