Abstract
IFTT-PIN is a self-calibrating version of the PIN-entry method introduced in Roth et al. (2004). In the original method, digits are split into two sets and assigned a color respectively. To communicate their digit, users press the button with the same color that is assigned to their digit, which can thus be identified by elimination after a few iterations. IFTT-PIN uses the same principle but does not pre-assign colors to each button. Instead, users are free to choose which button to use for each color. The button-to-color mapping only exists in the user's mind and is never directly communicated to the interface. In other words, IFTT-PIN infers both the user's PIN and their preferred button-to-color mapping at the same time, a process called self-calibration. We present online interactive demonstrations and introduce the key concepts making self-calibration possible, proposing it as a novel approach to protect users against shoulder surfing attacks.
Introduction: The Challenge of Shoulder Surfing
Traditional PIN-entry systems are vulnerable to "shoulder surfing," where an observer watches a user's physical actions to deduce their secret code. This vulnerability stems from a fixed, visible mapping: pressing the button labeled "1" always inputs the digit 1. While "cognitive trapdoor" methods have attempted to obscure this by using intermediate symbols like colors, they typically rely on pre-defined or visible mappings between buttons and those symbols.
IFTT-PIN addresses this by introducing a "chicken-and-egg" solution to interface design. To understand user intent, a machine needs a mapping; to learn a mapping, it needs to know the intent. IFTT-PIN aims to solve both simultaneously through a paradigm borrowed from brain-computer interfaces: self-calibration.
The IFTT-PIN Interface and Interaction
The system presents a user interface consisting of a 4-digit PIN display, a grid of digits 0-9, and two input buttons. The interaction follows a specific sequence:
- Mental Mapping: The user assigns colors (e.g., Yellow and Grey) to the available input buttons in their mind. This mapping is never explicitly shared with the machine.
- The Challenge: The interface displays the digits 0-9, randomly assigning each a color (Yellow or Grey). The system then prompts: "What color is your digit?"
- The Action: The user presses the button that corresponds to the color of their target digit, according to their private mental map.
Inference via the Consistency Principle
The core of IFTT-PIN is its ability to infer intent without a pre-defined key. The system maintains ten hypotheses—one for each possible digit. It operates on the Consistency Principle: a user is assumed to be consistent in their mental mapping (i.e., one physical button always represents one specific color for that user).
As the user provides inputs across multiple rounds, the system discards hypotheses that lead to contradictions. For example, if a specific digit hypothesis requires a button to have meant "Yellow" in the first round but "Grey" in the second, that hypothesis is eliminated. To optimize this process, the algorithm uses active planning, choosing color assignments for digits that maximize information gain in each round.
Evaluation and Real-World Testing
The authors evaluated IFTT-PIN through several practical lenses:
- The Vault Challenge: A transparent vault secured by IFTT-PIN was placed in a university common area. Despite having access to video recordings of the entry process, "hackers" required hours to crack the PIN, while authorized users could gain entry in seconds.
- Public Demonstrations: At public art events, hundreds of participants, including children, successfully used the system. Observations suggested that children often grasped the self-calibration concept faster than adults, indicating that defining an interface on-the-fly is an intuitive skill.
Discussion and Future Work
While IFTT-PIN offers robust security against shoulder surfing, it currently assumes a "perfect user." A single erroneous button press can break the inference logic, suggesting that future iterations must incorporate probabilistic models of human error. Additionally, providing user feedback on the inference progress could improve the user experience.
Beyond PIN entry, the self-calibration paradigm suggests a future where users actively and subjectively define how they interact with machines, potentially opening new doors for personalized interactive devices.
Summary updated