Been thinking about Pollard's kangaroo method lately and found a paper on it. It’s titled "Kangaroo Methods for Solving the Interval Discrete Logarithm Problem". Seems like the author is claiming he can beat the 4-kangaroo approach with a 5-kangaroo method.
Honestly, seems kinda limited. I mean, why not just subtract G from the initial pubkey? Keep both versions and at least one will have an even private key. Then you can 'divide' them and find your way into the right interval.
That could work, but what’s the catch? If the private key is 0x1, adding G would be better. If we go the other way, we have to check for the point at infinity, which slows things down a bit.
Totally agree on that. But then we’re left with two points to compare. Like, how do we know which one lies in the lower interval? It's a bit of a hassle searching both.
Papers like that are usually about getting degrees, no real innovation. Honestly, you won’t get better than K=1.7 no matter how many kangaroos you use. Super boring. Why not try a different approach?
Yeah, we all know that paper. I’ve even put it into my own Python script. It does bring C down to around 1.25-1.45, but the logic is a mess and it’s super complex to handle. Plus, starting points are costly.
Sure, here it is. Just got experimental proof that it works. It’s like mixing the 3-kangaroo method with the van Oorschot idea, but no dynamic programming involved.
You’re overselling it. Precomputing values risks ruining the randomness, which is what makes the birthday paradox work. You might just end up in a mess.
The birthday paradox relies on randomness, which lets you find collisions easier. If you make kangaroo more deterministic, you lose that edge and it just becomes brute force.