- cypherpunks - lists.cpunks.org

PNC Wants Court, Not Jury, To Hear Fintech Co. TM Defenses
by Gunnar Larson 05 Aug '24

05 Aug '24

PNC Wants Court, Not Jury, To Hear Fintech Co. TM Defenses <https://www.law360.com/banking/articles/1865990?nl_pk=feb6bbba-7878-4d39-be…> By Matthew Santoni PNC Financial Services Group wants a Pennsylvania federal judge, not a jury, to decide whether the bank did nothing with knowledge that Plaid Inc. was copying its logo and login screens when linking bank accounts to financial apps, PNC said in a court filing. 2 documents attached | Read full article » <https://www.law360.com/banking/articles/1865990?nl_pk=feb6bbba-7878-4d39-be…> | Save to favorites » <https://www.law360.com/banking/articles/1865990?nl_pk=feb6bbba-7878-4d39-be…>

1 0

SDNY Brass Looks To Future After String Of Courthouse Wins
by Gunnar Larson 05 Aug '24

05 Aug '24

SDNY Brass Looks To Future After String Of Courthouse Wins <https://www.law360.com/mergersacquisitions/articles/1866081?nl_pk=a34c25b5-…> By Pete Brush A series of high-profile convictions won by federal prosecutors from the Southern District of New York shows the office is pursuing justice for a diverse community without regard for politics, its senior leaders told Law360, adding they intend to continue on that path. Read full article » <https://www.law360.com/mergersacquisitions/articles/1866081?nl_pk=a34c25b5-…> | Save to favorites » <https://www.law360.com/mergersacquisitions/articles/1866081?nl_pk=a34c25b5-…>

1 0

Uber Legal Chief Takes Leave To Aid Kamala Harris' Campaign
by Gunnar Larson 05 Aug '24

05 Aug '24

Uber Legal Chief Takes Leave To Aid Kamala Harris' Campaign <https://www.law360.com/mergersacquisitions/articles/1866203?nl_pk=a34c25b5-…> By Lauren Berg Uber Technologies' Tony West, who has served as its chief legal officer since 2017, will take a leave of absence to volunteer for his sister-in-law Vice President Kamala Harris' presidential campaign, a representative of the ride-share company confirmed Friday. 1 document attached | Read full article » <https://www.law360.com/mergersacquisitions/articles/1866203?nl_pk=a34c25b5-…> | Save to favorites » <https://www.law360.com/mergersacquisitions/articles/1866203?nl_pk=a34c25b5-…>

1 0

CFPB, NY Say Complaint Updates Aren't Unfair To MoneyGram
by Gunnar Larson 05 Aug '24

05 Aug '24

CFPB, NY Say Complaint Updates Aren't Unfair To MoneyGram <https://www.law360.com/newyork/articles/1866144?nl_pk=ac5a3855-e47c-4403-be…> By Aislinn Keely The Consumer Financial Protection Bureau and the New York Attorney General's Office fired back on Friday at MoneyGram's assertions that it'd be "futile" for the agencies to tweak their lawsuit against the remittance service provider, saying "there are no pleading deficiencies to cure" in their initial complaint. Response attached | Read full article » <https://www.law360.com/newyork/articles/1866144?nl_pk=ac5a3855-e47c-4403-be…> | Save to favorites » <https://www.law360.com/newyork/articles/1866144?nl_pk=ac5a3855-e47c-4403-be…>

1 0

Bed Bath & Beyond Accuses Gamestop CEO Of Insider Trading
by Gunnar Larson 05 Aug '24

05 Aug '24

Bed Bath & Beyond Accuses Gamestop CEO Of Insider Trading <https://www.law360.com/newyork/articles/1865902?nl_pk=ac5a3855-e47c-4403-be…> By Dorothy Atkins The bankrupt big-box housewares retailer once known as Bed Bath & Beyond Inc. sued GameStop CEO Ryan Cohen and his company RC Ventures LLC in New York federal court Thursday, seeking to recover $47 million that it says the defendants made from insider trading not long before the company went belly-up. Complaint attached | Read full article » <https://www.law360.com/newyork/articles/1865902?nl_pk=ac5a3855-e47c-4403-be…> | Save to favorites » <https://www.law360.com/newyork/articles/1865902?nl_pk=ac5a3855-e47c-4403-be…>

1 0

Ex-Dropbox Exec Says JPMorgan Can't Block Arbitration
by Gunnar Larson 05 Aug '24

05 Aug '24

Ex-Dropbox Exec Says JPMorgan Can't Block Arbitration <https://www.law360.com/newyork/articles/1864681?nl_pk=ac5a3855-e47c-4403-be…> By Sydney Price A billionaire co-founder of the file sharing platform Dropbox has asked a California federal court to declare that he can take a JPMorgan wealth management unit to arbitration for more than $225 million in damages he says he suffered at the hands of an investment adviser who previously worked for units of First Republic Bank that JPMorgan acquired last year. Complaint attached | Read full article » <https://www.law360.com/newyork/articles/1864681?nl_pk=ac5a3855-e47c-4403-be…> | Save to favorites » <https://www.law360.com/newyork/articles/1864681?nl_pk=ac5a3855-e47c-4403-be…>

1 0

Manhattan DA Slams Trump's 'Regurgitated' Recusal Bid
by Gunnar Larson 05 Aug '24

05 Aug '24

Manhattan DA Slams Trump's 'Regurgitated' Recusal Bid <https://www.law360.com/newyork/articles/1866155?nl_pk=ac5a3855-e47c-4403-be…> By Frank G. Runyeon The Manhattan district attorney pilloried Donald Trump's renewed request for the judge overseeing his hush money case to recuse himself, branding it a "regurgitated" attempt to rehash issues the court already decided without any new facts — besides Kamala Harris' presidential bid. Read full article » <https://www.law360.com/newyork/articles/1866155?nl_pk=ac5a3855-e47c-4403-be…> | Save to favorites » <https://www.law360.com/newyork/articles/1866155?nl_pk=ac5a3855-e47c-4403-be…>

1 0

Ex-F1 Boss Sues HSBC For £5.5M For Wrongly Selling Bonds
by Gunnar Larson 05 Aug '24

05 Aug '24

Ex-F1 Boss Sues HSBC For £5.5M For Wrongly Selling Bonds <https://www.law360.co.uk/financial-services-uk/articles/1863047?nl_pk=a94fc…> By Sophia Dourou Former Formula One team boss Eddie Jordan has sued HSBC for £5.5 million ($7 million) for losses suffered as the result of the bank's allegedly false and negligent advice over a bond in which he invested tens of millions of pounds. Read full article » <https://www.law360.co.uk/financial-services-uk/articles/1863047?nl_pk=a94fc…> | Save to favorites » <https://www.law360.co.uk/financial-services-uk/articles/1863047?nl_pk=a94fc…>

1 0

[crazy][crazy] Design for a Time Lock Blockchain
by Undescribed Horrific Abuse, One Victim & Survivor of Many 03 Aug '24

03 Aug '24

Time lock cryptography is an old problem apparently originally described by Tim May, I had no idea. A solution to this was described at https://people.csail.mit.edu/rivest/pubs/RSW96.pdf long ago. Basically, a ciphertext is constructed such that it takes n operations of squaring a number (modulo) to recover it, So, a time lock blockchain might perform these squaring operations as proof of work, to mine its currency. Demand would be produced by people paying to encode time lock messages. Each block would announce decrypted messages. To keep the mining effective, as the time lock protocl and the security of the chain depend on the network being the strongest source of computation that exists, it could be important to make it very clear to miners how to improve the mining algorithm and tune it for new hardware. Rather than hiding it away in mining apps, a good focus might be on figuring out what parts can be factored into user control, and how to ease users writing their own GPU kernels, FPGA layouts, etc, with minimal training. It's obvious that this solution is obvious. However, it is not mentioned at https://crypto.stackexchange.com/questions/606/time-capsule-cryptography in my "internet", and the site gives me errors when I try to log in to post. First steps might include cloning code for an existing blockchain, and coding the Rivest algorithm into the mining and verification code. A websearch shows some of the existing attempts: - drand has timelock cryptography https://drand.love/docs/timelock-encryption but it fails if nodes are compromised since they hold the private material in a multiparty manner, whereas Rivest's solution does not have this failure uhhhhhhhhhhhhhh I ran into https://eprint.iacr.org/2023/189.pdf which mentions a number of other approaches to this, _none_ of which were mentioned on stack exchange ?? this is apparently an old and popular problem for 30 years. thinking the stack exchange is broken and others aren't logging in either or something. there are apparently a number of existing solutions to this, but it is not immediately obvious what their security guarantees are. > the LCS time-lock puzzle released in 1999 by Ron Rivest was anticipated to remain secure for 35 years, but was successfully solved fifteen years early in 2019 via two independent methods [csa19]: once by simply running sequential operations on a single core of a modern CPU for 3.5 years, and again by running the computation for only 2 months on dedicated FPGA hardware. Rabin and Thorpe take the alternative approach and propose the idea of relying on Pedersen distributed key generation, Feldman verifiable threshold secret sharing, and ElGamal encryption [RT06]. Therein multiple parties implement a “Time-Lapse Cryptography Service” by publishing public keys and then releasing the related private keys at given times. To the best of our knowledge such a Time-Lapse Cryptography Service has never been implemented and deployed in practice. Moreover this setting requires explicit action by a threshold of members to compute the decryption operation. In other words, the work required by the network is linear in the number of ciphertexts to decrypt. In [LJKW18] a timelock scheme is built on top of the concept of computational reference clocks using extractable Witness encryption, and a practical instantiation relying on Bitcoin is presented. However, this scheme relies on multilinear maps which have not yet been securely constructed. In [CDK+21] IBE and Witness encryption are again used to achieve a novel notion of “Encryption to the Current Winner” (ECW) where the receiver of an encrypted message is determined by the current state of a blockchain and is not yet known at the time of encryption. Ultimately this scheme is then transformed to achieve a different notion of “Encryption to the Future” (EtF) towards parties selected at arbitrary points in the future, unlike classical timelock or Timelapse encryption where any party knowing the ciphertext can decrypt once the specified time has passed. In direct contrast to our timelock definition which permits anyone to decrypt after the designated time, in EtF only some party with certain rights (the future winner) not yet determined in the present can decrypt ciphertexts. Notice that the setting is also quite different between both works, assuming dynamic committees with the strong assumption that “YOSO”—You Only Speak Once. This has not yet been implemented nor deployed in practice to the extent of our knowledge. Threshold networks have also been used as an encryption “recipient.” For example, Ferveo [Ano21] and Shutter [shu22] are systems where users encrypt their transactions for threshold networks (which can also be the validator set of a blockchain). Once encrypted transactions are included in a block, they get decrypted by the network, thereby preventing front-running. In comparison, our scheme does not rely on the threshold network to perform an operation per ciphertext, rather simply to emit one value per round and the decryption happens publicly. Anyone can compute a valid decryption of a ciphertext encrypted for this round.

1 1

[ot][aiml] Merge Higher-Performing Models Without Compute Across Modes?
by Undescribed Horrific Abuse, One Victim & Survivor of Many 02 Aug '24

02 Aug '24

from March via X (my X feed content is restabilizing after the change) https://arxiv.org/abs/2403.13187 https://github.com/SakanaAI/evolutionary-model-merge <https://github.com/SakanaAI/evolutionary-model-merge?tab=readme-ov-file> The Open LLM Leaderboard is now dominated by merged models, showcasing its potential for democratizing foundation model development. https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard 1. Automated Model Composition: We introduce Evolutionary Model Merge, a general evolutionary method to automatically discover optimal combinations of diverse open-source models for creating new foundation models with user-specified capabilities. This approach harnesses the collective intelligence of existing open models, enabling the creation of powerful models without the need for extensive training data or compute. 2. Cross-Domain Merging: We demonstrate that our method can discover novel ways to merge models from disparate domains (e.g., non-English language and Math, non-English language and Vision), potentially exceeding the capabilities achievable through conventional human design strategies. 3. State-of-the-Art Performance: We showcase the effectiveness of our method by automatically generating a Japanese LLM with Math reasoning capability and a Japanese Vision-Language Model (VLM). Notably, both models achieve state-of-the-art performance on various benchmarks, even without explicit optimization for those tasks. 4. High Efficiency and Surprising Generalizability: We observe that our 7B parameter LLM surpasses the performance of some previous 70B parameter Japanese LLMs on benchmark datasets, highlighting the high efficiency and surprising generalization capability of our approach. We believe this model can serve as a strong general-purpose Japanese LLM.

1 1