D3.3 - Complete Design and Evaluation of Verifiability mechanisms

This deliverable overviews the complete specification and evaluation of nine different TREDISEC primitives that enable a cloud customer to remotely verify the correctness of cloud operations including storage, processing. Thanks to these primitives, cloud users acquire more confidence to outsource their storage and processing operations to the cloud. On the other hand, thanks to their compatibility with cloud functional operations such as file deduplication or data replication, cloud servers can maintain the cost efficiency of current infrastructures and offer these new security guarantees at the same time. The proposed primitives are the following:

  • Two verifiable storage primitives, namely ML-PoR and SPORT that enable a cloud server to guarantee the correct storage of customers’ data while being able to perform file deduplication to achieve storage savings. ML-PoR basically is a generic solution that extends traditional PoR (proof of retrievability) schemes to make them compatible with file-level deduplication. It leverages a key server in order for all cloud users generating the same PoR parameters to encode the to-be-outsourced PoR encoded data. This way, cloud servers will still be able to perform deduplication. On the other hand, SPORT is a new PoR that transparently supports multi-tenancy with deduplication by enabling different cloud users to share the same PoR tags in order to verify the integrity of the same file. SPORT introduces a stronger adversary model.
  • One primitive, Mirror that provides the cloud customers the guarantee that the cloud correctly keeps multiple replicas of their data in addition to the retrievability guarantee for the original data and its replicas. Unlike previous schemes, Mirror outsources the replica generation function to the cloud server and makes use of cryptographic puzzles to prevent a malicious cloud from meeting the replication guarantee while not actually storing these replicas.
  • A proof of ownership primitive that allows a cloud server to verify that a user actually owns a file without the need for transferring it over the network. This immediately enables a secure client-side deduplication and thus achieves bandwidth savings for the storage of redundant data. The proposed primitive, OOPRF, makes use of an oblivious pseudo-random function in order for the user not to reveal any information about the file but still prove its ownership. An open-source implementation of two existing PoW solutions has been provided.
  • Three verifiable computation primitives that help customers efficiently verify the correctness of some outsourced operations, namely: polynomial evaluation, matrix multiplication, and biometric matching. While the first two primitives make use of some simple algebraic properties of the original operations, the verifiable biometric matching operation optimizes an existing verifiable computation protocol to be compatible with the inner product operation.
  • A verifiable document redacting primitive that empowers cloud users to easily remove some part of their already signed document without having an impact on the validity of the signature. Thanks to this new primitive, users will not disclose private information of the document that does not need to be shared with the destined party.
  • A system integrity verification primitive, TRAVIS, which makes use of a Trusted Platform Module technology to achieve remote attestation of virtual cloud systems.