Yes-no Bloom filter: A way of representing sets with fewer false positives

TL;DR

This paper introduces the yes-no Bloom filter, a new set representation method that significantly reduces false positives compared to traditional Bloom filters, while maintaining similar query efficiency.

Contribution

The paper proposes the yes-no Bloom filter, a novel variation that lowers false positives without increasing query complexity, inspired by applications in information-centric networking.

Findings

Yes-no BF has lower false positive rate than BF.

It maintains the same query processing requirements as BF.

Simulation results confirm improved false positive performance.

Abstract

The Bloom filter (BF) is a space efficient randomized data structure particularly suitable to represent a set supporting approximate membership queries. BFs have been extensively used in many applications especially in networking due to their simplicity and flexibility. The performances of BFs mainly depends on query overhead, space requirements and false positives. The aim of this paper is to focus on false positives. Inspired by the recent application of the BF in a novel multicast forwarding fabric for information centric networks, this paper proposes the yes-no BF, a new way of representing a set, based on the BF, but with significantly lower false positives and no false negatives. Although it requires slightly more processing at the stage of its formation, it offers the same processing requirements for membership queries as the BF. After introducing the yes-no BF, we show through simulations, that it has better false positive performance than the BF.