Data protection using a combination of data encryption, fragmentation, and dispersal is revisited in the context of distributed storage. First, the family of all-or-nothing transforms is extended with three new proposals protecting outsourced data against the exposure of the encryption key. Second, two ways of accelerating data fragmentation are introduced. They achieve better performance than the most common symmetric-key encryption algorithm. Finally, an additively homomorphic scheme for distributed storage inside Unattended Wireless Sensor Networks is introduced. It reduces by at least half the storage and transmission costs in comparison to the state-of-the art techniques.