USC Computer Science Professor Wins Most Influential Paper Award
USC Professor of Computer Science Nenad Medvidović was recognized with the honor of the Most Influential Paper Award at the 15th International Symposium on Software Engineering for Adaptive and Self-Managing Systems (SEAMS 2020). The award recognizes his 2007 paper, “An Architectural Style for Solving Computationally Intensive Problems on Large Networks,” co-authored with Yuriy Brun.
The award recognizes work published at least ten years prior that has had substantial impact in industry or academia. The award is decided by a committee specially appointed by the organizers of SEAMS, which is the annual symposium for researchers working on software engineering of self-adaptive and self-managing systems.
The 2007 paper discussed privacy around cloud-based systems, when this type of computing was not yet widely adopted. Medvidović said he was happy to hear this unexpected news and was grateful that the SEAMS committee recognized the paper, which was the dissertation work of Brun (Medvidović’s USC doctoral student at the time). Brun is now an Associate Professor at the College of Information and Computer Sciences (CICS) of University of Massachusetts Amherst. At USC, Medvidović has been Director of the Center for Systems and Software Engineering and Associate Chair of the Computer Science Department.
Although security in cloud computing and networks is in everyday parlance, it is still difficult to ensure certain aspects of security (for example, the confidentiality of data) on a public network. Recognizing the 2007 paper now shows the committee deemed it to have been ahead of its time.
“It was very difficult to find a ‘home’ for this work back then,” said Medvidović. “Several of the initial attempts at publishing it were rejected. The idea was different enough from research being pursued in my field at the time that we even got a ‘desk rejection’ from a major software engineering journal for submitting work that was outside the journal’s scope. It is wonderful to now have the impact of this work recognized by my community.”
The 2007 paper introduced sTile — a technology developed at USC in 2007 by Brun and Medvidović — which was a novel method for distributing computation onto a public network of computers in a way that makes two critical guarantees. First, the data used in the computation will remain confidential; second, the sTile-encoded software system computing on the data will be tolerant to faults such as a computer crashing.
According to Medvidović, the novel aspect of sTile is that it built a distributed software system based on an underlying mathematical model of how crystals grow. sTile’s individual components are extremely small, which means that very large numbers of them are required to solve a problem. In turn, this means that in order to figure out what data a sTile system’s components are operating on, an adversary would have to find a way to “peek” inside such a large number of the components that it is practically infeasible for them to do so.
Medvidović said this work has had an impact on researchers developing numerous systems, including distributed authentication protocols, permissioned blockchains, and runtime self-adaptive systems. “This idea has clear implications on ‘volunteer computing’ systems, as well as the cloud and smart mobile devices, which have become ubiquitous since the paper’s original publication,” he said.
“As we remain ‘plugged in’ and different facets of our private data are maintained in countless locations, we are constantly at risk of having that data exposed and stolen by yet another breach,” he continued. “The solutions that are often offered to the public in such situations after the fact are simply unsatisfactory. sTile showed how realistic systems can be built in a way that can provide much higher practical data confidentiality guarantees.”
After the 2007 paper, Medvidović and Brun continued their research in cloud and network security, publishing additional papers: “Keeping Data Private while Computing in the Cloud” (published in the IEEE International Conference on Cloud Computing in 2012) and “Entrusting Private Computation and Data to Untrusted Networks” (published in IEEE Transactions on Dependable and Secure Computing in 2013). These papers proved mathematical limits on the data confidentiality guarantees, including what fraction of the cloud a malicious adversary must compromise to gain access to the data, and empirically measured the efficiency cost of providing such data confidentiality, according to Medvidović.
–posted May 14, 2020 by USC Stevens Center for Innovation staff