My general hope/vague plan for bitcoinj based wallets is to get them all on
to automatic updates with threshold signatures. Combined with regular
audits of the initial downloads for new users, that should give a pretty
safe result that is immune to a developer going rogue.
On Wed, Apr 3, 2013 at
By the way, I have a download of the Bitcoin-Qt client and signature
verification running in a cron job.
On Thu, Apr 4, 2013 at 10:11 AM, Mike Hearn m...@plan99.net wrote:
My general hope/vague plan for bitcoinj based wallets is to get them all
on to automatic updates with threshold
gpg signing commits, like the Linux kernel
Though, honestly, when I ACK that means I read the code, which is more
important than the author really. github seems fine for that still,
though I do wonder if there is a race possible,
* just before I click pull, sneak rebases the branch to
I would rather we spend time working to make users' bitcoins safe EVEN IF
their bitcoin software is compromised.
Eliminate the if you get a bad bitcoin-qt.exe somehow you're in big
trouble risk entirely, instead of worrying about unlikely scenarios like a
timing attack in between ACKs/pulls.
Users will have available multisig addresses which require
transactions to be signed off by a wallet HSM. (E.g. a keyfob
Hardware is a good thing. But only if you do the crypto in the
hardware and trust the hardware and its attack models ;) For
instance, the fingerprint readers you see
On Tue, Apr 2, 2013 at 11:41 PM, Wladimir laa...@gmail.com wrote:
Maybe now that bitcoin is growing out of the toy phase it's an idea to start
gpg signing commits, like the Linux kernel
(https://lwn.net/Articles/466468/).
But I suppose then we can't use github anymore to merge as-is and need
I was just looking at:
https://bitcointalk.org/index.php?topic=4571.0
I'm just curious if there is a possible attack vector here based on the
fact that git uses the relatively week SHA1
Could a seemingly innocuous pull request generate another file with a
backdoor/nonce combination that slips
An attacker would have to find a collision between two specific pieces of
code - his malicious code and a useful innoculous code that would be
accepted as pull request. This is the second, much harder case in the
birthday problem. When people talk about SHA-1 being broken they actually
mean the
On 1 April 2013 20:28, Petr Praus p...@praus.net wrote:
An attacker would have to find a collision between two specific pieces of
code - his malicious code and a useful innoculous code that would be
accepted as pull request. This is the second, much harder case in the
birthday problem. When
On 2 April 2013 00:10, Will w...@phase.net wrote:
The threat of a SHA1 collision attack to insert a malicious pull request
are tiny compared with the other threats - e.g. github being compromised,
one of the core developers' passwords being compromised, one of the core
developers going rogue,
The attack Schneier is talking about is a collision attack (i.e. it
creates two messages with the same hash, but you don't get to choose
either of the messages). It's not a second preimage attack, which is
what you would need to be able to create a message that hashes to the
same value of an
And the moment I hit send I realised it's not necessarily true.
Conceivably, a collision attack might help you craft two commits (one
good, one bad) with the same hash.
But I still maintain what I just posted is true: if someone gets
malicious code into the repo, it's going to be by social
12 matches
Mail list logo