Naiakshina, Alena: Don't Blame Developers! Examining a Password-Storage Study Conducted with Students, Freelancers, and Company Developers. - Bonn, 2020. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5-60596
@phdthesis{handle:20.500.11811/8842,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5-60596,
author = {{Alena Naiakshina}},
title = {Don't Blame Developers! Examining a Password-Storage Study Conducted with Students, Freelancers, and Company Developers},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2020,
month = dec,

note = {Authentication systems are a major concern for the usable security and privacy community. Twenty years ago, the seminal work “Users Are Not The Enemy” [1] by Adams and Sasse initiated a user-centred approach to research. This work was followed by extensive research on end users’ security behavior around authentication systems, which resulted in many suggestions for improvement. Many of these proposals were passed on to software developers, who were considered experts who should know better and were expected to address the issues. However, the related work “Developers Are Not The Enemy” [2] by Green and Smith, citing Adams and Sasse [1], points out that, similar to end users, developers are usually not security experts. Thus, they also struggle with usability and security issues. This is highlighted by the high number of reported security breaches that have compromised millions of end-user passwords. In fact, much of the work invested in usable authentication systems might be in vain if software developers fail to securely store user passwords in databases.
Motivated by the recent security breaches, this thesis aims to provide deeper insights into developers’ security behavior and to clarify why software developers so often fail to store user passwords securely. Therefore, this thesis describes a Java-based password-storage study that was conducted with different samples of developers: computer science (CS) students, freelancers, and professional developers employed by different companies. Participants were instructed to complete the registration functionalities for a social network platform. In order to investigate whether software developers think about security without prompting, half the participants were told the study was about application programming interface (API) usability (non-prompted for security), while the other half were specifically instructed to securely store user passwords (prompted for security). The study also investigated whether an API’s level of password storage security support affects developers’ security behavior. Thus, half the participants used a framework offering opt-in support for secure password storage (Spring), while the other half were provided a programming frame with JavaServer Faces (JSF), which required them to implement password storage security without support.
Initially, a qualitative and a quantitative study were conducted with 20/40 CS students from the University of Bonn in a laboratory setting. The most important finding of this study was that all the students who were not prompted for security submitted solutions in which user passwords were stored in plaintext in the database. Furthermore, a number of the prompted participants who considered password storage security still chose weak security practices. However, some participants claimed that they would have stored user passwords securely in the database if they were solving the task for a real company. In order to test whether these findings were a study artifact, a follow-up study was conducted with 43 freelancers recruited via Freelancer.com. A pilot study with the freelancers suggested that a university context might lead them to believe that university students were hiring them to do their homework. Therefore, this time, the freelancers were not informed that the study was conducted by a research team. Instead, they were told that they were working for a start-up that had lost recently a developer from their team. In this study, freelancers behaved similarly to students with regard to user password storage. Freelancers also had often misconceptions about secure password storage and chose weak practices. Finally, 36 company developers were invited to take part in a password-storage study. They were recruited through their companies and also via the German business social platform Xing. Company developers submitted significantly more secure solutions than students, and they also chose significantly better password storage parameters than students or freelancers. Thus, in absolute terms, they performed better than students and freelancers. However, in relative terms, the results were similar: Security prompting and framework had a significant effect on password storage security for all samples. When prompted, more participants submitted secure solutions, and participants made better parameter choices for password storage security when they used Spring instead of JSF (freelancers used only JSF and thus were not tested for the variable framework).
Additionally, the four studies offered insights for the ecological validity of security studies with developers. The student studies provided an early indication that qualitative research might reveal essential insights without the need to conduct quantitative studies for specific use cases. Furthermore, if the usable security and privacy community is more interested in how security systems can be improved than in which developer group performs best, it might be valid to conduct studies with students rather than professionals. What is more, freelancers tended to behave similarly to students with regard to secure password storage, although they were not aware of the purpose of the study. This suggests that participants tended to ignore the security aspects of software even when working on a web application intended for the use in the real world.
Based on these findings, this thesis has made some recommendations for improving password storage security and regarding the methodological implications for security studies with developers.},

url = {https://hdl.handle.net/20.500.11811/8842}
}

The following license files are associated with this item:

InCopyright