Can we determine how 'k' is selected?

Hey all I found a curious case with the Bitcoin address "1LN4yp6rQALjwg53SKsi44teq1fp2v5wqR". It reused the nonce 'k' in two separate transactions signed with the same private key. Here are the details: Transaction 1: 65278a3cf6d4710bd5a556edeb6854e07de0324a5008569e33984da491df8c38 R = 00b42646001435ec60a40982735c4e73d31be13d0086be394eaf5ce10b0f38f689 S = 711a12a1a4db96159ec5760f2201beeb04043efc3d39e10ba17c9e66816ff9ad Z = 91002245c37eba47e4f5764f0c4780b3cd9ba0f3ea709368cfd21717d82be496 Transaction 2: d87d1bc2d41ed867f0f4b7b7ee9915f09cb82477388e306acac8279146edacb8 R = 00b426460...

Unless r has an obvious value (eg. 1*G) it is going to be very difficult to guess how it was chosen. The only method is to use a search engine to see where the address/transaction is mentioned and see the reason there. In the past when any of these broken keys were found, the owner is asked what wallet they were using and by analyzing the behavior and/or looking at its source code the broken part was found. For example when the users said they were using blockchain.info the source code was analyzed and it was found that they were using random.org to generate the ephemeral key and the site was returning an error message which the app still used as input which ended up being the same for everyone!

Surprisingly the way to generate the k from 1LN4yp6rQALjwg53SKsi44teq1fp2v5wqR doesn't seems weak. This my method to say that Step 1 : recovered every tx signed by 1LN4yp6rQALjwg53SKsi44teq1fp2v5wqR (total sent136 BTC!!!!) Step2 : identified the two input where  r==b42646001435ec60a40982735c4e73d31be13d0086be394eaf5ce10b0f38f689 Step 3: recovered k for this two input  with nonce reused formula this is the secret recovered: But 1 sample of k is too low to see a patern or some biased entropy so  I made the assumption that every k was generate with the same (probably) weak RNG. Step 4: recovered every k for 1LN4yp6rQALjwg53SKsi44teq1fp2v5wqR (priv = 28608884601749819904852249207473436882265395560722309163084982020290971201593) with this formula there is 675 different k for 1LN4yp6rQALjwg53SKsi44teq1fp2v5wqR removed the one of the two b42646001435ec60a40982735c4e73d31be13d0086be394eaf5ce10b0f38f689 duplicate ( not to false entropy) Step 5 : used a bunch of rng testing statiscal test from this google github repo on the list of k: http://[code]https://github.com/google/paranoid_crypto[/code] this is the result of the tests (p values) (0.7526555431379823, 'Frequency') (0.8922205625150528, 'LongestRuns') (0.17401029065157583, 'Runs') (0.09841635668060003, 'BinaryMatrixRank') (0.133636, 'LargeBinaryMatrixRank_64_*_64') (0.711212, 'LargeBinaryMatrixRank_128_*_128') (0.711212, 'LargeBinaryMatrixRank_256_*_256') (0.13004210277109796, 'OverlappingTemplateMatching') (0.9204344990860858, 'UniversalImpl1') (0.7418997106144236, 'UniversalImpl2') (0.9851583654051379, 'LinearComplexity_dist') (0.32331377641368286, 'LinearComplexity_extreme') (0.5797307681237995, 'ApproximateEntropy_0') (0.7235022428413989, 'ApproximateEntropy_1') (0.6577288253868533, 'ApproximateEntropy_2') (0.49821641104893855, 'ApproximateEntropy_3') (0.6156687031532114, 'ApproximateEntropy_4') (0.45974516471567906, 'ApproximateEntropy_5') (0.21884604607320576, 'ApproximateEntropy_6') (0.2414807578348784, 'ApproximateEntropy_7') (0.15020972293729182, 'ApproximateEntropy_8') (0.18459092265541552, 'Spectral') (0.4303553446832167, 'RandomWalk_0') (0.7000550793665439, 'RandomWalk_1') (0.7229207109153853, 'LatticeFindBias') !! A bad p_value indicating a weakness had to very low (for ex 10e-7) this is not the case here.. : NB:The latest test is the more important and perform a Lattice Reduction with samples of k. It is very powerful to identify a bias in a lcg and in a lot of weak prng (according to this docs of the library :https://github.com/google/paranoid_crypto/blob/main/docs/ecdsa_signature_tests.md