limits of network hacking/netsplits (was: Discovery/addr packets) [combined summary]



Individual post summaries: Click here to read the original discussion on the bitcoin-dev mailing list

Published on: 2013-05-07T21:07:50+00:00


Summary:

The author of the context disagrees with the notion that Bitcoin package maintainers are doing a good job for Linux users. They argue that Ubuntu users on the distro quantal are stuck on an outdated version of Bitcoin with no updates since 2012, and Debian Squeeze users are missing every fix since version 0.3.24. Additionally, the author notes that most distros link Bitcoin against libdb5.1, making wallets unusable for an official Bitcoin build. The author acknowledges that package maintainers may not be at fault for using outdated versions of libraries. The context also discusses the importance of secure binary distribution and the role of SSL in providing a secure connection. The author suggests that SSL is essential but requires finding alternative file hosting. They mention Damgards as a potential solution, but note that there is no choice in what to use. On Android, developers are limited in their ability to change the signing key after deployment. The author mentions the quorum-of-developers signing system using gitian and reproducible builds, but points out that people don't check the signatures, rendering the system less effective. The author suggests combining this system with an auto-update engine or other software distribution platform for better security.An email conversation between Mike Hearn and Adam explores the trustworthiness of Bitcoin binaries and ways to ensure their source. They discuss the website redesign and suggest checking PGP signatures using the Web of Trust (WoT). Some developers have their keys in a PGP Global Trust Register, which could provide a level of trust for those who take the WoT seriously. SSL-enabled websites are considered trustworthy up to the weakest link in CA security. The threshold RSA library is mentioned as a means of splitting RSA keys for updates, and Linux is identified as having the worst situation due to man-in-the-middle attacks by package maintainers. The conversation suggests multiple people signing releases to provide assurance, even if some signers are pseudonymous.The context also highlights the limitations of signatures for verifying mirror integrity, as they are often not checked and obtained from the same website serving the binary. The verify-bitcoin.sh script is mentioned as unreliable for checking signatures. The Bitcoin Wallet app on Android is protected by measures such as only accepting updates signed by the original key and being at the top of the Play Store search results. MacOS X binaries are signed under the legal identity of the Bitcoin Foundation, and unsigned binaries do not run by default. Windows antivirus companies operate binary whitelisting to fight viruses. However, Linux faces more challenges, including man-in-the-middle attacks and patches, prompting a need for a health warning on the Bitcoin website.The discussion on the Bitcoin-development mailing list revolves around various topics related to Bitcoin mining, including network manipulation through hacking pools or segmenting the network. Participants acknowledge that such attacks can be short-lived and inconsiderate. They also discuss the difficulty level in Bitcoin mining, which can be reduced every 2016 blocks. The conversation then shifts to the security of Bitcoin software packages. While some maintainers are doing a good job, there are concerns about SSL and CA infrastructure vulnerabilities. The use of gpg signatures and gitian downloader signatures is considered secure, but getting users to use them remains an open question.Email exchanges between Adam Back and others delve into the vulnerability of Bitcoin network attacks. Back argues that network attacks that induce splits are the primary vulnerability of Bitcoin. He suggests hacking pools or segmenting the network to have miners do the work. The issue of signatures is also discussed, with Back pointing out the lack of signatures for Linux binaries and tarballs. Maxwell claims that they are signed, but Back refutes this and highlights the absence of signatures. The ease of attacking binaries is mentioned, such as creating user-operated netsplits or emptying wallets via a mix.


Updated on: 2023-08-01T04:50:06.990061+00:00