Supersingular isogeny graph

A supersingular isogeny graph is determined by choosing a large prime number ${\displaystyle p}$ and a small prime number ${\displaystyle \ell }$, and considering the class of all supersingular elliptic curves defined over the finite field ${\displaystyle \mathbb {F} _{p^{2}}}$. There are approximately ${\displaystyle (p+1)/12}$ such curves, each two of which can be related by isogenies. The vertices in the supersingular isogeny graph represent these curves (or more concretely, their j-invariants, elements of ${\displaystyle \mathbb {F} _{p^{2}}}$) and the edges represent isogenies of degree ${\displaystyle \ell }$ between two curves.[1][2][3]

The supersingular isogeny graphs are ${\displaystyle \ell +1}$-regular graphs, meaning that each vertex has exactly ${\displaystyle \ell +1}$ neighbors. They were proven by Pizer to be Ramanujan graphs, graphs with optimal expansion properties for their degree.[1][2][4][5] The proof is based on Pierre Deligne's proof of the Ramanujan–Petersson conjecture.[4]

One proposal for a cryptographic hash function involves starting from a fixed vertex of a supersingular isogeny graph, using the bits of the binary representation of an input value to determine a sequence of edges to follow in a walk in the graph, and using the identity of the vertex reached at the end of the walk as the hash value for the input. The security of the proposed hashing scheme rests on the assumption that it is difficult to find paths in this graph that connect arbitrary pairs of vertices.[1]

It has also been proposed to use walks in two supersingular isogeny graphs with the same vertex set but different edge sets (defined using different choices of the ${\displaystyle \ell }$ parameter) to develop a key exchange primitive analogous to Diffie–Hellman key exchange, called supersingular isogeny key exchange.[2]

Additional cryptographic methods based on these graphs include signature schemes and public-key cryptography. They have been suggested as a form of post-quantum cryptography: As of 2018, there are no known subexponential-time methods for breaking these schemes, even on quantum computers.[6]