Towards understanding ATM security: a field study of real world ATM use

Alexander De Luca, Marc Langheinrich, Heinrich HussmannSOUPS 2010 · DOI

TL;DR

This paper presents a field study of real-world ATM usage, revealing that contextual factors like distractions and social interactions significantly impact PIN security and usability. The findings suggest that future ATM authentication systems should be designed with built-in security, rather than relying on user vigilance, and must be compatible with social norms and distracting environments.

What Problem Does This Paper Address?

The increasing prevalence of ATM fraud has spurred the development of new authentication mechanisms beyond traditional PINs. While these new methods are often evaluated on speed, security, and memorability in controlled lab settings, their real-world effectiveness remains unclear. This paper addresses the gap in understanding how contextual factors in actual ATM usage influence security and performance, and how these real-world dynamics should inform the design and evaluation of future authentication systems. The authors highlight that lab studies may not fully capture the complexities of public ATM interactions, leading to a potential disconnect between theoretical security and practical usability.

Methods

The study employed a mixed-methods approach, combining field observations with follow-up interviews.

  • Field Observations: The primary method involved unobtrusive observation of ATM users in public spaces across six locations in Munich and Delft. A single researcher conducted observations over two months, ensuring consistency. A detailed checklist was used to record data, including location, time of day, interaction time, queue length, security measures, and contextual comments. Observations were discarded if the checklist could not be completed with 100% confidence or if the observer's presence might have influenced user behavior.
  • Follow-Up: Public Interviews: To gain deeper insights into user security considerations, the influence of company, and queuing behavior, semi-structured interviews were conducted with 25 participants. These interviews aimed to understand user perceptions and decision-making processes related to ATM security.
  • Follow-Up: In-Depth Time Measures: A smaller, secondary field study focused on precisely measuring the different phases of ATM interaction (preparation, PIN entry, interaction, and cleanup) using a custom program on a smartphone. This was done to better understand the overhead beyond the core transaction.

Datasets

The paper does not introduce new benchmark datasets. Instead, it relies on data collected from:

  • Primary Field Observation: Data from 360 ATM users observed across 44 sessions.
  • Follow-Up Interviews: Responses from 25 interviewed participants.
  • In-Depth Time Measures: Data from 24 ATM interaction sessions collected during the follow-up time study.

Key Insights and Ideas

  • Context is King: Real-world ATM use is heavily influenced by contextual factors like distractions (phone calls, shopping bags, company), physical hindrance, and social dynamics. These factors can override security best practices.
  • Authentication is a Minor Task: The actual authentication (PIN entry) is a very small fraction of the overall ATM interaction time. This suggests that users may not tolerate significantly slower authentication methods, even if they offer enhanced security.
  • Security Needs to Be Built-In: The majority of users do not actively employ security measures, or do so ineffectively. This indicates that future authentication systems should not rely on users to actively protect their secrets but should have security features "built-in."
  • Social Norms Matter: Users may forgo security measures to avoid appearing untrusting of companions. Authentication systems need to be socially compatible.
  • Memorability is a Niche Problem, But Critical When It Occurs: While few users forgot their PINs, those who did resorted to insecure methods like writing them down. This highlights that memorability, though not a widespread issue, can lead to severe security breaches when it arises.
  • Distractions are Pervasive: A significant portion of users (11%) were distracted during their ATM use, with a small percentage experiencing errors due to these distractions.

Key Results

  • Interaction Time: The average interaction session was 45.9 seconds (SD: 15.1s). Preparation and cleanup phases accounted for approximately 27% of the total blocked time at the ATM.
  • Distractions: 11% of users were distracted during their ATM use, and for 3% of these, distractions led to errors.
  • Input Errors: Only 6 out of 360 users failed to authenticate correctly on the first attempt. Interactions with a failed authentication attempt were significantly longer.
  • Queuing Behavior: Large queues were rare. User decisions to queue were influenced by urgency, queue length, availability of alternatives, and perceived safety.
  • Observable Security Measures: Only about 35% of users made observable efforts to secure their PIN entry, with hiding the PIN with a hand being the most common method.
  • Social Influence: 17% of users were in company, and 6% allowed their companions to watch their PIN entry. Notably, only one user who was watched by companions applied security measures.

Limitations and Future Work

  • Geographic Limitations: The study was conducted in two central European cities, limiting its generalizability to other cultural areas or less urban settings.
  • Hardware Attack Observation: The unobtrusive nature of the observations prevented in-depth findings on users checking ATM hardware for manipulations.
  • Interview Bias: Field interviews may have been subject to social desirability bias, with participants potentially overstating their security practices.
  • Future Work: The authors suggest extending observations to other public terminals (e.g., ticketing machines, supermarket checkouts) where different circumstances might lead to different usage patterns. They also encourage cross-cultural studies to understand how findings apply in diverse contexts. Further research into the cost of deploying new authentication systems and resistance to vandalism is also noted as important.

Summary updated

BibTeX
@inproceedings{luca2010towards,
  author = {Alexander De Luca and Marc Langheinrich and Heinrich Hussmann},
  title = {{Towards understanding ATM security: a field study of real world ATM use}},
  booktitle = {SOUPS},
  year = {2010},
  doi = {10.1145/1837110.1837131},
}
Figure