Has Steve Gibson been informed?

[sukima]

If this is broken then the original author needs to be informed. Does Steve have anything to contribute?

[Sc00bz]

I informed him on February 4th and he has not gotten back. There are four things that can be changed (first two are important and the last two are bike shed):

• Remove "cryptographically strong PRNG" from the comment block. Maybe replace that with a warning about not being a CSPRNG and not to use it for cryptography.
• Move the return line in mash() to the end of the function:

-			return (n >>> 0) * 2.3283064365386963e-10; // 2^-32
		} else n = 0xefc8249d;
+		return (n >>> 0) * 2.3283064365386963e-10; // 2^-32

• Use a function like https://github.com/Sc00bz/ModRandom/blob/master/random.js but instead of window.crypto.getRandomValues(rand); it would be rand = rawprng() * 0x100000000;
• Replace s[i] = mash(Math.random()); with window.crypto.getRandomValues to set s[i].

Then the only thing left is the "slight bias between every 24th random number".

[firepacket]

I informed him on February 4th and he has not gotten back. There are four things that can be changed (first two are important and the last two are bike shed):

• Remove "cryptographically strong PRNG" from the comment block. Maybe replace that with a warning about not being a CSPRNG and not to use it for cryptography.
• Move the return line in mash() to the end of the function:

-			return (n >>> 0) * 2.3283064365386963e-10; // 2^-32
		} else n = 0xefc8249d;
+		return (n >>> 0) * 2.3283064365386963e-10; // 2^-32

• Use a function like https://github.com/Sc00bz/ModRandom/blob/master/random.js but instead of window.crypto.getRandomValues(rand); it would be rand = rawprng() * 0x100000000;
• Replace s[i] = mash(Math.random()); with window.crypto.getRandomValues to set s[i].

Then the only thing left is the "slight bias between every 24th random number".

So you go on to tell him how to make it cryptographically secure but then tell him to remove that description?

Is it not true that both your attacks on your algorithm require knowledge of its initial state?

[Sc00bz]

It is not cryptographically secure before or after the changes. Just because it once passed diehard and dieharder doesn't make it a CSPRNG. The output of the algorithm is its current state which makes it very much not cryptographically secure.

[firepacket]

Can you please take a look at my implementation here any feedback would be appreciated:
https://github.com/firepacket/CSharp-UHEPRNG-Ultra-High-Entropy-Pseudo-Random-Number-Generator-WITH-GUI/

I believe it should be cryptographically secure as long as nobody knows the starting state.

[Sc00bz]

as long as nobody knows the starting state.

You keep saying this thing about "knowing the starting state". I do not know the starting state. I ask for random numbers and build the state from that. This PRNG is returning its internal state. I record that data and can predict future output.

Also it is not cryptographically secure regardless of your believes because it's returning its internal state.