Professional Documents
Culture Documents
Modern Cryptography Volume 2: A Classical Introduction To Informational and Mathematical Principle 1st Edition Zhiyong Zheng
Modern Cryptography Volume 2: A Classical Introduction To Informational and Mathematical Principle 1st Edition Zhiyong Zheng
Modern Cryptography Volume 2: A Classical Introduction To Informational and Mathematical Principle 1st Edition Zhiyong Zheng
https://ebookmeta.com/product/a-modern-introduction-to-classical-
electrodynamics-1st-edition-michele-maggiore/
https://ebookmeta.com/product/a-modern-introduction-to-
mathematical-analysis-1st-edition-alessandro-fonda/
https://ebookmeta.com/product/classical-and-modern-cryptography-
for-beginners-rajkumar-banoth-rekha-regar/
https://ebookmeta.com/product/classical-mechanics-
volume-2-kinematics-and-uniformly-accelerated-motion-gregory-a-
dilisi/
California Studies in Classical Antiquity Volume 2
https://ebookmeta.com/product/california-studies-in-classical-
antiquity-volume-2/
https://ebookmeta.com/product/modern-app-development-with-dart-
and-flutter-2-a-comprehensive-introduction-to-flutter-1st-
edition-meiller/
https://ebookmeta.com/product/introduction-to-classical-and-
quantum-computing-1st-edition-thomas-wong/
https://ebookmeta.com/product/introduction-to-classical-legal-
rhetoric-a-lost-heritage-1st-edition-michael-h-frost/
https://ebookmeta.com/product/a-modern-introduction-to-dynamical-
systems-richard-brown/
Financial Mathematics and Fintech
Zhiyong Zheng
Kun Tian
Fengxia Liu
Modern
Cryptography
Volume 2
A Classical Introduction to Informational
and Mathematical Principle
Financial Mathematics and Fintech
Series Editors
Zhiyong Zheng, Renmin University of China, Beijing, Beijing, China
Alan Peng, University of Toronto, Toronto, ON, Canada
This series addresses the emerging advances in mathematical theory related to
finance and application research from all the fintech perspectives. It is a series of
monographs and contributed volumes focusing on the in-depth exploration of
financial mathematics such as applied mathematics, statistics, optimization, and
scientific computation, and fintech applications such as artificial intelligence, block
chain, cloud computing, and big data. This series is featured by the comprehensive
understanding and practical application of financial mathematics and fintech. This
book series involves cutting-edge applications of financial mathematics and fintech
in practical programs and companies.
The Financial Mathematics and Fintech book series promotes the exchange of
emerging theory and technology of financial mathematics and fintech between
academia and financial practitioner. It aims to provide a timely reflection of the state
of art in mathematics and computer science facing to the application of finance. As a
collection, this book series provides valuable resources to a wide audience in
academia, the finance community, government employees related to finance and
anyone else looking to expand their knowledge in financial mathematics and
fintech.
The key words in this series include but are not limited to:
a) Financial mathematics
b) Fintech
c) Computer science
d) Artificial intelligence
e) Big data
Zhiyong Zheng · Kun Tian · Fengxia Liu
Modern Cryptography
Volume 2
A Classical Introduction to Informational
and Mathematical Principle
Zhiyong Zheng Kun Tian
School of Mathematics School of Mathematics
Renmin University of China Renmin University of China
Beijing, China Beijing, China
Henan Academy of Sciences
Zhengzhou, China
Fengxia Liu
Artificial Intelligence Research Institute
Beihang University
Beijing, China
© The Editor(s) (if applicable) and The Author(s) 2023. This book is an open access publication.
Open Access This book is licensed under the terms of the Creative Commons Attribution 4.0 International
License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribu-
tion and reproduction in any medium or format, as long as you give appropriate credit to the original
author(s) and the source, provide a link to the Creative Commons license and indicate if changes were
made.
The images or other third party material in this book are included in the book’s Creative Commons license,
unless indicated otherwise in a credit line to the material. If material is not included in the book’s Creative
Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright holder.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Preface
v
vi Preface
calculation speed and small storage space. In 2009, the National Institute of
Standards and Technology wrote a survey report: there is no cryptosystem could
consider both public key encryption and digital signature, and resist the Shor
algorithm simultaneously. The NTRU encryption algorithm seems to be the most
likely choice among many lattice-based encryption schemes. The PQCRYPTO
program (Horizon 2020 ICT-645622) by European Union hopes to develop a new
European encryption standard based on the NTRU improved by Stehle-Steinfeld.
4. MacElience/Niderreiter cryptosystem (1998). Linear codes are the earliest error-
correcting codes in coding theory. Later, algebraic coding developed based on
the ideal theory greatly enriched and improved the linear coding theory. Cycle
code and Goppa code are the most important error-correcting codes in algebraic
coding. MacElience and Niderreiter constructed a new public key cryptosystem
by using the asymmetry of encoding algorithm and decoding algorithm of the
error-correcting code independently, which we call MacElience/Niderreiter cryp-
tosystem. Since a code (linear code or algebraic code) can be regarded as a lattice
on a finite field, the security of this cryptosystem is closely related to the closest
vector problem on the q-ary integer lattice. Recent studies have shown that coding
theory plays an important role in lattice-based cryptosystems.
5. LWE cryptosystem (2005). In 2005, O. Regev of Tel Aviv University in Israel
proposed the famous LWE cryptosystem based on the LWE distribution. Because
of this work, Regev won the highest award in the theoretical computer science
in 2018—the Godel Award. The LWE distribution (Learning With Errors) is
a random linear system with errors having Gauss distribution. Regev’s cryp-
tosystem encrypts a single bit of plaintext each time. Since the security of the LWE
problem has been clearly proved (see Chap. 3 of this book), LWE cryptosystem
is currently the most active and mainstream research topic.
6. Fully homomorphic encryption (FHE). In 1985, R.Rivest, C.Adleman and
M.Dertouzos first proposed the concept of data bank and the conjecture of fully
homomorphic encryption. Some individuals and organizations encrypt the orig-
inal data and store them in the data bank for privacy protection, which is obviously
a huge wealth. How to compute these encrypted data effectively? R. Rivest, C.
Adleman and M. Dertouzos presented the fully homomorphic encryption conjec-
ture. In 2009, C. Gentry of Stanford University partially solved the RAD conjec-
ture. Gentry’s work is based on the ideal lattice, that is, an integer lattice which
has a one-to-one correspondence to the ideal of polynomial ring. But the cryp-
tosystem of Gentry is a finite-time fully homomorphic encryption, and infinite
fully homomorphic encryption is still an unsolved public problem. In 2012 and
2013, the second and third fully homomorphic encryption algorithms based on
the LWE distribution were proposed one after another. Gentry won the 2022
Godel Award for his contributions.
In the book Modern Cryptography, we give a detailed introduction to the basic
theory of lattice and the first four kinds of lattice-based cryptosystems. The main
purpose of this book is to discuss the computational complexity theory of lattice
Preface vii
cryptosystems, especially Ajtai’s reduction principle, and fill the gap that post-
quantum cryptography focuses on the encryption and decryption algorithms, and
the theoretical proof is insufficient. In Chaps. 3, 4 and 6, we introduce the LWE
distribution, LWE cryptosystem and fully homomorphic encryption in detail. When
using stochastic analysis tools, there are many ‘ambiguity’ problems in terms of
definitions and algorithms, such as the ‘≈’ notation appeared in a large number of
papers and books, which is unprecise mathematically. The biggest characteristic of
this book is to use probability distribution to provide rigorous mathematical defi-
nitions and proofs for various unclear expressions, making it a rigorous theoretical
system to facilitate teaching and dissemination in class. Chapters 5 and 7 are based
on two papers published by the authors in the journal Journal of Information Security
(see references [63, 64]). These materials can be regarded as some important topics,
such as the further extension and improvement of cyclic lattices, ideal lattices and
generalized NTRU cryptosystems.
This book contains the most cutting-edge and hottest research topics in post-
quantum cryptography. Reading all the chapters requires a lot of mathematical
knowledge and a good mathematical foundation. Therefore, this book can be used
as a textbook for graduate students in mathematics and cryptography, or a reference
book for researchers in cryptography area. Due to the rush of time, all the mate-
rials are summarized from domestic and foreign research papers in the last 20 years,
and shortcomings and mistakes are inevitable. We welcome readers to criticize and
correct them.
ix
x Contents
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Notations
xi
Chapter 1
Random Lattice Theory
⎛ ⎞ ⎛ ⎞
x1 y1
⎜ .. ⎟ ⎜ .. ⎟
Let R be the Euclidean space of dimension n, x = ⎝ . ⎠, y = ⎝ . ⎠ are two vectors
n
xn yn
of Rn , the inner product of x and y is defined as
x · y = x1 y1 + x2 y2 + · · · + xn yn = x T y. (1.0.1)
The Euclidean norm |x| of vector x (also called the l2 norm) is defined as
1 √
|x| = (x12 + x22 + · · · + xn2 ) 2 = x · x. (1.0.2)
N (x0 , r ) = {x ∈ Rn | |x − x0 | r }, x0 ∈ Rn . (1.0.5)
In particular, N (0, r ) represents a sphere with origin as the center of the circle and
radius r . The discretization of a lattice is equivalent to the fact that the intersection
of L with any sphere N (x0 , r ) is a finite set, i.e.
#
{L ∩ N (x0 , r )} < ∞. (1.0.6)
n
F(B) = { xi βi | 0 xi < 1}. (1.0.7)
i=1
therefore, F ∗ (B) can also be a basic neighborhood of the lattice L. The following
property is easy to prove [see Lemma 2.6 in Chap. 7 in Zheng (2022)]
That is, the volume of the basic neighborhood of L is an invariant and does not
change with the choice of the generated matrix B. We denote det(L) = |det(B)| as
the determinant of the lattice L.
The basic properties of lattice can be found in Chap. 7 of Zheng (2022). The main
purpose of this chapter is to establish the random theory of lattice. If a lattice L is the
space of values of a random variable (or random vector), it is called a random lattice.
Random lattice is a new research topic in lattice theory, and the works of Micciancio
and Regev (2004), Regev (2004), Micciancio and Regev (2004), Micciancio and
Regev (2009) are pioneering. In this way, the study of random lattice is no more
than ten years. For technical reasons, only a special class of random lattices can be
defined and studied. That is, consider a random variable ξ defined in Rn from a Gauss
distribution, and limit the discretization of ξ to L so that L becomes a random lattice.
It is a special kind of random lattice, which we call the Gauss lattice. The main purpose
of this chapter is to introduce Gauss lattice, define the smoothing parameter on Gauss
lattice and calculate the statistical distance based on the smoothing parameter. The
mathematical technique used in this chapter is high dimensional Fourier transform.
1.1 Fourier Transform 3
and
L 2 (R) = { f : Rn → C | | f (x)|2 dx < ∞}. (1.1.2)
Rn
The following are some of the most common and fundamental properties of Fourier
transform.
f ∗ g(x) = f ∗ g(ξ )e−2πi x·ξ dξ
Rn
f ∗ g(x) = g(y)e−2πi x·y dy · f (y )e−2πi x·y dy = fˆ(x)ĝ(x),
Rn Rn
1.1 Fourier Transform 5
τa f (x) = f (ξ + a)e−2πi x·ξ dξ = f (y)e−2πi x·(y−a) dy
Rn Rn
property (ii) gets proved. Similarly, we can obtain (iii). Next, we give the proof of
(iv). Since δ = 0, and f δ (x) = f ( 1δ x), so
1
fˆδ (x) = f ( ξ )e−2πi x·ξ dξ = f (y)e−2πi x·δy |δ|n dy
δ
Rn Rn
f ◦ A(x) = f (Aξ )e−2πi x·ξ dξ.
Rn
f ◦ A(x) = f (y)e−2πi x·A
−1
y
|A|−1 dy = |A|−1 f (y)e−2πi((A
−1 T
) x·y)
dy
Rn Rn
Example 1.1 Let n = 1, a ∈ R, a > 0, define the characteristic function 1[−a,a] (x)
of the closed interval [−a, a] as
1, x ∈ [−a, a],
1[−a,a] (x) =
0, x ∈
/ [−a, a].
Then
sin 2πax
1̂[−a,a] (x) = . (1.1.6)
πx
6 1 Random Lattice Theory
Define the characteristic function 1[−a,a] (x) of the square [−a, a], then
sin 2πai xi
1̂[−a,a] (x) = i=1
n
. (1.1.7)
π xi
Example 1.2 Let f (x) = e−π|x| , x ∈ Rn , then f (x) ∈ L 1 (Rn ), and fˆ(x) = f (x),
2
namely f (x) is a fixed point of Fourier operator, which is also called a dual function.
Proof Clearly, f (x) ∈ L 1 (Rn ). To prove the fixed point property of f (x), by defi-
nition
Rn Rn Rn
+∞
e−π y dy = 1,
2
(1.1.8)
−∞
e−π|y| dy = 1.
2
(1.1.9)
Rn
From the property of f (x) = e−π|x| under the Fourier operator introduced in the last
2
section, and high dimensional Poisson integral formula (1.1.9), we can generalize
f (x) as the density function of a random variable from the normal Gauss distribution
to a general Gauss distribution in Rn . We first discuss the Gauss function on Rn .
Definition 1.2.1 Let s > 0 be a given positive real number, c ∈ Rn is a vector. The
Gauss function ρs,c (x) centered on c with parameter s is defined as
π
ρs,c (x) = e− s2 |x−c| , x ∈ Rn
2
(1.2.1)
and
ρs (x) = ρs,0 (x), ρ(x) = ρ1 (x) = e−π|x| .
2
(1.2.2)
s
and
ρs (x) = ρs (x1 ) . . . ρs (xn ).
Lemma 1.2.1 The Fourier transform of Gauss functions ρs (x) and ρs,c (x) are
and
ρ̂s,c (x) = e−2πi x·c s n ρ1/s (x). (1.2.5)
The last equation follows from Example 2 in the previous section, therefore, (1.2.4)
holds. By the property (ii) of Lemma 1.1.2, we have
−2πi x·c
ρ̂s,c (x) = τ
−c ρs (x) = e ρ̂s (x) = s n e−2πi x·c ρ1/s (x).
Lemma 1.2.2 ρs,c (x) is uniformly continuous in Rn , i.e. for any > 0, there is
δ = δ( ), when |x − y| < δ for x ∈ Rn , y ∈ Rn , we have
Proof By definition, 0 < ρs,c (x) 1, hence ρs,c (x) is uniformly bounded in Rn , we
will prove ρs,c (x) is also uniformly bounded in Rn . We only prove the case of c = 0.
Since ρs (x) = ρs (x1 ) = · · · = ρs (xn ), without loss of generality, let n = 1, t ∈ R,
then
2π π 2
ρs (t) = − 2 te− s2 t .
s
When |t| M, it is clear
π 1
e− s 2 t
2
.
|t|2
2π 2π
|ρs (t)| 2 .
s |t|
2 s M
For |t| < M, By the continuity of ρs (t) we have ρs (t) is bounded. This gives the
proof that ρs,c (x) is uniformly continuous in Rn . Let |ρs,c (x)| M0 , ∀x ∈ Rn . By
the differential mean value theorem, we have
Let δ = M0
, then
|ρs,c (x) − ρs,c (y)| < , if |x − y| < δ.
Definition 1.2.2 For s > 0, c ∈ Rn , define the continuous Gauss density function
Ds,c (x) as
1
Ds,c (x) = n ρs,c (x), ∀x ∈ Rn . (1.2.6)
s
The definition gives that
1
Ds,c (x)dx = ρs,c (x)dx = 1.
sn
Rn Rn
The continuous Gauss density function Ds,c (x) is also called the continuous Gauss
measure. In order to implement the transformation from continuous measure to dis-
crete measure and define random variables on discrete geometry in Rn , the following
lemma is an important theoretical support.
1 1 − π2 |x−c|2
Ds,c (L) = ρs,c (x) = e s .
s n x∈L s n x∈L
By the property of the exponential function et , there exists a constant M0 > 0, when
|x − c| > M0 ,
π s2
e− s2 |x−c|
2
. (1.2.8)
π |x − c|2
Thus, we can divide the points on the lattice L into two sets. Let
A1 = L ∩ {x ∈ Rn | |x − c| M0 } = L ∩ N (c, M0 ).
and
A2 = L ∩ {x ∈ Rn | |x − c| > M0 }.
Based on (1.2.8),
π s2
e− s2 |x−c|
2
< ∞. (1.2.9)
x∈A2 x∈A2
π |x − c|2
10 1 Random Lattice Theory
Since A2 is a countable set, the right hand side of the above inequality is clearly a
convergent series. Combining the above two estimations, we have Ds,c (L) < ∞, the
lemma is proved.
P T T P = diag{δ1 , δ2 , . . . , δn }.
Hence,
|Ax|2 = x T T x = x T P(P T T P)P T x.
We call A for the matrix norm of A, and Lemma 1.2.4 shows that
√ √
δ ||A|| δ ∗ , ∀A ∈ G L n (R). (1.2.12)
Another proof of Lemma 1.2.3: Let L = L(B) be any full-rank lattice, B is the
generated matrix of L. By definition we have
1 − π2 |x−c|2 1 − π2 |Bx−c|2
Ds,c (L) = Ds,c (x) = n
e s = n e s . (1.2.13)
x∈L
s x∈L s n x∈Z
1.2 Discrete Gauss Measure 11
|B −1 x|
||B −1 || ⇒ |B −1 x| ||B −1 || |x|, ∀x ∈ Rn .
|x|
1 √
|y| ||B −1 || |By| ⇒ |By| −1
|y| |y|/ δ ∗ , ∀y ∈ Rn . (1.2.14)
||B ||
π s 2n
e− s2 |Bx−c|
2
. (1.2.15)
π n |Bx − c|2n
x∈Zn ,|Bx−c|>M x∈Zn ,|Bx−c|=0
Since
|Bx − c|2n = |B(x − B −1 c)|2n |x − B −1 c|2n /(δ ∗ )n .
n
|x − B −1 c|2n = ( (xi − u i )2 )n (n n i=1
n
(xi − u i )2 )n = n n i=1
n
(xi − u i )2 .
i=1
By (1.2.15),
s 2n s 2n (δ ∗ )n 1
· n
x∈Zn ,|Bx−c|=0
π n |Bx − c| 2n
x∈Zn ,|Bx−c|=0
π nn n i=1 (xi − u i )2
s 2n (δ ∗ )n 1 1 1
= ··· ,
π nn n (x1 − u 1 ) 2 (x2 − u 2 ) 2 (xn − u n )2
x1 ∈Z x2 ∈Z xn ∈Z
every infinite series on the right hand side of the above equation converges, hence,
Ds,c (L) < ∞.
By Lemma 1.2.3, we define the discrete Gauss density function D L ,s,c (x) as
Trivially, we have
D L ,s,c (x) = 1.
x∈L
12 1 Random Lattice Theory
Definition 1.2.4 Let L = L(B) ⊂ Rn be a lattice with full rank, s > 0 is a given
positive real number, c ∈ Rn is a given vector, define the discrete Gauss measure
function g L ,s,c (x) as a function defined on the basic neighborhood F(B) of L,
1
g L ,s,c (x) = Ds,c (x̄) = ρs,c (x + y), x ∈ F(B). (1.2.17)
s n y∈L
1 1
g L ,s,c (x)dx = ρs,c (x + y)dx = ρs,c (x)dx = 1. (1.2.18)
s n y∈L sn
F(B) F(B) Rn
Thus, the density function g L ,s,c (x) defined on the basic neighborhood F(B) corre-
sponds to a continuous random variable on F(B), denoted as Ds,c modL.
Lemma 1.2.5 The random variable Ds,c modL is actually defined in the additive
quotient group Rn /L.
Proof F(B) is a set of representative elements of the additive quotient group Rn /L,
and we only prove that for any set of representative elements of Rn /L, the discrete
Gauss function g L ,s,c (x) remains constant, then Ds,c mod L can be regarded as
a random variable on the additive quotient group Rn /L. Actually, if x1 , x2 ∈ Rn ,
x1 ≡ x2 (mod L), we have g L ,s,c (x1 ) = g L ,s,c (x2 ). To obtain the result, by definition
1
g L ,s,c (x1 ) = Ds,c (x¯1 ) = ρs,c (x1 + y).
s n y∈L
Since x1 = x2 + y0 , where y0 ∈ L, so
1 1
g L ,s,c (x1 ) = n
ρs,c (x1 + y) = n ρs,c (x2 + y0 + y)
s y∈L s y∈L
1
= ρs,c (x2 + y) = Ds,c (x¯2 ) = g L ,s,c (x2 ).
s n y∈L
By x1 ≡ x2 (mod L), then x¯1 = x¯2 are the same additive cosets in the quotient
group Rn /L. Thus, the discrete Gauss measure g L ,s,c (x) can be defined on any basic
neighborhood of L, and the corresponding random variable Ds,c mod L is actually
defined on the quotient group Rn /L.
1.3 Smoothing Parameter 13
For a given full-rank lattice L ⊂ Rn , in the previous section we defined the dis-
crete Gauss measure g L ,s,c (x), and the corresponding continuous random variable
Ds,c mod L on the basic neighborhood F(B) of L. In this section, we discuss an
important parameter on Gauss lattice—the smoothing parameter. The concept of
smooth parameters was introduced by Micciancio and Regev in 2007 Micciancio
and Regev (2004). For a given vector x ∈ Rn , we have the following lemma.
or equally
lim ρ1/s (x) = 0.
s→∞
x∈L\{0}
Proof By the property of the exponential function, when |x| > M0 (M0 is a positive
constant) then
1
e−πs |x|
2 2
.
π s 2 |x|2
So
1 1
e−πs |x|2
e−πs |x|2
2 2
ρ1/s (x) = + .
x∈L x∈L |x|M0 ,x∈L
π s2 |x|>M0 ,x∈L
|x|2
The first part of the equation above only has a finite number of terms, so
e−πs |x|2
2
lim = 1.
s→∞
|x|M0 ,x∈L
1 1
lim = 0.
s→∞ π s2 |x|>M0 ,x∈L
|x|2
Definition 1.3.1 Let L ⊂ Rn be a lattice with full rank, L ∗ is the dual lattice of L,
define the smoothing parameter η (L) of L: For any > 0, define
Equally,
η (L) = min{s | s > 0, ρ1/s (L ∗ \{0}) < }. (1.3.2)
Definition 1.3.2 Let A ⊂ Rn be a finite or countable set, X and Y are two discrete
random variables on A, the statistical distance between X and Y is defined as
1
(X, Y ) = |Pr {X = a} − Pr {Y = a}|. (1.3.3)
2 a∈A
( f (X ), f (Y )) (X, Y ).
From (1.2.17) in the last section, Ds,c mod L is a continuous random variable
defined on the basic neighborhood F(B) of the lattice L with the density function
g L ,s,c (x). Let U (F(B)) be a uniform random variable defined on F(B) with the
density function d(x) = det(L)
1
. The main result of this section is that the statistical
distance between Ds,c mod L and the uniform distribution U (F(B)) can be arbitrar-
ily small.
Theorem 1.1 For any s > 0, given a lattice with full rank L = L(B) ⊂ Rn , L ∗ is the
dual lattice of L, then the statistical distance between the discrete Gauss distribution
and the uniform distribution on the basic neighborhood F(B) satisfies
1
(Ds,c mod L , U (F(B))) ρ1/s (L ∗ \{0}). (1.3.5)
2
Particularly, for any > 0, and any s η (L), we have
1.3 Smoothing Parameter 15
1
(Ds,c mod L , U (F(B))) . (1.3.6)
2
To prove Theorem 1.1, we first introduce the following lemma.
1 ˆ
f (x) = f (y),
x∈L
det(L) y∈L ∗
Since F(u) is a periodic function of the lattice Zn , namely F(u + x) = F(u), for
∀x ∈ Zn , we have the following Fourier expansion
F(u) = a(y)e2πiu·y . (1.3.7)
y∈Zn
= f (z)e−2πi z·y dz = f (z)e−2πi z·y dz = fˆ(y).
x∈Zn x+[0,1]n Rn
16 1 Random Lattice Theory
Take u = 0, we have
F(0) = f (x) = fˆ(y),
x∈Zn y∈Zn
the lemma is proved for L = Zn . For the general case L = L(B), since L ∗ =
L((B −1 ) ), then
f (x) = f (Bx) = ( f ◦ B)(x),
x∈L x∈Zn x∈Zn
where f ◦ B(x) = f (Bx). Replace f (x) with f ◦ B, then f ◦ B still satisfies the
conditions of this lemma, so
f ◦ B(x) = f ◦ B(y).
x∈Zn y∈Zn
f ◦ B(y) = f (Bt)e−2πi y·t dt.
Rn
1 −1
f ◦ B(y) = f (x)e−2πi y·B x
dx
|det(B)|
Rn
1 −1
= f (x)e−2πi(B ) y·x
dx
|det(B)|
Rn
1
= fˆ((B −1 ) y).
|det(B)|
Above all,
1 1
f (x) =
f ◦ B(y) = fˆ((B −1 ) y) = fˆ(y).
|det(B)| |det(B)| y∈L ∗
x∈L y∈Zn n y∈Z
The proof of Theorem 1.1 The density function of the continuous random variable
Ds,c mod L defined on the basic neighborhood F(B) of L is g L ,s,c (x), from Eq.
(1.2.17) and Lemma 1.3.2, we have
1 1
g L ,s,c (x) = n
ρs,c (x + y) = n ρs,c−x (y).
s y∈L s y∈L
1
g L ,s,c (x) = e2πi y·(x−c) ρ1/s (y). (1.3.8)
|det(B)| y∈L ∗
The density function of the uniformly distributed random variable U (F(B)) on F(B)
1
is |det(B)| , based on the definition of statistical distance,
1 1
(Ds,c mod L , U (F(B))) = |g L ,s,c (x) − |dx
2 |det(B)|
F(B)
1 1
= | e2πi y·(x−c) ρ1/s (y)|dx
2 |det(B)| y∈L ∗ ,y=0
F(B)
1
Vol(F(B))det(L ∗ ) max | e2πi y·(x−c) ρ1/s (y)|
2 x∈F(B)
y∈L ∗ \{0}
1 1
ρ1/s (y) = ρ1/s (L ∗ \{0}).
2 y∈L ∗ \{0}
2
1
(Ds,c mod L , U (F(B))) .
2
Thus, Theorem 1.1 is proved.
18 1 Random Lattice Theory
Proof By Definition 1.2.1, the left hand side of the sum in the above inequality can
be written as √
π
ρ√a (x) = e− a |x| , s = a.
2
√
Obviously ρs (x) is a monotone increasing function of s, take s = a 1, then
ρ√a (x) = a 2 det(L ∗ ) ρ √1 (x) a 2 det(L ∗ )
n n
ρ(x)
a
x∈L x∈L ∗ x∈L ∗
e−π|x| .
n n 2
= a2 ρ(x) = a 2
x∈L x∈L
N = {x ∈ Rn | |x| 1}.
Lemma 1.3.4 Suppose L ⊂ Rn is a lattice with full rank, c > √1 is a positive real
√ 2π
number, C = c 2π e · e−πc , v ∈ Rn , then
2
√ √
ρ(L\c n N ) < C n ρ(L), and ρ((L + v)\c n N ) < 2C n ρ(L).
That is,
e−π|x| < C n e−π|x| ,
2 2
(1.3.10)
√
x∈L ,x ∈c
/ nN x∈L
e−π|x| < 2C n e−π|x| .
2 2
√
x∈L+v,x ∈c
/ nN x∈L
1.3 Smoothing Parameter 19
Proof We will prove the first inequality, ler t be a positive real number, 0 < t < 1,
then
e−πt|x| = eπ(1−t)|x| · e−π|x|
2 2 2
x∈L x∈L
eπ(1−t)|x| · e−π|x|
2 2
>
x∈L ,|x|2 c2 n
eπ(1−t)c e−π|x| .
2 2
n
x∈L x∈L
Hence,
e−π|x| < e−π(1−t)c e−πt|x| e−π(1−t)c n t − 2 e−π|x| .
2 2 2 2 n 2
n
It implies that
√
ρ(L\c n N ) < (t − 2 e−π(1−t)c )n ρ(L).
1 2
Let t = 1
2πc2
, then
√ √
ρ(L\c n N ) < (c · 2π e · e−πc )n ρ(L),
2
The second inequality can be proved in the same way. Lemma 1.3.4 holds.
Based on the above inequality, we can give an upper bound estimation of the
smoothing parameter on lattice, which is a very important result about the smoothing
parameter.
Theorem 1.2 For any n dimensional full-rank lattice L ⊂ Rn , we have
√
η2−n (L) n/λ1 (L ∗ ). (1.3.11)
√ Cn √
ρ(L ∗ \ n N ) < ρ(L ∗ ∩ n N ).
1−C n
√
If s > n/λ1 (L ∗ ), for all x ∈ L ∗ \{0},
√ √
|sx| s · λ1 (L ∗ ) > n ⇒ s L ∗ ∩ n N = {0}.
Hence, √
ρ1/s (L ∗ ) = ρ(s L ∗ ) = 1 + ρ(s L ∗ \ n N )
Cn √
<1+ ρ(s L ∗ ∩ n N )
1−C n
Cn 2−2n
=1+ < 1 + = 2−n + 1.
1 − Cn 2−n
According to the proof of Theorem 1.2, we can further improve the upper bound
estimation of the smoothing parameter.
√ Cn 1
2π e · e−πc ⇒
2
C =c· < n. (1.3.15)
1−C n 2
√ Cn √
ρ(L ∗ \c n N ) < ρ(L ∗ ∩ c n N ).
1−C n
√
If s > c n/λ1 (L ∗ ), for any x ∈ L ∗ \{0},
√
|sx| sλ1 (L ∗ ) > c n.
Hence, √
s L ∗ ∩ c n N = {0}.
Therefore,
√ Cn 1
ρ1/s (L ∗ ) = ρ(s L ∗ ) = 1 + ρ(L ∗ \c n N ) < 1 + < 1+ n.
1 − Cn 2
4√
η2−n (L) n/λ1 (L ∗ ). (1.3.16)
5
Proof Take c = 4
5
in Lemma 1.3.4, then c > √1 ,
2π
and
√ Cn 1
2π e · e−πc ⇒
2
C =c· < n.
1−C n 2
√ Cn √
ρ(L ∗ \c n N ) < ρ(L ∗ ∩ c n N ).
1−C n
22 1 Random Lattice Theory
√
If s > c n/λ1 (L ∗ ), for any x ∈ L ∗ \{0},
√
|sx| sλ1 (L ∗ ) > c n.
Hence, √
s L ∗ ∩ c n N = {0}.
We get
√ Cn 1
ρ1/s (L ∗ ) = ρ(s L ∗ ) = 1 + ρ(L ∗ \c n N ) < 1 + < 1+ n,
1 − Cn 2
In the following, we give another classical upper bound estimation for the smooth-
ing parameter. For any n dimensional full-rank lattice L ⊂ Rn , we have introduced
the definition of minimal distance λ1 (L) on lattice, which can actually be generalized
to the general case. For 1 i n,
λi (L) is also called the i-th continuous minimal distance of lattice L. To give an
upper bound estimation of the smoothing parameter, we first prove the following
lemma.
= det(L ∗ ) e−2πic·y ρ̂s (y)
y∈L ∗
det(L ∗ ) ρ̂s (y) = ρs (L),
y∈L ∗
1.3 Smoothing Parameter 23
where we have used ρ̂s,c (y) = e−2πic·y ρ̂s (y), the lemma gets proved.
Theorem 1.3 For any n dimensional full-rank lattice L, > 0, we have
ln(2n(1 + 1/ ))
η (L) λn (L), (1.3.19)
π
where λn (L) is the N -th continuous minimal distance of the lattice L defined by
(1.3.17).
Proof Let
ln(2n(1 + 1/ ))
s= λn (L),
π
we need to prove ρ1/s (L ∗ \{0}) . From the definition of λn (L), there are n linearly
independent vectors v1 , v2 , . . . , vn in L satisfying |vi | λn (L), and for any positive
integer k > 1, we have vi /k ∈ / L, 1 i n. The main idea of the proof is to take a
segregation of L ∗ , for any integer j, let
Si, j = {x ∈ L ∗ | x · vi = j} ⊂ L ∗ ,
j2 j2
(x − ju i ) · ju i = j x · u i − j 2 u i · u i = − = 0.
|vi |2 |vi |2
Therefore,
|x|2 = |x − ju i |2 + | ju i |2 .
So
e−πs |x|2
2
ρ1/s (Si, j ) =
x∈Si, j
= e−πs | ju i |2
e−πs |x− ju i |2
2 2
x∈Si, j
= e−πs | ju i |2
2
ρ1/s (Si, j − ju i ). (1.3.21)
Since the inner product of any vector in Si, j − ju i with vi is 0, then Si, j − ju i
is actually a translation of Si,0 , namely there is a vector w satisfying Si, j − ju i =
24 1 Random Lattice Theory
e−π(s/λn (L))
2 2
j
ρ1/s (Si,0 )
j=0
2
ρ1/s (Si,0 )
eπ(s/λn (L))2 −1
2
= (ρ1/s (L ∗ ) − ρ1/s (L ∗ \Si,0 )).
eπ(s/λn (L))2 − 1
So we get
2
ρ1/s (L ∗ \Si,0 ) ρ1/s (L ∗ ).
eπ(s/λn (L))2 +1
From (1.3.20),
n
2n
∗
ρ1/s (L \{0}) ρ1/s (L ∗ \Si,0 ) ρ1/s (L ∗ ).
i=1
eπ(s/λn (L))2 + 1
2n 2n
ρ1/s (L ∗ \{0}) < π(s/λ (L))2 = .
eπ(s/λn (L))2 + 1 − 2n e n − 2n
1.4 Some Properties of Discrete Gauss Distribution 25
At the end of this section, we present an inequality for the minimal distance on
lattice, which will be used in the next chapter when we prove that the LWE problem
is polynomial equivalent with the hard problems on lattice.
λ1 (L ∗ )
= ρ1/s (L ∗ \{0}) ρ1/s (v) = e−πs
2 2
.
Hence,
ln 1/ 1
s .
π λ1 (L ∗ )
That is, the first inequality in this lemma holds. For the second inequality, Theorem
2.1 in Banaszczyk (1993) implies that
In this section we introduce some properties about the discrete Gauss distribution.
First we give the definition of the expectation of discrete Gauss distribution.
26 1 Random Lattice Theory
as the expectation of f (ξ ).
Lemma 1.4.1 For any n dimensional full-rank lattice, L ⊂ Rn , c, u ∈ Rn , |u| = 1,
0 < < 1, s 2η (L), ξ is a random variable from the discrete Gauss distribution
D L ,s,c , then we have
s
|E[(ξ − c) · u]| , (1.4.3)
1−
and
s2 s2
|E[((ξ − c) · u)2 ] − | . (1.4.4)
2π 1−
Proof Let L = L/s = { xs | x ∈ L}, c = c/s, ξ is a random variable from the dis-
crete Gauss distribution D L ,c , for any x ∈ L , we have
ξ
E[(ξ − c) · u] = s E[( − c ) · u] = s E[(ξ − c ) · u],
s
the inequality (1.4.3) is equivalent to
1
|E[((ξ − c ) · u)2 ] − | . (1.4.6)
2π 1−
So we only need to prove the two inequalities for s = 1. Denote ξ as a random variable
from the discrete Gauss distribution D L ,c , the condition s 2η (L) in Lemma 1.4.1
1.4 Some Properties of Discrete Gauss Distribution 27
becomes η (L) 21 . We prove that the two inequalities (1.4.5) and (1.4.6) hold if
u = (1, 0, . . . , 0) firstly. For any positive integer j, let
g j (L)
E[((ξ − c) · u) j ] = E[(ξ1 − c1 ) j ] = .
ρc (L)
In order to estimate ρ̂c (L ∗ ), from Lemma 1.2.1 we get ρ̂c (x) = e−2πi x·c ρ(x), thus,
|ρ̂c (x)| = ρ(x), note that η (L) 21 < 1,
|ρ̂c (L ∗ )| = |1 + ρ̂c (x)| 1 − |ρ̂c (x)| = 1 − ρ(L ∗ \{0}) 1 − .
x∈L ∗ \{0} x∈L ∗ \{0}
(1.4.8)
( j)
To estimate ĝ j (L ∗ ), assume ρc (x) is the j order partial derivative of ρc (x) about
the first variable x1 , i.e.
∂ j
ρc( j) (x) = ( ) ρc (x).
∂ x1
If j = 1, 2, it is easy to get
It follows that
1 (1)
g1 (x) = − ρ (x).
2π c
1 (2) 1
g2 (x) = ρ (x) + ρc (x).
4π 2 c 2π
( j)
Since ρc (x) = (2πi x1 ) j ρ̂c (x), we have
1
ĝ2 (x) = ( − x12 )ρ̂c (x).
2π
28 1 Random Lattice Theory
|x|2
According to the inequality |x1 | |x|2 e 2 and η (L) 21 ,
|x|2
|ĝ1 (L ∗ )| e−π|x|
2
|x1 | · |ρ̂c (x)| = |x1 |ρ(x) e 2
π
e− 4 |x| = ρ2 (L ∗ \{0}) .
2
(1.4.9)
x∈L ∗ \{0}
|ĝ1 (L ∗ )|
|E[(ξ − c) · u]| = .
|ρ̂c (L ∗ )| 1−
For a general unit vector u ∈ Rn , there exists an orthogonal matrix M ∈ Rn×n such
that Mu = (1, 0, . . . , 0). Denote η as a random variable from the discrete Gauss
distribution D M −1 L ,M −1 c , for any x ∈ L,
−1 −1
ρ M −1 c (M −1 x) e−π|M x−M c|
2
−1
Pr {η = M x} = =
ρ M −1 c (M −1 L) ρ M −1 c (M −1 L)
e−π|x−c|
2
= = Pr {ξ = x} = Pr {M −1 ξ = M −1 x},
ρc (L)
which implies that the distributions of η and M −1 ξ are the same, hence,
Above all the inequality (1.4.3) holds, and inequality (1.4.4) could be proved in the
same way. We complete the proof of Lemma 1.4.1.
and
1
E[|ξ − c|2 ] ( + )s 2 n. (1.4.11)
2π 1−
n
1
E[|ξ − c|2 ] = E[((ξ − c) · u i )2 ] ( + )s 2 n.
i=1
2π 1−
Proof From the proof of Lemma 1.4.1, here we only need to prove for the case
s = 1. Since
√ ρv (x)
Pr {|ξ − v| > n} =
√ ρv (L)
x∈L ,|x−v|> n
√
ρ(x − v) ρ((L − v)\ n N )
= = ,
√ ρv (L) ρv (L)
x∈L ,|x−v|> n
That is,
√ ρ(L)
Pr {|ξ − v| > n} < 2−n . (1.4.13)
ρv (L)
|det(L ∗ )|(1 − |e−2πi x·v ρ(x)|) = |det(L ∗ )|(1 − ρ(x))
x∈L ∗ \{0} x∈L ∗ \{0}
Similarly,
ρ(L) = |ρ(L)| = |det(L ∗ )ρ̂(L ∗ )|
= |det(L ∗ ) ρ(x)| = |det(L ∗ )|(1 + ρ(x))
x∈L ∗ x∈L ∗ \{0}
√ 1 + −n
Pr {|ξ − v| > n} 2 .
1−
min |x − y|.
y∈A
Proof From the proof of Lemma 1.4.1, we only need to prove for the case s = 1.
Let
g(x) = e2πi x·v ρc (x).
By Lemma 1.3.2,
therefore,
|ĝ(L ∗ )| = | ρ(x − v)e−2πi(x−v)·c | ρ(x − v) = ρ(L ∗ − v).
x∈L ∗ x∈L ∗
√
Since dist(v, L ∗ ) n, we know
√
ρ(L ∗ − v) = ρ((L ∗ − v)\ n N ).
Above all,
ĝ(L ∗ ) 1 + −n
|E[e2πiξ ·v ]| = | | 2 .
ρ̂c (L ∗ ) 1−
1.4 Some Properties of Discrete Gauss Distribution 31
1 + −n
|E[cos(2π(ξ + w) · v)]| 2 . (1.4.17)
1−
1 + −n
|E[cos(2π(ξ + w) · v)]| |E[e2πi(ξ +w)·v ]| = |E[e2πiξ ·v ]| 2 .
1−
m √
Proof By Cauchy inequality we get i=1 |z i | m|z|, so
m
E[| z i vi |2 ] = z i z j E[vi · v j ] = z i2 E[|vi |2 ] + z i z j E[vi ] · E[v j ].
i=1 i, j i i= j
(1.4.19)
The first term of the right hand side in (1.4.19) has the estimation
z i2 E[|vi |2 ] z i2 l = l|z|2 .
i i
The second term of the right hand side in (1.4.19) has the estimation
1
z i z j E[vi ] · E[v j ] |z i ||z j | · (|E[vi ]|2 + |E[v j ]|2 )
i= j i= j
2
|z i ||z j | ( |z i |)2 m |z|2 .
i= j i
32 1 Random Lattice Theory
m
E[| z i vi |2 ] (l + m )|z|2 .
i=1
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.
Chapter 2
Reduction Principle of Ajtai
In 1996, the famous scholar Ajtai proposed the reduction principle from the worst
case to the average case at the 28th Summer Symposium of the American Computer
Society (ACM), named the Ajtai reduction principle [see Ajtai (1996), Ajtai (1999)
and Ajtai and Dwork (1997)]. Subsequently, Ajtai and Dwork presented the first
lattice-based cryptosystem, which is called the Ajtai-Dwork cryptosystem in the
academic circles. The proof of this cryptosystem resisting Shor’s quantum computing
is to apply Ajtai reduction principle to transform searching for collision points of
the Hash function into the SIS problem, and Ajtai reduction principle proves that the
difficulty of solving the SIS problem is polynomially equivalent to the shortest vector
problem on lattice. The main purpose of this chapter is to prove the Ajtai reduction
principle.
where x, y, z are random variables on Zqm and Zqn , respectively. This random lin-
ear system plays an important role in modern cryptography. We prove some basic
properties in this section.
1
Pr {x = x0 } = .
qn
1
Pr {y = y0 } = Pr {x = x0 } = .
qn
Remark 2.1.1 In fact, for the above linear system, x and y are random variables with
the same distribution when A is an invertible square matrix. However, this property
doesn’t hold if A is not a square matrix.
Let a ∈ R be a real number, [a] be the greatest integer no more than a, i.e. [a] is
the only integer satisfying the following inequality,
[x] is called the integer vector of x. We say x is a random vector, which means each
element x j is a random variable, and the n random variables are mutually indepen-
dent.
Proof Since all the components of x are independent, we only prove for n = 1.
If a ∈ [0, 1) is a continuous random variable uniformly distributed, then for any
i = 0, 1, . . . , q − 1, we have
i i +1 1
Pr {[qa] = i} = Pr {i qa < i + 1} = Pr { a< }= .
q q q
These nurseries rest on the clay-like framework of the nest, but are not
attached thereto; they in no way support it, or one another, indeed they
have the appearance of being constantly added to on their upper margins
and constantly eaten away on their under parts. Fig. 240 represents the
appearance of the upper boundary of a nursery taken from a nest of
Termes angustatus. The small white bodies, mentioned above, have
disappeared: the mycelium of the fungus, though not shown in the figure,
is still visible on the specimen from which it was drawn, and gives rise to a
whitish, glaucous appearance.
Termitidae are classed by some naturalists with the Orthoptera, and they
have a great deal in common with some of the cursorial division of that
Order, more particularly Forficulidae and Blattidae; but they differ from
Orthoptera in the nature and form of the wings. They are also classed by
some, with a few other forms, as a separate Order of Pseudo-Neuroptera
called Corrodentia, but this is not a very satisfactory course, as the
Termitidae do not agree closely with the forms associated with them,
while the aggregate so formed is far from being very distinct from other
forms of Neuroptera. On the whole the best plan appears to be to treat
the Termitidae as forming a distinct family of the Order Neuroptera, or to
make it a distinct Order, as proposed by Grassi. Packard now associates
Termites in an Order with the biting-lice, and calls it Platyptera.
The Psocidae are without exception small and soft-bodied Insects, and
are only known to those who are not entomologists by the wingless forms
that run about in uninhabited or quiet apartments, and are called dust-lice
or book-lice. They are perhaps more similar to Termitidae than to any
other Insects, but the two families differ much in the structure of their
wings, and are totally dissimilar in the nature of their lives.
The life-history has never been satisfactorily sketched. The young greatly
resemble the old, but have no ocelli or wings, and sometimes the tarsi are
of two joints, while in the adult they have three. The antennae have also
in these cases a less number of joints in the young stage. The food is
animal or vegetable refuse substances; many live on fungoid matter of
various kinds, mouldy chaff being, it is said, a favourite pabulum; the
mould on palings is a source of food to many; others live on the rust-fungi
of leaves, and many frequent the bark of trees. They are able to spin
webs, probably by the aid of the lingual glands; the eggs are deposited, in
some cases, on leaves and covered with a web. Hagen says that a
peculiar organ, possibly a gland—he calls it a hose[309]—exists at the
base of the tarsal claws. In our climate most of the species pass the
winter in the egg-state. There may be two generations in a year, perhaps
more.
In the young the wings first appear as buds, or outgrowths of the sides of
the meso- and meta-thorax; afterwards the prothorax decreases, while
the other two thoracic segments and the wing-rudiments attached to them
increase. The wings from their very origin appear to be different from
those of the Orthoptera, and the changes that take place in the thoracic
segments in the course of the development, differ from those that occur in
Orthoptera.
Fig. 246.—Micropterous form of Mesopsocus unipunctatus. a, a, Wings.
(After Bertkau.)
There are several peculiarities connected with the wings. Frequently they
exist, though of no use for flight; some Psocidae that have perfectly-
formed wings are so reluctant to use them that, M‘Lachlan says, they will
allow themselves to be crushed without seeking to escape by flight. At
certain periods, however, some Psocidae float on the wing in
considerable numbers, especially in a moist still atmosphere, and then
drift about like the winged Aphididae, which are frequently found with
them. There is evidence that individuals, or generations, of some of the
winged species occur with only rudimentary wings; although this has
been denied by Kolbe, there can be no doubt about it. The form figured
above (Fig. 246) was described by Bertkau[312] as a distinct genus, but
was afterwards recognised by him[313] to be a short-winged form of
Mesopsocus unipunctatus. It is probable that the adult female of this
species has the wings always micropterous, while the male has these
organs of the full size. In other species the condition of the rudimentary
wings seems to be quite constant. The facts concerning the wings of
Psocidae are so peculiar that Kolbe came to the conclusion that the
organs exist not because they are of use for flight, so much as because it
is the nature of an Insect to develop wings.[314]
Some of the species of Psocidae have never any trace of wings. These
apterous forms are mostly included in the division Atropinae, and are
usually very minute; it has been again and again erroneously stated that
they are the young state of winged forms. Hagen kept a large colony of
Atropos divinatoria for some years in confinement, so that he saw
numerous generations as well as many specimens. He found the
apterous condition quite constant.
The Rev. W. Derham, who two hundred years ago was Rector of
Upminster, in Essex, and was well known as a distinguished writer and
philosopher, gave an account of the ticking of death-watches to the Royal
Society.[318] This gentleman was a most accurate and minute observer;
he was well acquainted with the ticking of the greater death-watch—
Anobium—which he describes very accurately, as well as the acts
accompanying it, the details he mentions being exactly such as occur at
the present time. He not only heard the ticking of the Psocid or lesser
death-watch, but repeatedly witnessed it. He says: "I am now so used to,
and skilful in the matter as to be able to see, and show them, beating
almost when I please, by having a paper with some of them in it
conveniently placed and imitating their pulsation, which they will readily
answer." He also states that he could only hear them beating when it was
done on paper, and that this death-watch will tick for some hours together
without intermission, with intervals between each beat, so that it much
resembles the ticking of a watch. The act of ticking was accompanied by
rapping the front of the head on the paper, but Mr. Derham could not be
sure that the sound was produced in that manner, because each stroke
was also accompanied by a peculiar shudder, or recoil. After a prolonged
ticking he observed that another individual of the other sex made its
appearance. The species figured by Mr. Derham more resembles a
Hyperetes than it does either of our two known book-lice, Atropos and
Clothilla.
Fig. 248.—The lesser death-watch of Upminster. (After Derham.) A,
magnified; B, natural size.
Fam. V. Perlidae.
Insects of moderate or large size, furnished with four membranous
wings; these are usually complexly reticulate; the hind pair are much
the larger, and have a large anal area of more simple venation, which
becomes plicate when folded. The coxae are small, the legs widely
separated. The larvae are aquatic in habits; the metamorphosis is
slight.
The antennae of the Perlidae are long, very flexible, and composed of a
very large number of joints. The parts of the mouth vary a good deal. The
mandibles and maxillae are usually rather small, and all the parts of the
mouth are of feeble consistence or even membranous; the maxillary palpi
are, however, well developed and exserted from the mouth, five-jointed.
The labium is short and but little conspicuous. The mandibles in some
forms are almost membranous, but in other genera they are firmer and
are toothed. The labium is composed of a very large mentum, beyond
which is a large piece, usually undivided, bearing the four terminal lobes;
the three-jointed palpus is seated on the side of the large middle sclerite,
which is no doubt of composite nature. Considerable variety as to the
lower lip prevails. The head is broad and flat; there is an indistinctly-
indicated clypeus, three—more rarely two—ocelli, and on each side an
eye neither very large nor perfect. The prothorax is free, and has a flat,
margined notum. The meso- and the meta-thorax are large, equal
segments. The pro-, meso-, and meta-sternum are large pieces; between
the first and second, and between the second and third there is an
intervening membrane. The metasternum is much prolonged backwards,
and has on each side a peculiar slit; similar orifices exist on the other
sterna (Fig. 254, o). Newport, who has examined them in Pteronarcys,
says that they are blind invaginations of the integument; he calls them the
sternal or furcal orifices.[321] According to this naturalist these very
peculiar openings pass into the body "as strong bone-like tubes, diverging
from the axis to the periphery of the body in the immediate vicinity of
some of the principal tracheae, but that they do not in any way
communicate with them, as they terminate abruptly as caecal structures."
He thinks them analogous with the endo-skeleton of other Insects; a view
which cannot be considered sufficiently established. Laboulbène
states[322] that when Perla parisina is seized and placed on its back, it
does not move, but emits a liquid at the base of the articulation of the
legs. This suggests that it may come from these sternal orifices. The
abdomen consists of ten dorsal plates, the first being short, and of nine
ventral; the dorsal plates are much more ample transversely than the
ventral. Frequently the hind body is terminated by two long, many-jointed
cerci, looking like antennae. The coxae are small, not prominent, and are
directed outwards. The legs are slender, the tibiae often grooved. The
tarsi are three-jointed, terminating in two claws and a more or less distinct
pad. In the genus Isopteryx an auditory organ has been described as
existing in the legs, in a position similar to that of the analogous structures
in Termitidae and Blattidae. The wings when closed repose flat on the
back, and fold and overlap so that only one is seen (Fig. 251); in this state
the costal portion of each front wing is turned downwards, so as to protect
to some extent, the sides of the body.
Fig. 251.—Perla maxima. (After Pictet.)
The early stages are known, but have not been described minutely, and
there appears to be very little information as to the youngest life. All the
species are, when immature, aquatic in their habits; the larvae greatly
resemble the perfect Insects in form, though differing in not possessing
wings and in the ocelli being merely opaque spaces. They have rather
large compound eyes; the future wings are represented by lobe-like
prolongations—varying in length according to age—of the meso- and
meta-notum. In the Nemourae the cerci are absent in the imago though
present in the young. The larvae of Perlidae are carnivorous and are able
to swim well, the legs being provided with abundant swimming hairs; they,
however, as a rule, prefer to walk at the bottom of the pool, or on rocks or
boulders in the water they live in.
Pictet, Dufour, Newport, and Imhof[326] have studied the internal anatomy.
The alimentary canal is remarkable for the enormous oesophagus; there
is no distinction between this and the crop. A proventriculus is quite