TL;DR
This paper introduces "cognitive trapdoor games" as a novel method for secure PIN entry that is resilient to shoulder surfing, even when the attacker records the entire interaction. The approach leverages human cognitive limitations to create a system that is easy for the legitimate user to operate but difficult for an observer to decipher, offering a practical balance between security and usability.
What Problem Does This Paper Address?
Shoulder surfing, where an attacker observes a user entering their PIN, is a persistent threat to the security of financial transactions and access control systems. While adding "something you have" (like a card) to "something you know" (like a PIN) is common, attackers can still obtain both. Existing methods to combat shoulder surfing are often insufficient, especially with the advent of miniature cameras that can record PIN entry. This paper addresses the need for a PIN entry method that is inherently resistant to observation, even by sophisticated attackers with recording devices, while remaining usable for the average person.
Methods
The core innovation is the "cognitive trapdoor game." This is an interactive challenge-response protocol where the verifier (e.g., an ATM) presents information in a way that is easy for the prover (the user) to interpret and respond to, but difficult for an observer to deduce the underlying PIN.
The primary mechanism involves presenting PIN digits through a series of visual cues (color patterns on keys) that are partitioned into two sets. The user indicates which set their digit belongs to (e.g., by pressing a "white" or "black" key). The verifier uses these responses to narrow down the possible digits.
Two main variants are proposed:
- Immediate Response: The user responds immediately after each visual cue.
- Delayed Response: The user sees a sequence of cues and then responds with the sequence of their choices.
To enhance security against recording devices, a "probabilistic cognitive trapdoor game" is introduced. This involves providing slightly ambiguous information, resulting in a "shadow set" of possible PINs. The verifier can efficiently check candidates in this set, while an attacker without the master secret can only guess randomly from the shadow set, significantly reducing their success probability.
A tactile variant using Braille-like displays is also proposed for visually impaired users, where output is hidden by the user's fingertips.
Datasets
Not discussed in the paper. The paper focuses on a theoretical design and experimental studies with human subjects.
Key Insights and Ideas
- Leveraging Cognitive Limitations: The most significant insight is that human short-term memory (STM) limitations can be actively exploited to create security. By requiring an observer to memorize a complex sequence of information, their cognitive load can be made much higher than the user's.
- "Cognitive Trapdoor" Concept: The idea of a game that is easy to win if you know the secret (the PIN) but hard to win if you don't, even with full observation, is a powerful security paradigm.
- Shadow Sets for Camera Resilience: The probabilistic modification, creating a "shadow set" of potential PINs, is a clever way to provide resilience against camera recordings by introducing uncertainty. This turns a deterministic attack into a probabilistic one with a manageable success rate.
- Usability vs. Security Trade-off: The paper acknowledges that increased security often comes with a usability cost (longer entry times, higher attention required). The goal is to find a practical balance, and the "immediate choice" method appears to achieve this.
Key Results
- Security Against Shoulder Surfing: User studies showed that subjects with no specific training could guess only one or two digits of PINs entered using the new methods, compared to correctly guessing all digits in the regular method. This suggests a substantial increase in resistance to shoulder surfing.
- Usability of Immediate Choice (IOC): The "immediate choice" method (IOC) was rated comparably to the regular PIN entry method in terms of usability, despite requiring more effort. This suggests it's a practical option.
- Usability of Delayed Choice (DOC): The "delayed choice" method (DOC) was rated significantly less usable and more demanding, making it less suitable for practical application.
- Learning Effect: Subjects acquired the skills to operate the new methods within three repetitions, indicating a manageable learning curve.
- Error Rates: While initial error rates were higher for the new methods, they became comparable to the regular method after learning. Fatigue also impacted error rates in the IOC and DOC conditions.
Limitations and Future Work
- Observer Training: The study's results on observer guessing are based on untrained subjects. It's acknowledged that determined adversaries might improve their skills with training, and further investigation into the degree of training improvement is needed.
- DOC Method Demands: The delayed choice method was found to be too demanding for practical application.
- Tactile Variant Security: While the tactile design offers privacy, the paper notes that sophisticated observers might develop new ways to capture pin states (e.g., via microphones).
- Future Work: The authors plan to conduct larger-scale usability studies, ideally in field tests, and to compare their designs with other proposed methods. They also express interest in studying the usability of Matsumoto's designs.
Summary updated