Abstract:
Crowdsourcing is the practice of getting information or input for a task or project from a number
of people by floating out the task at hand to a pool of people who are usually not full-time
employees. Use of Crowdsourcing to get work done is on the rise due to the benefits that it offers. Trends and surveys show that the conventional workforce is moving towards gig-economy, and by 2020, 43% of the US workforce is expected to be on-demand workforce [9]. In our work, we focus on higher level software development/engineering tasks and how they could potentially lead to Confidentiality Loss in the process of giving the task and getting it done. We look at different stages and components of the crowdsourcing cycle to examine potential information leak sources. We conducted a survey to study how people perceive this problem and what is their level of understanding when it comes to sharing information online. We also analyzed a dataset of tasks posted online previously to get insight if the problem persists in the real world or not. Some conversations between the task posters and the workers were studied along with a dataset of reviews for workers to get a deeper insight into the problem and its detection. Based on the analysis, we propose NLP based techniques to detect such potential leaks and nudge the task poster before the information is dissipated. Having such an additional layer of scrutiny at the level of companies before a task is given for crowdsourcing out will keep in check any loss of confidential information.
