An experimental study of Kannan’s embedding technique for the search LWE problem

Yuntao Wang, Yoshinori Aono, Tsuyoshi Takagi

研究成果: Chapter in Book/Report/Conference proceedingConference contribution

5 被引用数 (Scopus)

抄録

The learning with errors (LWE) problem is considered as one of the most compelling candidates as the security base for the post-quantum cryptosystems. For the application of LWE based cryptographic schemes, the concrete parameters are necessary: the length n of secret vector, the moduli q and the deviation σ. In the middle of 2016, Germany TU Darmstadt group initiated the LWE Challenge in order to assess the hardness of LWE problems. There are several approaches to solve the LWE problem via reducing LWE to other lattice problems. Xu et al.’s group solved some LWE Challenge instances using Liu and Nguyen’s adapted enumeration technique (reducing LWE to BDD problem) [14] and they published this result at ACNS 2017 [23]. In this paper, we study Kannan’s embedding technique (reducing LWE to unique SVP problem) to solve the LWE problem in the aspect of practice. The lattice reduction algorithm we use is the progressive BKZ [2, 3]. At first, from our experimental results we can intuitively observe that the embedding technique is more efficient with the embedding factor M closer to 1. Then especially for the cases of σ/q= 0.005, we will give an preliminary analysis for the runtime and give an estimation for the proper size of parameters. Moreover, our experimental results show that for n≥ 55 and the fixed σ/q= 0.005, the embedding technique with progressive BKZ is more efficient than Xu et al.’s implementation of the enumeration algorithm in [21, 23]. Finally, by our parameter setting, we succeeded in solving the LWE Challenge over (n, σ/q) = (70, 0.005) using 2 16.8 s (32.73 single core hours).

本文言語英語
ホスト出版物のタイトルInformation and Communications Security - 19th International Conference, ICICS 2017, Proceedings
編集者Sihan Qing, Dongmei Liu, Chris Mitchell, Liqun Chen
出版社Springer Verlag
ページ541-553
ページ数13
ISBN(印刷版)9783319894997
DOI
出版ステータス出版済み - 1 1 2018
イベント19th International Conference on Information and Communications Security, ICICS 2017 - Beijing, 中国
継続期間: 12 6 201712 8 2017

出版物シリーズ

名前Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
10631 LNCS
ISSN(印刷版)0302-9743
ISSN(電子版)1611-3349

その他

その他19th International Conference on Information and Communications Security, ICICS 2017
Country中国
CityBeijing
Period12/6/1712/8/17

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

フィンガープリント 「An experimental study of Kannan’s embedding technique for the search LWE problem」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル