I think multiple  ideas  here are viable and similarly good.  I'll think about which one I'd want to implement and in the mean time, this discussion may strike up the interest of anyone with skills more applicable to either of these devices

Benjamindees:

Smartcards are an interesting suggestion.  If there were an ECDSA card out there, it sounds like a private key could not be extracted from it, even for backup purposes.  Is this correct?


If the implementation were done on a non-ECDSA javacard, something like a computer or smartphone would still have to interface with it, correct?

My concern is that a compromised machine could tamper with the input to the javacard, such that the destination address and/or amount is changed when a user wants to make a transfer.

On the other hand, the idea of using a smart card kept in a safe as a bitcoin savings account seems pretty sexy from a usability standpoint.  For people like allinvain who want to keep around 25,000 BTC for a long period of time, is a smartcard something he'd want to keep it on?

(ben-abuya you posted right around when I was going to and it looks like we have similar concerns about the smart card as a savings account)

Awesome!  I've set up a github page for this: https://github.com/bitcoin-idea/bitcoin-savings

It's just an empty folder with a description right now.  Feel free to push/pull to it.  As I mentioned earlier, I don't have my own Arduino, but I can help with research.  I'm familiar with java (and c++) and I already have the bitcoinJ source on my computer.  I'm going to keep monitoring this thread for ideas to tie into the project.

(as an aside, I may want to move ideas/dialogue about the project there or reddit/r/bitcoin.  The bitcoin.org forum loads insanely slow for me)


Bert:
This idea is a response to a thread I linked to in my first post of this thread.  I've yet to see a thread titled "help, my husband died and I can't access our 25k bitcoins".  When that happens, someone can address that problem then.  Right now the bigger threat to Bitcoin seems to be private key theft via viruses or keyloggers.

If someone can get a hardware prototype running, I could just look through the Google bitcoinj source for the implementation details of how they store a signed transaction and push it to the network.  From there, it's just adapting their code to make a mini script.  Once a proof of concept works and can add transactions to the network, then it'll be easier to propose an addition to the bitcoin client.

Thanks for the search idea, it should've occurred to me 

That's a very solid way of doing it.  Any ideas what subforum or site would have the most arduino programmers?

I conceived of this idea with the arduino since I know about it from my friends, but there could be other platforms that would make this simpler or faster to develop.  I'd probably want to stay away from mobile devices since they're non-dedicated and I'm guessing that trojans/rootkits can be written for them to target BC users

This is basically the inspiration for my idea.  If you go look at allinvain's post (the guy who lost 25k BC), there's this back-and-forth between people who say how simple the solution is (i.e. linux, separate machine, etc) and the others.

I'm not proposing this arduino solution for everyone, certainly not the former group as they're already aware.  It's meant for the latter.  That group might include our parents or grandparents, maybe people who are somewhat tech-savvy but for one reason or another can't or won't spend time learning about linux.  It could also be for the tech-savvy who just want something "that works (tm)".



Wow 1 and 2 really simplify things.  This is why I like the idea of open source!

For the public/private key backup, there are many routes to go.  One could program a single button, button combination (hold 2+ together) or a button sequence to write a file to the usb drive.  If someone wants to get really fancy, integrate some identity check like a fingerprint reader, but I say make the barebones functionality first.

An alternative approach is to use an SD card for its memory store, though the memory card and reader would ultimately add to the cost of the unit.  As an open source project, this could still be developed as a fork or plugin and buyers (or businesses) could offer a different variety with simpler backup.

Edit: how do you see a password/usb backup working together if the unit doesn't support a keyboard? Pre-programmed codes that differ for each unit?

To get started with the open source option, you'd want to set up an online repo.  I've done some basic development on github.com, which offers free source hosting for open source projects.  Sourceforge is another option.  I like github as it integrates with git very nicely.  I don't *believe* that there are language restrictions for github projects, so an arduino project should be fine.

After that, I think this comes down to acquiring an arduino and determining what add-ons will be needed.  Starter ideas:
-usb interface
-some sort of keyboard/keypad?  a usb port to accept a keyboard?
-an adequate display, maybe

The remainder is developing the basic functionality
-generating a public/private key on an arduino
-generating a transaction hash (i.e. signing a transaction)

I can help with reading, research and documentation.  Eventually I can help with testing and development, but as I'm unemployed I'd like to limit my spending to what's necessary to live.  If only I could eat arduinos  .

An arduino, as a turing machine, should be able to perform the crypto functions.  Your question does lead to this one though: "Can the arduino perform the crypto functions in a reasonable amount of time?"  On that, I don't have an answer.

I'm not sure about the specifics of the bitcoin public/private key address generation.  So someone with more knowledge of the Bitcoin implementation would be required here.  I could also look through the source when I have more free time and report back when I find out.  I've spent some time looking for documentation and I couldn't find anything about the public/private key address algorithm (not to be confused with the hash confirmation algorithm, which is based on elliptic curve cryptography).

A post 2 days ago drew attention to how one can easily lose massive amounts of Bitcoins quickly if proper precautions aren't taken.  I read through most of the posts yesterday and there's quite a disconnect between tech-savvy users and the rest.  I believe that this mentality will hold Bitcoin back as a currency, so I'm reposting an idea I put up on reddit's r/bitcoin. (Original post here)

Proposed Solution: create an arduino bitcoin wallet to serve as a hardware wallet for a savings account. The arduino will:

• generate a new public/private keypair so that the private key is never seen by a computer (multiple posts suggest that the private key is unencrypted in your computer's RAM for ~100 ms, which may be enough time for an already installed virus/trojan to grab it.)
• display public and private keys
• perform an encryption/transaction in 2 steps:
•     1.  generate the transaction (use the public/private keypair along with a (user-inputted) destination public key address)
•     2.  publish this transaction to an external device (I.E. usb key, wifi network, ethernet connection)

To use:

• When your (computer's) bitcoin wallet becomes too fat: make a transfer to the arduino public key. The arduino doesn't even need to know how much is in the account, the transactions just need to be published to the bitcoin network
• The arduino can generate transaction hashes when you want to transfer money out of the arduino savings account. The user would then have to publish this transaction hash to the bitcoin network. Overspending will simply be rejected by the bitcoin network.  (Technical note: this avoids a trojan/virus stealing the user's private key, unless the arduino can be compromised)

While this account hack was terrible for the account owner, allinvain, it was bound to happen eventually to someone.  The silver lining here is that many people will want to address this potential security hole with the currency.  This is both a business problem and a bitcoin problem. I see two directions for this idea:

closed source: write the arduino code and offer the finished arduino project as a product

benefits:

• more profit for the developer
• higher entry-cost for would-be competitors

open source: arduino code is published online and peer-reviewed

benefits:

• more collaboration will generally offer better security
• will help the bitcoin community grow more by offering a low-cost method for easy security (which also generates more trust for the bitcoin currency)
• can still support a private business model where someone loads an arduino (or orders a mass-production fabrication for even lower costs) and offers a simple out-of-the-box product for users who want to do no configuration (this is the Redhat Linux model)

For anyone that wants to take this idea and run with it, it's yours. I'd personally like to see an open source solution developed, so users can
1.  buy an arduino
2.  review the code (if desired)
3.  transfer the code to an arduino to have a secure, physical wallet

Step 2 is an important optional step, IMO, as open source code is harder to hide backdoors.  However, even a blackbox-type product would still benefit the community, as it will introduce more confidence in bitcoin.