Score:1

Non probabilistic algorithm : Given secret key $d$ we can factorize $n$ assuming $e$ is small

vi flag

I read in an introduction to a paper that if $e$ is small enough and we were given secret key $d$ in RSA, then there is an efficient deterministic algorithm to factorize $n$. I've searched about that and I've found the probabilistic one: Algorithm to factorize $N$ given $N$, $e$, $d$

I guess, the fact that $e$ is small must play some role here. But I was able to come up with something. Do you know anything?

Daniel S avatar
ru flag
See [this paper](https://iacr.org/archive/crypto2004/31520213/det.pdf) by Alexander May.
fgrieu avatar
ng flag
Or Jean-Sebastien Coron and Alexander May's [Deterministic Polynomial Time Equivalence of Computing the RSA Secret Key and Factoring](https://eprint.iacr.org/2004/208), in [JoC 2007](https://doi.org/10.1007/s00145-006-0433-6), which is later, and may have some refinements.
Score:1
pe flag

The May-Coron result for balanced RSA moduli can be seen as essentially an application of the Coppersmith theorem to the polynomial $$ f(x) = n - x \bmod (de-1)\,, $$ where we are looking for the root $p+q-1 \approx n^{1/2}$ modulo $\varphi(n) \approx n$—a divisor of $de-1$.

The Coppersmith theorem says that we can find a root of $f(x)$ smaller than $(de-1)^{\beta^2}$ modulo an unknown divisor $\varphi(n)$ of $de-1$ of size $(de-1)^{\beta}$. Translating this to our setting, suppose $de-1 = n^{k}$. Then $n \approx (de-1)^{1/k}$. Setting $\beta=1/k$ and taking logarithms, we have that $k\cdot (1/k)^2 \ge 1/2$ implies $k \le 2$, which means that the factorization can be found in polynomial time as long as $de-1 \le n^2$. As such, it is not strictly $e$ that needs to be sufficiently small, but the product $ed$.

tonythestark avatar
vi flag
maybe you meant to say $f(x) = x^2 - n$ ? because that's when $ n^{1/2}$ is a root
tonythestark avatar
vi flag
Why we assume $p+q-1 \approx n^{1/2} \mod \phi(n)$? We have $\phi(n) = (p-1)(q-1)$ $p+q-1= (p-1) + (q-1) + 1 \mod \phi(n) = 1$
pe flag
$n - (p + q - 1) \equiv 0 \pmod{\varphi(n)}$. When I use $\approx$ above, it means "roughly of the same size".
mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.