Towards an architecture for end-to-end-encrypted file synchronization systems

Users often utilize Dropbox and similar services to store their data in a cloud. They protect their data through encryption services offered by the cloud provider. But how reasonable is such a protection? The cloud provider is usually able to (at least theoretically) read the encrypted data, since he is the one holding the encryption keys. And even if you trust a cloud provider, what happens if the cloud provider is acquired by another company? Do you also trust the acquiring company? Global surveillance has become a daily issue, outlined by disclosures of files from the United States National Security Agency (NSA). To keep one's data protected from unauthorized access, a user optimally needs to trust as few other parties as possible. We should aim for a future, in which users are able to protect their data without having to trust the cloud provider who stores their data. This can be achieved by using strong, auditable client-side encryption. This paper presents a first step towards this goal. Starting from a basic requirement-the principle of least privilege-requirements are derived that again result in an architecture to build end-to-end-encrypted file synchronization systems. The resulting architecture's practical applicability is shown by a concrete implementation.