Fragmented-Iterated Bloom Filters for Routing in Distributed Event-Based Sensor Networks

Abstract

In this research, we propose the construction of a new architecture of Fragmented–Iterated Bloom Filters (FIBFs) to redirect complex events in a distributed event-based sensor network. We introduce two novel structures of Bloom Filters (BFs): Fragmented BFs (FBFs) and Iterated BFs (IBFs). The aim of IBFs is to discard single events that do not match any subscription. Then, FBFs deal with conjunctive and disjunctive set of events. Whether a match is found at the FBFs the publication is forwarded. Our strategy is theoretically and practically compared to the use of Standard BFs. The results show that FBFs lead to save memory and computational resources at the membership test. Moreover, we show that there is no memory cost for dividing a BF in smaller BFs using the same: (1) number of elements to insert and (2) probability of false positives. Then, we prove that FBFs may use fast hash functions that present a complexity of $O \left( \log _2\left( x\right) \right)$ while Standard BFs use hashes with a complexity of $O \left( \left( \log _2\left( y\right) \right) ^2\right)$. The hash output is represented by x, y $\in \mathbb {N}$ so that $x<y$. Additionally, it is shown that the use of the double hashing technique does not improve the computational complexity. Finally, we show that the construction of a structure of IBFs using an Iterated Hash Function (IHF) reduce the complexity because smaller filters and less hash functions are required.