Dokument

Summary:
While security research has made significant progress in the development of theoretically secure methods, software and algorithms, software still comes with many possible exploits, many of those using the human factor. The human factor is often called ``the weakest link'' in software security. To solve this, human factors research in security and privacy focus on the users of technology and consider their security needs. The research then asks how technology can serve users while minimizing risks and empowering them to retain control over their own data. However, these concepts have to be implemented by developers whose security errors may proliferate to all of their software's users. For example, software that stores data in an insecure way, does not secure network traffic correctly, or otherwise fails to adhere to secure programming best practices puts all of the software's users at risk. It is therefore critical that software developers implement security correctly. However, in addition to security rarely being a primary concern while producing software, developers may also not have extensive awareness, knowledge, training or experience in secure development. A lack of focus on usability in libraries, documentation, and tools that they have to use for security-critical components may exacerbate the problem by blowing up the investment of time and effort needed to "get security right". This dissertation's focus is how to support developers throughout the process of implementing software securely. This research aims to understand developers' use of resources, their mindsets as they develop, and how their background impacts code security outcomes. Qualitative, quantitative and mixed methods were employed online and in the laboratory, and large scale datasets were analyzed to conduct this research. This research found that the information sources developers use can contribute to code (in)security: copying and pasting code from online forums leads to achieving functional code quickly compared to using official documentation resources, but may introduce vulnerable code. We also compared the usability of cryptographic APIs, finding that poor usability, unsafe (possibly obsolete) defaults and unhelpful documentation also lead to insecure code. On the flip side, well-thought out documentation and abstraction levels can help improve an API's usability and may contribute to secure API usage. We found that developer experience can contribute to better security outcomes, and that studying students in lieu of professional developers can produce meaningful insights into developers' experiences with secure programming. We found that there is a multitude of online secure development advice, but that these advice sources are incomplete and may be insufficient for developers to retrieve help, which may cause them to choose un-vetted and potentially insecure resources. This dissertation supports that (a) secure development is subject to human factor challenges and (b) security can be improved by addressing these challenges and supporting developers. The work presented in this dissertation has been seminal in establishing human factors in secure development research within the security and privacy community and has advanced the dialogue about the rigorous use of empirical methods in security and privacy research. In these research projects, we repeatedly found that usability issues of security and privacy mechanisms, development practices, and operation routines are what leads to the majority of security and privacy failures that affect millions of end users.

Das Dokument ist im Internet frei zugänglich - Hinweise zu den Nutzungsrechten