Techniques for Secure and Reliable Computational Outsourcing

Page Content

Research Areas: Assured Identity and Privacy

Principal Investigator: Mike Atallah

We are designing techniques for the secure outsourcing of computations, where a computationally limited client uses remote (and computationally more powerful) servers for carrying out computationally intensive tasks in a confidentiality-preserving manner and cheating-resilient manner. This is done without revealing to the remote servers the identity of the computational power being used, either its input data or the outcome of the data computation, and while detecting with high probability incorrect answers from the servers. Moreover, the client needs to reliably detect cheating by a remote server who returns wrong answers.

Such techniques are useful in a wide range of applications that require lightweight and computationally slow or battery-limited entities (e.g., sensors, roving small vehicles) to carry out computationally intensive tasks.  It enables them to use the computational power of remote entities that are neither trusted nor secure.  The proposed techniques are also applicable when solving massive computational tasks on remote super-computers that cannot be considered fully secure (e.g., at major universities). This kind of technology is useful even when the remote servers are trustworthy, as a means of compartmentalization and “defense in depth": The damage resulting from the compromise of a server becomes confined to that server's data, and does not extend to the clients that use it (their data remains secure even if the server suffers a break-in, a malware / spyware infection, insider misbehavior, etc).

Personnel

Other PIs: Marina V. Blanton

Keywords: cryptography, Privacy, secure collaboration, supply chain, zero knowledge