Resolving a common vulnerability in secret sharing scheme–based data outsourcing schemes

Abstract

Summary Every day, the amount of generated data from different resources are increasing significantly and posing new serious challenges in terms of data storage and maintaining. As a solution, outsourcing data to a public cloud server with high storage and processing capacity sounds reasonable in comparison with storing in local data storage. Therefore, the database as a service (DaaS) paradigm has gained much popularity over the last decade since the introduction of cloud services. Storing confidential data into a third party arises some concerns regarding security and privacy of the data. One popular approach to address these problems in data outsourcing is utilizing secret sharing and distributing sensitive data among several databases. In this paper, we introduce a flaw that exhibits a deficiency relating to current schemes. This flaw indicates an intrinsic vulnerability to such methods and necessitates a reconsideration of these schemes. Moreover, a new way will be developed to resolve this problem in the previous schemes, which can be employed in all current schemes to mitigate the mentioned concern. Finally, a comprehensive implementation demonstrates that the proposed method is scalable to meet increasing demands in large databases.

Secret Sharing-Based IoT Text Data Outsourcing: A Secure and Efficient Scheme

“…As observed in previous works [22]- [24], potentially there are two types of attacks against secret sharing-based outsourcing schemes. The first type of attacks, called inference attacks, will be explained in detail in Section III-C1.…”

“…Gcd-based attacks were first discovered in [24] whereas compromised cloud servers are able to retrieve the distribution vector X, by recovering each of the individual reconstruction element x i based on the shares they received. It has been claimed in [24] that, subtracting any two shares that are associated with the same secret will result in a multiple of the reconstruction element x i . To see this, readers can refer to Equation ( 5) below, where (c vn * x i + v) is the share generated from the secret v for an n-th time and (c vm * x i + v) is the share for an m-th time.…”

“…One can easily see that the right hand side of Equation ( 5) is essentially a multiple of x i . The value of x i has been successfully retrieved in [24] by observing that x i turns out to be the greatest common divisor (gcd) appearing most frequently in a set of data called Med.…”

“…In other words, we only need a constant size of storage overhead while the secure partitioning approach requires an extra storage which grows linearly in the domain size of secret values (readers can recall that D v is the domain of unique values originated from data owners.). * When compared with the fake record approach [24] which tackled gcd-based attacks by introducing an additional 10% overhead, our scheme is more efficient thanks to the constant size storage overhead achieved.…”

“…The second type of attacks, termed as gcd-based attacks, will be described in Section III-C2. A gcd-based attack has been discovered by Ghasemi in the literature [24] where a countermeasure was also proposed by inserting 10% extra records as fake records in order to protect the real records. We refer to this approach as fake record approach in the later sections.…”