Schneier on Security

maqp • September 21, 2019 3:14 PM

@Vesselin Bontchev, all

The cado-nfs debug messages can still be seen on the video:
https://www.youtube.com/watch?v=uEVJrQEVd0I#t=7m45s

Info:root: Set tasks.threads=32 based on detected logical CPUs

The amount of RAM is also being displayed.

Was that not the same press conference video? If not, what was on the other video?

Nicholas Weaver noticed this but I’m not sure if he was the first one:

https://twitter.com/ncweaver/status/1175493814296858624

Finally, gotta appreciate the sense of humor from the maintainer of the satire repository:

https://github.com/CrownSterlingIO/SemiPrimeFactorization/commit/b43ea3cce3a37024c077456e33e43190434c7028

maqp • September 26, 2019 12:18 PM

@Vesselin Bontchev

Haha. So the same story as was with the first TimeAI video they uploaded: its commenting was also disabled after it received a lot of critique.

maqp • October 27, 2019 8:17 PM

@All

In the previous post I tried to evaluate the efficiency of the Crown Sterling reciprocal factoring method. Turns out there is no memory requirement for it. But it’s still not feasible:

Looking at https://www.geeksforgeeks.org/find-length-period-decimal-value-1n/ the way the algorithm

def getPeriod(n) : 
    # Find the (n+1)th remainder after decimal point in value of 1/n  
    rem = 1 # Initialize remainder  
    for i in range(1, n + 2):  
        rem = (10 * rem) % n 
  
    # Store (n+1)th remainder  
    d = rem 
      
    # Count the number of remainders before 
    # next occurrence of (n+1)'th remainder 'd'
    count = 0
    rem = (10 * rem) % n 
    count += 1
    while rem != d : 
        rem = (10 * rem) % n 
        count += 1
      
    return count 

produces the period, is it first iterates over the range (1, n+2), so the first step has n+1 steps.

The remainder is then stored, and remainders are iterated backwards until the remainder repeats: the second part only requires as many steps as the period has length.

So, for RSA-2048, you need 2²⁰⁴⁸ + len(period) steps to determine the period.

The most efficient way to find the factors is, while you’re iterating over the 2²⁰⁴⁸+1 steps, you keep log₁₀sqrt(2²⁰⁴⁸) digits long substring (308 for RSA 2048) that is the RSA prime factor candidate (best case scenario).

When you produce the next decimal with the algorithm above, you need to slice off from the other end in order to prevent the memory cost from reaching insane levels: Like I said in the previous post, for just the RSA-100 (that has been insecure since the 80s) you’d have to store 15226050279225333605356183781326374297180681149613
80688657908494580122963258952897654000350692006139-1 digits, i.e. you’d have to store in the order of 10 petabytes of digits inside every atom in the observable universe.

Once you’ve added the next decimal of the reciprocal to the substring, you have 309 digit long substring. Thus, you have to slice one digit from the other end to get back to the 308 digit substring. This substring is most likely not prime, but checking that with a really efficient test like Rabin-Miller takes much much longer than just seeing whether it divides the semi-prime evenly. So to see if the substring is a prime factor, you just divide the semi-prime with it.

Because you sliced the one digit from the other end after every iteration, you have to keep doing the division after every step, because otherwise you might miss the prime factor candidate.

The question is, how efficient is this?

Well, I took time to implement as efficient algorithm as I could based on the non-existent documentation (read: the instagram post) with Python (the language should matter very little as we’re comparing efficiency between the algorithms by counting the steps, not by comparing the execution time).

The source code along with example output can be found here: https://gist.github.com/maqp/b04e27e795ab01c5cdefdcddc59e3a02

The tl;dr is, for that particular game of 100 rounds with 32-bit semiprimes, the Crown Sterling Reciprocal Factoring Method was on average 5.85 times slower than random division (brute force).

Note that the implementation of the reciprocal factoring method we’re using does not first have to iterate over the entire reciprocal, it’s testing the period on the fly so it exits as soon as possible.

Considering the comparison, it can be concluded the reciprocal factoring algorithm is just a crappier implementation of trial division, and that both run in exponential time (which in turn means Crown Sterling has achieved nothing.)

This makes it a lot less efficient than GNFS of the cado-nfs program they were using to break RSA256, and as shown above, slower than brute force.

While I’ve been thinking about this, I asked the good folks at cryptography.stackexchange.com about Crown Sterling’s reciprocal factoring method:

https://crypto.stackexchange.com/questions/74614/what-is-the-efficiency-of-the-new-crown-sterling-semiprime-factoring-method

The top answer turned out to be a fascinating read and technical dissection of the snake oil claims by the company. I highly recommend reading it.

maqp • October 28, 2019 3:37 AM

@Clive Robinson

They’ve been re-hashing the mod24 quasi-prime wheel for very long time.

Apart from the history Bruce linked above, most information has been in the instagram (go figure):

The prime wheel first appeared in August 7, 2018.

The findings from Shakespeare Sonnets with Alan Green appeared in Sept 29, 2018.

A Picture from March 12 this year shows Grant being interviewed for the Thrive movie https://www.youtube.com/watch?v=lEV5AFFcZ-s — I tried skimming it and the end credits, but didn’t find him anywhere.

March 21 shows him sketching the Time AI logo.

He did the privacy piracy interview, Ted Talk, The prime wheel interview for Hold my Ark. Then there was the blackhat talk, cado-nfs presentation, and now apparently they hosted a new age conference called Conference on Precession of Ancient Knowledge, or, CPAK: https://www.instagram.com/p/B3Cen7WnNo0/?igshid=a4usg3kn0xfu

My guess is Crown Sterling is still in its baby steps.

But if you look at https://www.youtube.com/watch?v=8paZWjiJIk8 you’ll see the company portfolio of Strathspey Crown (the parent company?) and all the snake oil in the world from Audiophile crap, medical patches with hundreds of medications in them, something called GeneRADAR that looks A LOT like the Theranos blood scanner, Botox alternative called Jeuveau, some social media app for physicians, some refrigerant..

And then there’s Torus Tech:

Torus Tech focuses specifically on quantum vacuum energy technologies across a wide area of application including energy production, gravitational control, health and medical treatments and many more. Resonance science and unified physics theory lie at the core of all these applications carried by a strong partnership between our research team and the core engineering team.

They have a product called “Harmonic Flux Resonator”:

Replicating the magnetohydrodynamics occurring in a variety of astrophysical objects.

Energy extracted directly from the vacuum. Nothing to burn, nothing to consume, nothing to destroy. No fumes, no toxins, no limitations. Nothing short of a paradigm shift which will alter the course of human kind forever.

So… free energy… Oh, they have another product:

ARK® crystals are a revolutionary technology that greatly boosts the body’s natural ability to attune with the vitalistic and expansive zero-point field of the quantum vacuum. The quantum vacuum represents the revelatory understanding in modern physics that space is not empty; on the contrary, it is the one thing that connects all things.”

They are serious about selling you $51,200 piece of machined bullshit: https://arkcrystals.com/product/64-ark-crystal-bundle/ (I really like the looks though, maybe it passes as art, at least in their opinion).

My favorite part on that site however is the “this is all bullshit” legal disclaimer they have to put there:

DISCLAIMER – Statements made on this site have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease. Consult your own doctor.

So, with that being said, considering Time AI is the last item on that portfolio, it could also be the newest. The reason Crown Sterling hasn’t done anything else yet is because they’ve been really busy building an empire of bullshit.

I’m fairly sure these guys will make the main story on John Oliver by the end of the next decade.