Designing Safe Discussion Spaces by Anonymizing Communication Style
Abstract
Statement of Work:The grantee will perform the following tasks:Task 1: Anonymity Metric Design; The goal of this task is to design a new anonymity metric that applies to stylometric applications, and to design mechanisms that enforce this novel anonymity metric. Task 2: Measuring Anonymity; The goal of this task is to design methods for evaluating the amount of anonymity offered by a (possibly human-generated) process. This is easy for automatic processes, but non-trivial when humansare involved. The grantee will then perform crowdsourcing studies to determine the impact of different parameters on a crowdsourcing system that anonymizes communication. This is divided into three tasks (Tasks 3-5). Task 3: Length of text; The crowdsourcing system designed will break down text into chunks and give it to other participants toanonymize. The grantee will examine the impact of length of the text chunks on anonymization performance through pilot crowdsourcing studies. Task 4: Number of participants; The crowdsourcing system will determine the effect of the number of participants on anonymization performance through crowdsouring. Task 5: Design the reassembly process; The goal of this task is to do pilot crowdsourcing studies to figure out impact of the reassembly process. The grantee will also look at crowdsourced reassembly as a local optimization process and examine its convergence properties.Objective:In many situations, free speech is only possible under the guise of anonymity. For example, in a country like Egypt, where agents of the state may be monitoring the conversation, participants who engage in political speech may only be able to do so under anonymity. However, if writing samples from individuals are available, it is possible to determine the authorship of a piece of text from the writing style using modern stylometric methods, even when there are a large number of candidates for authorship. The objective of this project is to design algorithms and principles that will allow us to anonymize text according to writing style using crowdsourcing. These algorithms may then be used to design discussion spaces where individuals can communicate freely without fear of being recognized. Approach:The grantee will address the algorithmic aspects of the problem, and conduct small pilot studies in order to validate the proposed approaches. In particular, the grantee will (a) come up with a new principled framework for anonymity that is suitable for this problem and propose methods that enforce this kind of anonymity (b) come up with a way of evaluatingthe amount of anonymity achieved empirically and (c) analyze the anonymization process and carry out small scale crowd-sourcing studies to validate our analysis.Overall Merit and ONR Mission/Relevance:This project is directly related to ONR s mission of creating a safe and secure cyber environment, which enhances anonymity of activities in cyber-world, and ensures that eavesdroppers cannot effectively exploit cyberspace and impair its safe and secure use. Designing algorithms and principles that anonymize text by writing style will allow us to design anonymized communication fora where participants can safely speak their mind without fear of their identity being compromised by eavesdroppers and malicious adversaries.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 12, 2016
- Source ID
- N000141612616
Entities
People
- Kamalika Chaudhuri
Organizations
- Office of Naval Research
- United States Navy
- University of California, San Diego