Search

A query is said to be secure against inference attacks by a user if there exists no database instance for which the user can infer the result of the query, using only authorized queries to the user. In this paper, first, the security problem against inference attacks on object-oriented databases is formalized. The definition of inference attacks is based on equational logic. Secondly, the security problem is shown to be undecidable, and a decidable sufficient condition for a given query to be secure under a given schema is proposed. The idea of the sufficient condition is to over-estimate inference attacks using over-estimated results of static type inference. The third contribution is to propose subclasses of schemas and queries for which the security problem becomes decidable. Lastly, the decidability of the security problem is shown to be incomparable with the static type inferability, although the tightness of the over-estimation of the inference attacks is affected in a large degree by that of the static type inference.

Detecting security flaws is important in order to keep the database secure. A security flaw in object-oriented databases means that a user can infer the result of an unpermitted method only from permitted methods. Although a database management system enforces access control by an authorization, security flaws can occur under the authorization. The main aim of this paper is to discuss the detection problem of security flaws for database schemas. This problem is to decide whether or not, when a schema S, an authorization A, and a term to be verified are given, there exists a database instance of S such that a security flaw on exists with respect to A. This paper shows that the problem is undecidable and proposes a decidable sufficient condition for a given schema to have no security flaw. Furthermore, this paper shows that the sufficient condition is also a necessary one if the given schema is monadic (i.e., every method is unary), and evaluates the time complexity to decide the condition.

application/pdf, A query is said to be secure against inference attacks by a user if there exists no database instance for which the user can infer the result of the query, using only authorized queries to the user. In this paper, first, the security problem against inference attacks on object-oriented databases is formalized. The definition of inference attacks is based on equational logic. Secondly, the security problem is shown to be undecidable, and a decidable sufficient condition for a given query to be secure under a given schema is proposed. The idea of the sufficient condition is to over-estimate inference attacks using over-estimated results of static type inference. The third contribution is to propose subclasses of schemas and queries for which the security problem becomes decidable. Lastly, the decidability of the security problem is shown to be incomparable with the static type inferability, although the tightness of the over-estimation of the inference attacks is affected in a large degree by that of the static type inference.

application/pdf, Detecting security flaws is important in order to keep the database secure. A security flaw in object-oriented databases means that a user can infer the result of an unpermitted method only from permitted methods. Although a database management system enforces access control by an authorization, security flaws can occur under the authorization. The main aim of this paper is to discuss the detection problem of security flaws for database schemas. This problem is to decide whether or not, when a schema S, an authorization A, and a term to be verified are given, there exists a database instance of S such that a security flaw on exists with respect to A. This paper shows that the problem is undecidable and proposes a decidable sufficient condition for a given schema to have no security flaw. Furthermore, this paper shows that the sufficient condition is also a necessary one if the given schema is monadic (i.e., every method is unary), and evaluates the time complexity to decide the condition.

Detecting security flaws is important in order to keep the database secure. A security flaw in object-oriented databases means that a user can infer the result of an unpermitted method only from permitted methods. Although a database management system enforces access control by an authorization, security flaws can occur under the authorization. The main aim of this paper is to discuss the detection problem of security flaws for database schemas. This problem is to decide whether or not, when a schema S, an authorization A, and a term to be verified are given, there exists a database instance of S such that a security flaw on exists with respect to A. This paper shows that the problem is undecidable and proposes a decidable sufficient condition for a given schema to have no security flaw. Furthermore, this paper shows that the sufficient condition is also a necessary one if the given schema is monadic (i.e., every method is unary), and evaluates the time complexity to decide the condition., technical report