Breaking Assumptions: Distinguishing Between Seemingly Identical Items Using Cheap Sensors

Abstract

Intuitively, we understand that physical items are distinct --- we are able to

pick up and manipulate them. Yet, we often treat similar items as if they are

indistinguishable from one another. Undue reliance on the assumption that seemingly identical items can be treated as if they are identical can

lead to unintended, negative consequences. Conversely, the ability to

distinguish between seemingly identical objects often enables

new applications. This thesis examines three artifacts

that carry with them false assumptions about their indistinguishability: blank

sheets of paper, `anonymous' bubble forms, and loudspeakers.

Blank sheets of paper, e.g. those in a ream of copy paper, are often treated

as if they are identical to one another. We develop a new method to identify

sheets of paper by measuring their inherently unique surface texture using

commodity equipment. This permits a number of applications, e.g. counterfeit

currency detection, and has important implications for paper-based elections,

which are discussed in detail.

Next, we turn to optical-scan bubble-forms which often rely on the

assumption that they do not reveal the respondent's identity. We demonstrate

that individuals tend to mark bubbles in distinctive ways, unintentionally

conveying their identity. This has important implications for anonymous

surveys and the publication of completed ballots after an election. We

describe a number of mitigation techniques and procedural changes to limit the

risk of accidentally revealing identifying information.

The final artifact, loudspeakers, are increasingly ubiquitous devices designed

to accurately reproduce an audio signal. It is commonly

assumed that multiple instances of `identical'

loudspeakers will generate the same output given identical inputs. This

assumption is false. We demonstrate that individual loudspeakers,

even those of the same make and model, induce unique distortions on the

generated sound, identifying the individual loudspeaker. We develop methods to

identify loudspeakers, enabling a new method of device authentication.

By examining the common underlying assumptions of each artifact, we

develop a common methodology used in identifying distinguishing features. This

general framework is successfully applied to each artifact, suggesting that

other seemingly identical objects may become distinguishable in the future.