Hardware-oriented optimization of Bloom filter algorithms and architectures for ultra-high-speed lookups in network applications

Sateesan, Arish; Vliegen, Jo; Daemen, Joan; Mentens, Nele

doi:10.1016/j.micpro.2022.104619

Cited by 8 publications

(2 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For the non-terminal set, we gathered the elementary operators used by state-ofthe-art NC hash functions as depicted in Table 2. The majority of the techniques use the multiplication operator as it produces significant entropy in the output, but it also consumes a lot of hardware resources on FPGA as is shown in [30], thus we did not use this operator. The use of a division operator is also avoided for the same reasons [2,24].…”

Section: Terminal and Non-terminal Setmentioning

confidence: 99%

“…Therefore, to have a fair comparison, we first implemented Xoodoo-NC on the same FPGA (Virtex Ultrascale+ xcvu7p-flvb2104-2-i) used in our work and by the authors of [6]. We chose the number of rounds for Xoodoo-NC to be 2.5 as the avalanche properties are satisfied at this stage [30]. We compared the latency, throughput, maximal operating frequency, and hardware resources.…”

Section: Hardware Evaluationmentioning

confidence: 99%

See 1 more Smart Citation

Evolving Non-cryptographic Hash Functions Using Genetic Programming for High-speed Lookups in Network Security Applications

Hassan¹,

Sateesan²,

Vliegen³

et al. 2023

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Non-cryptographic (NC) hash functions are the core part of many networking and security applications such as traffic flow monitoring and deep packet inspection. For these applications, speed is more important than strong cryptographic properties. In Terabit Ethernet networks, the speed of the hash functions can have a significant impact on the overall performance of the system when it is required to process the packets at a line rate. Hence, improving the speed of hash functions can have a significant impact on the overall performance of such architectures. Designing a good hash function is a challenging task because of the highly non-linear and complex relationship between input and output variables. Techniques based on Evolutionary Computation (EC) excel in addressing such challenges. In this paper, we propose novel fast noncryptographic hash functions using genetic programming, and we call the resulting hash functions the GPNCH (Genetic Programming-based Non-Cryptographic Hash) family. We choose to employ avalanche metrics as a fitness function because the networking and security applications we consider require hash functions to be uniform and independent. We evaluate the performance of GPNCH functions on FPGA and compare the delay, throughput, and resource occupation with the state-of-the-art NC hash functions that satisfy the avalanche criteria. We show that GPNCH functions outperform the other algorithms in terms of latency, operating frequency, and throughput at the modest cost of hardware resources.

show abstract

Section: Terminal and Non-terminal Setmentioning

confidence: 99%

Section: Hardware Evaluationmentioning

confidence: 99%