AcBF: A High Accuracy Membership Filter using rDBF

Abstract

Bloom Filter is a data structure for membership query which is deployed in diverse research domains to boost up system’s performance and to lower on-chip memory consumption. However, there are still lacking of a high accuracy Bloom Filterwithoutcompromisingtheperformanceandmemoryspace. Moreover, the scalability causes more memory consumption as well as time complexity. Therefore, in this paper, we present a novel Bloom Filter, called accurate Bloom Filter (acBF), which features: a) an impressive guaranteed accuracy of 99.98%, b) a maximum false positive probability of 0.00015, c) lower collision probability, d) free from false negative, e) optimal insertion and membershipquerycost,andg)≤ 8−bits ofmemoryconsumption per item. acBF deploys eight multidimensional Bloom Filter. ThesemultidimensionalBloomFilterseliminatethefalsepositives at eight stages without sacrificing the system performance. We have conducted rigorous experiments to validate the accuracy of acBF which is unprecedentedly high. Also, acBF is compared with Scalable Bloom Filter (SBF) and Cuckoo Filter (CF). Experiments show acBF outperforms SBF and CF in terms of accuracy, and scalability. Moreover, performance of acBF outperforms CF in lookup operation. But, CF outperforms acBF in insertion. However, accuracy of acBF is incomparable with both SBF and CF.