FACTORING CRYPTOSYSTEM MODULI WHEN THE CO-FACTORS DIFFERENCE IS BOUNDED

International Journal of Information Technology, Modeling and Computing (IJITMC) Vol.1, No.3,August 2013
DOI : 10.5121/ijitmc.2013.1301 1
FACTORING CRYPTOSYSTEM MODULI WHEN THE
CO-FACTORS DIFFERENCE IS BOUNDED
Omar Akchiche1
and Omar Khadir2
1,2
Laboratory of Mathematics, Cryptography and Mechanics, Fstm, University of Hassan
II Mohammedia-Casablanca, Morocco
omar.akchiche@hotmail.fr
khadir@hotmail.com
ABSTRACT
LET P, Q BE TWO LARGE PRIMES AND N= . WE SHOW, IN THIS PAPER, THAT IF | − | ≤ 2 WHERE IS
THE BIT-SIZE OF AND ∈ ℕ, THEN WE CAN COMPUTE EFFICIENTLY THE INTEGERS AND IN AT MOST
COMPARISONS. OUR APPROACH CAN BE USED TO BUILD ATTACKS AGAINST RSA OR RABIN CRYPTOSYSTEMS.
KEYWORDS
Integer factorization problem, RSA, Rabin cryptosystem, Public key cryptography.
1. INTRODUCTION
Factoring large integers is a central issue in cryptography. No efficient deterministic algorithm is
known. Widely used cryptographic protocols like RSA [1], Rabin [2] system or Saryazdi [3]
digital signature rely on this fact.
In many cryptosystems, each user must randomly choose two large prime numbers and to
produce his own keys. These integers have to be sufficiently large to ensure that it is not
computationally possible for anyone to factor the modulus = . Generally, the running time
for generating primes takes the most important part in the total running time. Menezes and all. [4,
p. 133] give several algorithms for prime number generation and primality testing. Also, in [5],
the authors made experimental tests and concluded by suggesting some rapid procedures.
In literature, there exist various integer factorization methods, but they are not efficient. The
oldest and simplest one is the trial divisions. Fermat [6, p. 143] proposed a technique for factoring
integers that are product of two primes which are close to one another. The continued fraction
algorithm [7] was developed in 1931. Some decades later, John Pollard conceived his − 1 and
methods [8,9] respectively in 1974 and 1975. At the end of 1970s, with the advent of the public
key cryptography [10,1,2], integer factorization problem becomes of crucial importance.
Numerous papers, proposing ingenious and sophisticated methods, were published. Pomerance
[11] discovered the quadratic sieve algorithm in 1984. Less than two years later, Lenstra [12]
suggested to factor large numbers by means of finite elliptic curves. Today, the fastest known
algorithm is the General Number Field Sieve (GNFS) [13, p. 103]. It was invented by Pollard in
1988 and allows to factor natural integers with more than 110 digits. Stinson [14, p. 232] has

2
evoked the possibility of factoring the RSA modulus if the two factors are too close. In 1999,
Boneh and al. [15] described a polynomial time algorithm for factoring = when the
exponent is large. Some years later, in 2007, Coron and May [16] presented the first
deterministic algorithm for factoring the RSA modulus in polynomial time, but they used the
public and the secret key pair ( , ).
Our work is devoted to present original results related to integer factorization problem. Indeed,
we improve many statements established in two previous articles [17,18]. Furthermore, our
approach can be used to build attacks against cryptosystems like RSA [1], Rabin [2] or Saryazdi
digital signature [3].
The paper is organised as follows. In section 2, we recall the main facts stated in papers [17] and
[18]. In section 3, we present our own results and we conclude in section 4.
Throughout all the sequel, we will use standard notations. In particular ℕ is the set of all natural
integers 0,1,2,3, … and ℕ∗
= ℕ − {0}. The largest integer which does not exceed the real is
denoted by ⌊ ⌋. It is also the integer part and the floor of . Thus we have ⌊ ⌋ ≤ < ⌊ ⌋ + 1. The
bit-size of a positive integer is the number of bits in its binary representation. So, the bit-size of
is ⟺ = ∑ 2 with every ∈ {0,1} and = 1. Moreover, the bit-size of
satisfies the relation2 ≤ < 2 . Two positive integers and are said to be co-factors of
if = .
We start by recalling some known results.
2. PRELIMINARIES
In this section, we recall results established in 2008 and 2010 and described in papers [17] and
[18]. More precisely, we review sufficient conditions under which one can factor = where
, are co-factors not necessary prime but close to each other. For the sake of completeness, we
give all the proofs. First we present a statement published in [17]. Before that, we need the
following lemma.
Lemma 1. [17] Let < be two elements of ℕ∗
and let , denotes the number of perfect
squares such that < ≤ . Then we have: , <
√ √
+ 1.
Proof. Consider the set = { ∈ ℕ | ≤ }. Since is also ∈ ℕ | ≤ √ , its
cardinality is √ + 1, and then , = √ − √ . If we put = √ and = √ which
means that ≤ √ < + 1 and ≤ √ < + 1, we obtain ≤ √ and – < 1 − √ . Hence
, = − < √ − √ + 1 =
√ √
+ 1.
□
The following theorem was one of the main results in paper [17].
Theorem 1. [17] Let ∈ ℕ be a composite integer whose bit-size is . If its two prime factors
and satisfy the inequality:
| − | ≤ 2 (1)
then we can compute them efficiently.

3
Proof. Without loss of generality, we assume that 2 < < . As the prime factors and are
odd, we put = + 2 where ∈ ℕ. Since = , + = ( + ) . By relation (1), | −
| ≤ 2 . It follows that = ≤ 2 and then ≤ 2 . Let = + where > 0. We
have > . By Lemma 1, the number , of perfect squares between and satisfies the
inequality , <
√ √
+ 1. We deduce that , <
√
+ 1. The bit-size of is , so
≥ 2 . Thus 2√ ≥ 2 . Therefore , < + 1 = 2. However , is an integer, so
, = 1 This means that the only perfect squares between and is = + = ( + ) .
But the first perfect square greater than is = ( √ + 1) . This allows us to compute the
factors and . Indeed, since − = is a perfect square, = ( − )( + ). Then, we
have = − and = + . Hence = √ − + 1 and = √ + + 1.
□
Now we move to results published in paper [18]. But first, we give two definitions.
Definition 1. [18] Let ∈ ℕ be a composite integer. The minimal distance ( ) of is the
smallest distance between its co-factors:
( ) = {| − |, , ∈ ℕ, = } (2)
Next fact is easy to prove.
Theorem 2. [18] For every composite integer ∈ ℕ, there exists a unique couple ( , ) of
divisors of such that = and ( ) = | − |.
Definition 2. [18] Let ∈ ℕ be a composite integer. The unique couple of positive integers ( , )
such that = and ( ) = | − | is called the weak decomposition of .
For all the proofs below, we denote by = the odd natural integer to be factorized where
( , ) is its weak decomposition and ( ) is its minimal distance as stated in relation (2). For
simplicity, we assume that 2 < < . We also define = + such that − = ( ) = 2 ,
∈ ℕ∗
since is odd. According to previous notations, it is readily seen that = ( + ) .
Observe that as > 0, > . Furthermore, we let , be the number of perfect squares
between and .
The next theorem was the main result in [18]. In the sequel, the term comparison means the
operation consisting of checking if a given natural integer is a perfect square or not.
Theorem 3. [18] For every composite odd integer ∈ ℕ whose bit-size is , if there exists a
number such that the minimal distance of verifies:
( ) ≤ 2 (3)
then one can find the weak decomposition of in at most 2 comparisons.
Proof. By Lemma 1, since , is the number of perfect squares such that < ≤ ,
, <
√ √
+ 1 <
√
+ 1. On the other hand, by relation (3) ( ) ≤ 2 , so ≤ 2 .
As is the bit-size of , ≥ 2 . We obtain , <
.
+ 1 < 2 + 1. Since , is a

4
natural integer, it is clear that , ≤ 2 . The only 2 perfect squares immediately greater than
are ( √ + ) , 1 ≤ ≤ 2 . As + is a perfect square between and , there exists
necessary an integer ∈ {1,2, … , 2 } such that:
+ = ( √ + )
This relationship implies that: = √ + − ( √ + + ). If we put =
√ + − and = √ + + , then − = 2 = ( ). Therefore the couple
( , ) is a weak decomposition of . However, by Theorem 2, = and = which ends
the computation of the two factors and .
We find by making at most 2 comparisons since ∈ {1,2, … , 2 }.
□
In the following section, we improve and extend the main results established in papers [17] and
[18].
3. OUR RESULTS
In this section, we present our main contribution by extending Theorem 1. First, we need a
slightly modified version of Lemma 1.
Lemma 2. If < are two elements of ℕ∗
and if , denotes the number of perfect squares
such that < ≤ . Then we have:
√ √
− 1 < , <
√ √
+ 1 (4)
Proof. The right inequality in relation (4) was proved in Lemma 1. It is not difficult to see that
, = √ − √ . If we put = √ and = √ which means that ≤ √ < + 1 and
≤ √ < + 1, we obtain > √ − 1 and − ≥ −√ . Hence , = − > √ − √ −
1 =
√ √
− 1which ends the proof. Note that we also have | , −
√ √
| < 1.
□
Throughout the sequel, we make use of the following lemma.
Lemma 3. Let be a composite odd integer. The first natural integer such that √ + −
is a perfect square is exactly , . Moreover, if one can compute , , then it is possible to find
the weak decomposition of .
Proof. The number of perfect squares between and is , . Thus, the perfect squares such
that < ≤ have the form √ + , ∈ 1,2, … , , . The largest one is √ +
, . Hence, we must have = √ + , . Therefore √ + , − is a perfect
square . The first assertion is then proved. In order to justify the second one, assume that we
know , , Since + = √ + , , we deduce that = √ + , − ( √ +
, + ). Let = √ + , − and = √ + , + . As − = 2 = ( ), the
couple ( , ) is a weak decomposition of . Consequently, by Theorem 2, = and =
which ends the computation of the two co-factors of .
□

5
Now, we give our first main result: an extension of Theorem 1.
Theorem 4. For every composite odd integer ∈ ℕ whose bit-size is , if there exists a positive
number such that the minimal distance of verifies:
( ) ≤ 2 (5)
then one can find the weak decomposition of in at most comparisons.
Proof. As ( ) ≤ 2 , =
( )
= 2 , and then ≤ 2 . Since by Lemma 1, , <
√
+ 1, it follows that , < + 1. Therefore , < + 1. But , is a natural
integer, so , ≤ . By Lemma 3, in order to compute , , one must determine the first
positive integer , ≤ , such that √ + − is a perfect square. Thus, for this purpose,
we need at most comparisons. By the same Lemma 3, once , is known, we are able to find
the weak decomposition of .
□
Remark 1. Our result is also an improvement of Theorem 3 from paper [18] by taking = 2 ,
∈ ℕ∗
.
Theorem 4 leads to the following efficient algorithm where comments are delimited with braces.
Algorithm:
Input: A composite odd positive integer .
Output: The weak decomposition ( , ) of .
1. input( ); { is the composite natural integer to be factored}
2. ⟵ 1; {We initialise the value of in relation (5)}
3. ⟵ 0; {The program ends when flag becomes 1}
4. while = 0 do
4.1. ⟵ ( − 1) + 1; {We initialise the value of . We look for such that √ + − is
a perfect square}
4.2. while ≤ and = 0 do
4.2.1 ⟵ √ + ;
4.2.2 ⟵ − ;
4.2.3 if is a perfect square then {Here = √ + − }
4.2.3.1 ⟵ √ ;
4.2.3.2 ⟵ − ; { is the first co-factor of }
4.2.3.3 ⟵ + ; { is the second co-factor of }
4.2.3.4 ⟵ 1; {We stop the program since is factored by previous instructions}
4.2.3.5 output ( , ); {The algorithm computes the weak decomposition of }
4.2.4 ⟵ + 1; {We increment }
4.3 ⟵ + 1; {We enlarge the coefficient such that | − | ≤ 2 }
Next corollary improves Theorem 4 if more information is known.

6
Corollary 1. For every composite odd integer ∈ ℕ whose bit-size is , if is the smallest
positive integer such that:
( ) ≤ 2 (6)
with < 2 , then we can determine the weak decomposition of in at most −
( )
√
+ 1
comparisons.
Proof. We proved in Theorem 4 that , < , let us here show that , ≥
( )
√
. The bit-
size of is , so ≥ 2 and then 2√ ≥ 2 . By hypothesis (6) =
( )
≤ 2 . We
assumed that < 2 , so 2 < 2 . Therefore < 2√ . The fact that is the bit-size of
leads to < 2 and then 2√ < 2 . This implies that + 2√ < 2. 2 = 2 . So we obtain
√
> . Recall that is the smallest integer which satisfies inequality (6). That means
( ) > ( − 1)2 . As =
( )
, > ( − 1)2 and then > ( − 1) 2 . By Lemma 2,
, >
√
− 1. Hence , >
( )
− 1 =
( )
√
− 1. Since , is an integer, we must
have , ≥
( )
√
. We proved that
( )
√
≤ , ≤ . By Lemma 3, to compute , , we
have to determine the first integer such that √ + − is a perfect square. We are sure
that
( )
√
≤ ≤ . It is clear that we will need at most −
( )
√
+ 1 comparisons.
Then, by the same Lemma 3, knowing , , we can find the weak decomposition of .
□
Example 1. Let us take n = 84009841 as in paper [19]. With the help of Maple software, the
first positive integer for which √ + − is a perfect square is j = 370.Therefore the two
factors of are p = 6907 and q = 12163. On another hand, the first natural integer such that
( ) ≤ 2 is N = 21, and we check that is belonging to the interval
( )
√
, .
In the next result, we improve Theorem 3 under certain assumptions. But, first we need the
following theorem.
Theorem 5. For every composite odd integer ∈ ℕ whose bit-size is , if we can find a natural
integer and a positive real such that:
2 < ( ) ≤ 2 (7)
with 2 < + 1, then one can find the weak decomposition of in at most 2 − ⌊2 ⌋ + 1
comparisons.
Proof. In the proof of Theorem 3, we have seen that , ≤ 2 . Now we show that , ≥ ⌊2 ⌋.
As is bit-size of , ≥ 2 and then 2√ ≥ 2 . Since ( ) ≤ 2 , =
( )
≤ 2 .
By hypothesis 2 < + 1, thus 2 < 2 . Therefore < 2√ . Moreover, is the bit-size of
, so 2√ < 2 . It follows that + 2√ < 2. 2 < 2 . We then get
√
> . Since, by

7
relation (7), ( ) > 2 , we have =
( )
> 2 . So, we deduce that > 2 . By
Lemma 2, , >
√ √
− 1. Therefore , >
√
−1. Finally, we obtain , > −
1 > 2 − 1. But , is a integer, so , ≥ ⌊2 ⌋. We proved that ⌊2 ⌋ ≤ , ≤ 2 . By Lemma
3, the integer , verifies that √ + , − is a perfect square. That means that in order
to compute , , one must checks if √ + − is a perfect square for such that ⌊2 ⌋ ≤
≤ 2 . Obviously, we need at most 2 − ⌊2 ⌋ + 1 comparisons. Then, by the same Lemma 3,
once we know , , we can find the weak decomposition of .
□
The following corollary, our second main result, slightly improves Theorem 3 [18]. It reduces the
number of comparisons to perform.
Corollary 2. For every composite odd integer ∈ ℕ whose bit-size is , if we can find the
smallest positive integer such that:
( ) ≤ 2 (8)
With 2 < + 1, then one can find the weak decomposition of in at most 2 − 2 + 1
comparisons.
Proof. We share the value of the exponent of the first term in relation (7) into two part: 2 + 8 +
2x = x + 5 + (3 + 2x). Since is the smallest positive integer that satisfies the inequality (8),
( ) > 2
( )
. Therefore, in order to apply Theorem 5, we must have 3 + 2x = 2(l − 1). In
other words, = − . Hence, the assertion is proved.
□
The following theorem enlarges the bound in relation (3) without adding much cost.
Theorem 6. For every composite odd integer ∈ ℕ whose bit-size is , if there exists a positive
integer such that the minimal distance of verifies:
( ) ≤ 2 +2 (9)
with 2 < + 5, then one can find the weak decomposition of in at most 2 + 3 comparisons.
Proof. For simplicity, let = 2 . So the hypothesis (9) becomes ( ) < 2 + . As
=
( )
, ≤ 2 + . Therefore, it is clear that ≤ 2 + + 2 . The bit-size of
is , so
√
< . Since, by Lemma 1, , <
√
+ 1, we have:
, <
2
2
+
4. 2
+
2
2
+ 1
If we substitute with its value, then we get , < 2 + + 2 + 1. Recall that , is an
integer, so , ≤ 2 + 3. Consequently, by Lemma 3, in order to find a natural integer such

8
that √ + − is a perfect square, we need at most 2 + 3 comparisons. By the same
Lemma 3, it is possible to determine the weak decomposition of since = , , which ends the
proof.
□
4. CONCLUSION
In this paper, we presented new conditions under which one can factor a large composite integer
and we gave an extension of previous results published in [17, 18]. Our work can be applied to
build attacks against several cryptosystems based on the hardness of integer factorization like
RSA [1] and Rabin [2] algorithms or like Saryazdi digital signature [3]. We recommend to
cryptosystem designers to avoid all these insecure composite modulus mentioned in our results.
REFERENCES
[1] R. RIVEST & A. SHAMIR & L. ADELEMAN, (1978) “A METHOD FOR OBTAINING DIGITAL SIGNATURES
AND PUBLIC KEY CRYPTOSYSTEMS”, COMMUNICATIONS OF THE ACM, VOL. 21, PP. 120-126. .
[2] M. O. RABIN, (1979) “DIGITALIZED SIGNATURES AND PUBLIC KEY FUNCTION AS INTRACTABLE AS
FACTORING”, MIT/LCS/TR, VOL. 212.
[3] S. SARYAZDI, (1990) “AN EXTENSION TO ELGAMAL PUBLIC KEY CRYPTOSYSTEM WITH A NEW
SIGNATURE SCHEME”, PROCEEDINGS OF THE 1990 BILKENT INTERNATIONAL CONFERENCE ON NEW
TRENDS IN COMMUNICATION, CONTROL, AND SIGNAL PROCESSING, NORTH HOLLAND: ELSEVIER
SCIENCE PUBLISHERS, PP. 195-198.
[4] A. J. MENEZES & P. C. VAN OORSCHOT & S. A. VANSTONE, (1997) “HANDBOOK OF APPLIED
CRYPTOGRAPHY”, CRC PRESS, BOCA RATON, FLORIDA.[5]O. KHADIR & L. SZALAY, (2009)
“EXPERIMENTAL RESULTS ON PROBABLE PRIMALITY”, ACTA UNIV. SAPIENTIAE, MATH., 1, NO. 2, PP.
161-168.AVAILABLE AT HTTP://WWW.EMIS.DE/JOURNALS/AUSM/C1-2/MATH2-6.PDF
[6] N. KOBLIZ, (1994) “A COURSE IN NUMBER THEORY AND CRYPTOGRAPHY”, GRADUATE TEXTS IN
MATHEMATICS, SPRINGER-VERLAG, NEW YORK.
[7] D. H. LEHMER & R. E. POWERS, (1931) “ON FACTORING LARGE NUMBERS”, BULL. AMER. MATH. SOC.
VOL. 37 (10), PP. 770-776.
[8] J. M. POLLARD, (1974) “THEOREMS OF FACTORIZATION AND PRIMALITY TESTING”, PROCEEDINGS OF
CAMBRIDGE PHILOSOPHICAL SOCIETY, VOL. 76, PP. 521-528.
[9] J. M. POLLARD, (1975) “A MONTE CARLO METHOD FOR FACTORIZATION”, BIT, VOL. 15, PP. 331-334.
[10] W. DIFFIE & M. E. HELLMAN, (1976) “NEW DIRECTIONS IN CRYPTOGRAPHY”, IEEE TRANSACTIONS
ON INFORMATION THEORY, VOL. IT-22, PP. 644-654.
[11] C. POMERANCE, (1985) “THE QUADRATIC SIEVE FACTORING ALGORITHM”, IN PROC. OF
EUROCRYPT'84, WORKSHOP ON ADVANCES IN CRYPTOLOGY: THEORY AND APPLICATION OF
CRYPTOGRAPHIC TECHNIQUES, SPRINGER-VERLAG NEW YORK, PP. 169-182.
[12] H. W. LENSTRA, (1987) “FACTORING INTEGERS WITH ELLIPTIC CURVES”, ANNALS OF MATHEMATICS,
VOL. 126 (2), PP. 649-673.
[13] A. K. LENSTRA & H. W. LENSTRA, JR., (1993) “THE DEVELOPMENT OF THE NUMBER FIELD SIEVE”,
SERIES LECTURE NOTES IN MATHEMATICS, VOL. 1554, SPRINGER-VERLAG.
[14] D. R. STINSON, (2006) “CRYPTOGRAPHY, THEORY AND PRACTICE”, 3RD EDITION, CHAPMAN &
HALL/CRC.
[15] D. BONEH & G. DURFEE & N. A. HOWGRAVE-GRAHAM, (1999) “FACTORING N=P^R Q FOR LARGE R”,
IN PROCEEDING OF CRYPTO'99, LNCS, VOL. 1666, SPRING-VERLAG, BERLIN, PP. 326-337.
[16] J. S. CORON & A. MAY, (2007) “DETERMINISTIC POLYNOMIAL-TIME EQUIVALENCE OF COMPUTING THE
RSA SECRET KEY AND FACTORING”, JOURNAL OF CRYPTOLOGY, VOL. 20, PP.39-50.
[17] O. KHADIR, (2008) “ALGORITHM FOR FACTORING SOME RSA AND RABIN MODULI”, J. DISCRETE
MATH. SCI. CRYPTOGRAPHY, VOL. 11 (5), PP. 537-543.
[18] O. KHADIR & L. SZALAY, (2010) “SUFFICIENT CONDITIONS FOR FACTORING A CLASS OF LARGE
INTEGERS”, J.. INTERDISCIP. MATH., VOL. 13, NO. 1, PP. 95-103.
[19] J. MCKEE, (1999) “SPEEDING FERMAT'S FACTORING METHOD”, MATHEMATICS OF COMPUTATION,
VOL.68, PP. 1729-1737.

FACTORING CRYPTOSYSTEM MODULI WHEN THE CO-FACTORS DIFFERENCE IS BOUNDED

More Related Content

What's hot (20)

Viewers also liked (17)

Similar to FACTORING CRYPTOSYSTEM MODULI WHEN THE CO-FACTORS DIFFERENCE IS BOUNDED (20)

Recently uploaded (20)

FACTORING CRYPTOSYSTEM MODULI WHEN THE CO-FACTORS DIFFERENCE IS BOUNDED