Hacker News Reader

Comment by Luc | Hacker News Reader

How the comparison is supposed to be displayed:

  +-----[first]-----+    +-----[second]----+
  |       .=o.  .   |    |        ..=o   . |
  |     . *+*. o    |    |       o 0+.. o  |
  |      =.*..o     |    |        X.o. o   |
  |       o + ..    |    |       . * .  .  |
  |        S o.     |    |        S o  .   |
  |         o  .    |    |         o    .  |
  |          .  . . |    |          .    ..|
  |              o .|    |               .o|
  |               E.|    |               Eo|
  +-----------------+    +-----------------+

That's way too much alike IMHO. A more chaotic algorithm would create images that differ more substantially.

View on HN

Interested mobile users: Please rotate your screen to see "How the comparison is supposed to be displayed"

Interested HN mods: Yet another request for horizontally scrolling code blocks instead of line wrapping.

Is this better?

  +-----[first]-----+
  |       .=o.  .   |
  |     . *+*. o    |
  |      =.*..o     |
  |       o + ..    |
  |        S o.     |
  |         o  .    |
  |          .  . . |
  |              o .|
  |               E.|
  +-----------------+
  +-----[second]----+
  |        ..=o   . |
  |       o 0+.. o  |
  |        X.o. o   |
  |       . * .  .  |
  |        S o  .   |
  |         o    .  |
  |          .    ..|
  |               .o|
  |               Eo|
  +-----------------+

>yet another request

I believe it used to work this way, but they changed it after people complained. Go figure

A fun trick is to cross the eyes so the differences become obvious - a bit easier with more space between them:

    +---[ 1       ]---+                        +---[       2 ]---+
    |       .=o.  .   |                        |        ..=o   . |
    |     . *+*. o    |                        |       o 0+.. o  |
    |      =.*..o     |                        |        X.o. o   |
    |       o + ..    |                        |       . * .  .  |
    |        S o.     |                        |        S o  .   |
    |         o  .    |                        |         o    .  |
    |          .  . . |                        |          .    ..|
    |              o .|                        |               .o|
    |               E.|                        |               Eo|
    +-----------------+                        +-----------------+

The chaotic algorithm is meant to be the hash, not the art-ification. So of course if you take two very similar hashes they will appear very similar.

Doing an extra round of hashing wouldn't hurt though, right?

Either the hashes are very different and you can already tell from them (but the art will also be different) or they are very similar (someone trying to sneak a close-enough hash by you) and then the second round of hashing leads to very different art.

Doing an extra round of hashing wouldn't do anything. If they were different before, now they might be the same. If the hashing algorithm is worth its salt (pun intended) then the original hash will be just as random as the hash of the hash, providing nothing of value.

I think it might work if you display both the hash and the image, and ask the user to compare both.

If the hashes are very different, it doesn't matter that the images are similar (assuming the user indeed verifies both).

If the hashes are similar, another round of hashing before producing the image, might make the images very different.

But maybe I misunderstood?

The problem is that an attacker can churn out public keys until they end up with a hash that's very similar.

Picking a random number and doing a keyed hash of both inputs, then comparing those, would bring out small differences in a way the attacker cannot control.

I think that would work well. A high-quality hash is designed such that changing a single input bit will flip approximately half of the output bits. So hashing both values will make it very easy for a human to spot whether the input values were different.

it might be, but its much easier to spot a difference for a technical eye.

Ideally you'd want spot a change without comparing them side by side, I think.

Reply on news.ycombinator.com