Security in Big Data and Cloud Computing - Large Scale and Big Data: Processing and Management - page 587

19.3.6 v eriFying o utsourCeD C omPutation
Research Question 6 : How can we (efficiently) verify the accuracy of outsourced
computation [12]?
Users of clouds often outsource large and complex computations to a cloud. However,
doing so exposes the cloud user to a new issue: what guarantees that the cloud provider
will accurately execute the program and provide a correct value as a result?
Users have several options: first, clients can redo the computation. However, for
costly computations, the clients often would lack the capability to do so (and which
is precisely why they outsource the computations). Next, users can do redundant
computation by sending the computation to multiple clouds and later take majority
voting or other consensus schemes to determine correctness. For large computations,
this also may not be practical.
A slightly different approach was developed by Du et al. [12], who used run-
time attestation. In their scheme, the same data in a DataFlow programming system
is routed via multiple paths, and results are compared in each pair of cloud nodes
performing the same computation on the same data. Based on agreements between
the results from the two nodes, an attestation graph is created. From that graph, the
maximal clique of nodes is computed. If that clique has more than half of the nodes,
then it is assumed to be trustworthy, and results coming from the nodes belonging to
the maximal clique are also considered trustworthy.
19.3.7 v eriFying C aPability
Research Question 7 : How can a client remotely verify the capability and resource
capacity of a cloud provider [8]?
Verifying capability of a service provider is difficult, and even more so when the
service provider does not allow inspection of its infrastructure. Therefore, verify-
ing the capability of a cloud to store data or run applications is a complex problem.
Researchers have only recently developed techniques for verifying the storage capa-
bility of cloud service providers. Bowers et al. [8] developed a strategy to determine
whether a cloud is indeed storing multiple replicas of a file, and therefore is capable
of recovering from crashes. In this approach, file read latencies are used to determine
the presence of multiple physical replicas. Similar research has also looked into veri-
fying the capability of storing files in geographically separate data centers [5].
19.3.8 C louD F orensiCs
Research Question 8 : How can we augment cloud infrastructures to allow forensic
investigations [23]?
Cloud forensics is the application of computer forensic principles and procedures
in a cloud computing environment. Traditional digital forensics strategies and prac-
tices often fail when the suspect uses a cloud. As an example, a suspect using a tradi-
tional file storage to store his incriminating documents would be easy to convict and
prosecute—the law enforcement investigators can make an image of his hard drives
and run forensic analysis tools there.