issikebrokenyet / issikebrokenyet.github.io Goto Github PK

https://eprint.iacr.org/2022/1019

The current JS waits for the change event on checkboxes and then toggles classes. This works fine when starting from a blank state, but if I toggle a box and soft-reload the page (ctrl+R), the box stays checked but the classes get reset, so now the logic is inverted.

delfs-galbraith:
  name:
    short: DG
    long: Delfs-Galbraith
  complexity: exp(1/2)
  references:
    DG13: "https://arxiv.org/abs/1310.7789"
  comment: >-
    $\exp(1/4)$ reduction of the supersingular isogeny path problem to
    the vectorization problem for supsersingular curves over
    $\mathbb{F}_p$.

Pretty sure complexity: exp(1/2) should be complexity: exp(1/4)? (from the paper's abstract) but wanted to double check before editing it, as I wasn't sure if this was the specific attack for curves over F_p, or the full attack which first looks for curves over F_p then performs the e^1/4 attack which i think still has complexity e^1/2?

https://eprint.iacr.org/2021/1583

some schemes and assumptions have the same id, leading to id clashes in the html. The simplest solution may be to prefix ids, so that we're fine as long as ids are unique within a .yml file.

SeaSign and CSI-FiSh

Table column width is computed from the contents of the visible rows. This has the side effect that when some hidden rows are toggled the layout is recomputed: the width of the columns may change, and there may be a visible jump in vertical scrolling.

Using a combination of height: 0 and overflow: hidden should lead to more predictable results, however table rows do not like having their height set, so this may require a bit of fiddling.

Maybe with a modal on mouse over the attack column?

It could be useful to have a finer hierarchy than just poly < subexp < exp. For example, both Delfs-Galbraith and Biasse-Jao-Sankar break the isogeny path problem in quantum exponential time, but the latter is preferred (and so far not showing up in the row, because of the ordering in the .yml file).

For those who are coming to the site without visiting GitHub, maybe it would help to have the "read me" as an "about page".

As the site stands it might not be immediately obvious what the purpose is.

Add the basic meta stuff: Twitter/Facebook meta tags, SEO, etc.

I suppose we should put them somewhere. Footer?

I have not looked at OSIDH in details, but presenting an identical classical and quantum security level is a bit surprising. From what I understood from [DD21], the cost of the classical attack comes from lattice sieving, thus using a quantum sieving algorithm would directly give a better attack.
A few quantum sieving algorithms have been proposed, to the best of my knowledge the smallest claimed exponent is 0.2563 in https://eprint.iacr.org/2022/676 by me, Chailloux, Schrottenloher and Shen.

See https://eprint.iacr.org/2021/1023. Some of these are still safe.

Most of these were broken before the attacks on SIDH proper.

At least those that have bizarre security assumptions

Highlighted by Peter on twitter: https://twitter.com/kutasp/status/1565330594824802306
Attack said to be fully exponential in the OSDIH paper: https://eprint.iacr.org/2020/985.pdf

Using title=... (for acronyms and links to attacks) is not friendly to touchscreens. We need to think of better ways to visualize this extra info.

Very low priority, but I could make a second :root for the css to allow for prefers-color-scheme for those who like dark websites

We're starting to pile up a lot of variants of assumptions and schemes. E.g., the various versions of parallelization, SIDH/B-SIDH, etc.

We may consider regrouping variants as subfields of a main object. E.g.:

parallelization:
  name:
    long: Parallelization
  aliases:
    - short: GA-CDH
      long: Group Action Computational Diffie-Hellman
  references:
    Cou06: 'https://eprint.iacr.org/2006/291'
  comment:
    The anaologue of CDH for Group Actions.
  variants:
    ordinary:
      reduces_to:
        - vector>ordinary
    supersingular:
      references:
        CD19: 'https://eprint.iacr.org/2019/1404'
      reduces_to:
        - vector>supersingular
      comments: Introduced for CSIDH
    oriented:
      references:
        CK20: 'https://eprint.iacr.org/2020/985'
        Wes21: 'https://eprint.iacr.org/2021/1583'
      reduces_to:
        - vector>oriented

This requires some thinking on the semantics (e.g., does it make sense to have reduces_to for the main object?) and the presentation (foldable rows?), though. Low priority.

Not post-quantum!

If we try and format a list in markdown via yml we would do:

  comment: |-
    - this
    - is a
    - list

Here we need |- to ensure the line breaks appear to tell the markdown parser we have bullet points.

However, with |- we are forced to obey all line breaks, meaning long comments will have very long lines, which isn't so nice. (We currently use >- which works nicely, but breaks the markdown list parsing.

Currently, the idea I have is to just not support lists but this isn't so nice, as users may expect this.