11.07.2014 Views

Cryptanalysis of RSA Factorization - Library(ISI Kolkata) - Indian ...

Cryptanalysis of RSA Factorization - Library(ISI Kolkata) - Indian ...

Cryptanalysis of RSA Factorization - Library(ISI Kolkata) - Indian ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

141 7.6 Improved Results for Larger k<br />

by the LLL algorithm”, the complexity becomes poly{loga,k}. This happens in<br />

practice as observed in [86] too.<br />

In Table 7.5, we present a comparison <strong>of</strong> our experimental results with those<br />

in [86, Table 1, Section 6.2]. One may note that both our results and the results<br />

<strong>of</strong> [86] are <strong>of</strong> similar quality. We have implemented the method <strong>of</strong> [86] for<br />

comparison, and the data is presented in Table 7.5.<br />

α k Theoretical bound Results <strong>of</strong> [86] Our results<br />

(same for both) Experiments Time in seconds Experiments Time in seconds<br />

0.25 3 375 377 < 1 376 < 1<br />

0.35 10 389 391 < 1 390 < 1<br />

0.40 100 405 408 50.36 407 28.21<br />

0.44 50 449 452 7.09 451 4.04<br />

0.48 100 485 492 68.88 488 36.36<br />

Table 7.5: For 1000 bit N, theoretical and experimental data <strong>of</strong> the number <strong>of</strong><br />

shared LSBs in [86] and shared LSBs in our case.<br />

As we have already discussed, in the approach <strong>of</strong> [40], the number <strong>of</strong> shared<br />

MSBs should be greater than or equal to k<br />

k−1 αlog 2N +6 for k ≥ 3. In our case,<br />

putting the upper bound <strong>of</strong> β, number <strong>of</strong> shared bits should be greater than<br />

( ( ))<br />

2k −1<br />

1−α− 1−<br />

k −1 α log 2 N = k<br />

k −1 αlog 2N.<br />

We have implemented the method <strong>of</strong> [40] for comparison with our strategy. Note<br />

that the data presented in Table 7.6 match with those in [40, Tables 4, 5].<br />

Advantages <strong>of</strong> our approach over [40] are as follows.<br />

1. Our theoretical result in this section is slightly better than that <strong>of</strong> [40] in<br />

terms <strong>of</strong> number <strong>of</strong> shared MSBs.<br />

2. In this section, the matrix corresponding to the lattice is a square one, but it<br />

is rectangular in the method <strong>of</strong> [40]. Hence, the calculation <strong>of</strong> determinant<br />

for the lattices is easier in our method.<br />

3. In the presence <strong>of</strong> k many <strong>RSA</strong> moduli, we have to reduce a k ×k matrix,<br />

whereas the size <strong>of</strong> the matrix in [40] is k × 1 k(k + 1). Hence in practical<br />

2<br />

circumstances, the matrix reduction step in case <strong>of</strong> [40] takes more time than<br />

ours. Further, fromtheexperimentalresultspresentedinTable7.6, itisclear<br />

that our strategy requires much less time than the method <strong>of</strong> [40].

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!