Conventional secure concurrency controllers in the literature unfortunately fail to block conspiracy of transactions in diverse confidentiality levels. Some of them attempted to close internal communication channels under the assumption of restricted data resources. Fairness in transaction scheduling does matter in them in the sense that they forced higher level transactions to be sacrificed under the circumstances that any transaction conflicts with the other transactions in different confidentiality levels. To reduce unfairness, we propose a new concurrency control, named Triple Version Validation Scheme, TV2 for short.
TV2 closes the covert channel between more than two transactions. It also guarantees fairness between each confidentiality levels with restricted three data resources. TV2 needs to maintain at least three different data versions of each data objects for the purpose of preserving serializability. One of them is offered when a transaction requests the data of same confidentiality level. If some transactions require the data of lower confidentiality level, they are offered data items from another separated version. In addition, extra version is used for preserving serializability of database systems. By means of maintaining the three data versions and Serializable Schedule Graph, TV2 is able to produce serializable schedules without making covert channel which is produced by interfering a low level transaction. Performance results show that TV2 outperforms the Optimistic Orange Locking, OOL for short, in a terms of response time under the circumstance that DBMS classifies more than two confidentiality levels. And they show that TV2 improves equilibrium between transactions of each confidentiality levels. But abort ratio of TV2 appears higher than that of OOL under the circumstance that DBMS has only one confidentiality level. However we can say that abort ratio of TV2 is always lower than that of OOL in the Multilevel Secure DBMS because every Multilevel Secure DBMS classify more than 2 confidentiality levels.