Skip to content

Isogeny Atlas

Supersingular Isogeny Graph · eprint 2026/1431

Renders a real supersingular isogeny graph over GF(431²) — computed live in your browser from the modular polynomials — and walks the seven foremost open problems of isogeny-based cryptography as paths, cycles, and endomorphisms on that graph.

Exhibit 1

The atlas, computed — not drawn

Each dot is a supersingular elliptic curve, each line an isogeny between two of them — all computed just now, in your browser, by factoring the modular polynomial Φ_ℓ(j, Y) over GF(431²). Tap, click, or focus a vertex — or use the Inspect menu — to see its curve. Squares are the GF(431) "spine" (problem six).

Real math, toy size: p = 431, so the whole graph fits on screen; at cryptographic sizes (p ≈ 2²⁵⁶) it would have ~2²⁵² vertices. Not production crypto — a teaching demo.

Edge set
  • curve (j ∈ GF(431²))
  • spine curve (j ∈ GF(431))
  • start
  • target

Select a vertex (tap, click, or use the menu) to inspect its curve.

How the graph is computed (and how it checks itself)

Vertices are found by breadth-first search: the ℓ-isogeny neighbors of a vertex j are the roots of the classical modular polynomial Φ_ℓ(j, Y) in GF(431²), found by square-free reduction and Cantor–Zassenhaus splitting — real polynomial factorization, seeded so every build is identical. The result self-checks against classical theorems, live:

    A separate road to the same answer: the test suite rebuilds every vertex as an actual curve y² = x³ + ax + b and verifies with point arithmetic that its Frobenius trace is supersingular. Raw adjacency (ℓ = 2), straight from the computation:

    What is this?

    An elliptic curve can be boiled down to a single number, its j-invariant. An isogeny is a structure-preserving map between two curves. Put a dot for every special ("supersingular") curve and a line for every small isogeny between them, and you get the graph above — a tangled maze with no discernible pattern. The central hard problem of isogeny-based cryptography is embarrassingly simple to state: given two dots, find a path between them. At real-world sizes the maze has about 2²⁵² rooms (the vertex count is ≈ p/12, with p ≈ 2²⁵⁶) and nobody knows how to navigate it — and the field's frontier is a short list of precisely-stated open questions about this maze, which this page walks through one by one.

    What this isn't: not a key-exchange demo, not an attack lab, and not cryptographic-size — this is a real graph over the small prime p = 431, honestly labelled. For the SIDH key exchange, its 2022 collapse, and a working toy CSIDH, see crypto-lab-isogeny-gate. Not production crypto — a teaching demo.

    Exhibit 2

    Find the path — feel the hard problem

    Pick two curves and try to walk from one to the other by clicking neighbors — you are now doing, by hand, the exact task whose difficulty protects isogeny-based signatures. Then let breadth-first search do it and watch how much of the graph it has to flood: that flood is what costs ~√p steps, hopeless when p has 256 bits.

    Choose a start and target, then walk yourself or run the search.

    Why this stops working at real sizes

    This graph has 37 vertices because p = 431 and the vertex count is always ⌊p/12⌋ + ε. The graph is an expander: short paths exist between any two vertices (here, at most 7 steps) — the hardness is purely about finding them. The best known generic attack meets in the middle at cost about √p; quantum algorithms improve on that somewhat, but no known attack is efficient:

    How the graph and attack costs scale with the prime
    Prime p Vertices (~p/12) Meet-in-middle (~√p) Feasible?
    431 (this page) 37 ~2⁴ work instant — you just did it
    ~2⁶⁴ ~2⁶⁰ ~2³² work minutes on a laptop
    ~2²⁵⁶ (SQIsign) ~2²⁵² ~2¹²⁸ work beyond any conceivable computer

    Costs are the standard generic estimates; scheme-specific security also depends on the problems in Exhibit 3.

    Exhibit 3

    The seven open problems, drawn on the graph

    The frontier of isogeny-based cryptography is not one equation — it is a short list of precisely-stated open questions about this graph. Each one below is shown as a concrete structure: a path, a cycle, a walk, a subset of vertices. Newcomer text first; expand "The precise problem" when you want the sharp statement.

    The seven-problem framing follows Castryck, De Feo, Galbraith, Kutas, Reijnders and Wesolowski, "The Isogeny Problems" (ePrint 2026/1431), which collects expert write-ups of the field's foremost unsolved problems. The wording on this page is a teaching paraphrase of these long-studied problems — read the paper for the experts' precise statements.