4 ms·

While we could implement a system to have tamper proof verification of stuff that's been archived, how do you prove that the archive (snapshot) is what was actu

by cm3 10y ago

While we could implement a system to have tamper proof verification of stuff that's been archived, how do you prove that the archive (snapshot) is what was actually at the target at creation time and hasn't been modified before archival? I'm thinking in purely technical terms here, and that I can easily argue that such evidence is inadmissible, unless there's a certain number of independent, trusted snapshot services that would archive each link too, and allow one to compare the decrypted archive. To achieve that, one would probably want to use an indirection for linking so that such archives do not have to archive everything that exists, but just those that go through said indirection. Then, the last question is how many independents you need to put trust in the verification, considering that motivated attackers could circumvent, say, the 3 most used archiving services. Or you could bind the archival of a link in a federated number snapshot services to a consensus algorithm. Then you'd ask yourself how many networks (which consist of multiple services) you'd trust.

sonoffett 10y ago
TLSNotary is one approach: https://tlsnotary.org https://tlsnotary.org, although I'm not certain it provides trusted timestamps.
- Natanael_L 10y ago
  It doesn't, but you can feed the content through a digital timestamping system
sbierwagen 10y ago
Hash it, store the hash in a blockchain: https://www.gwern.net/Timestamping https://www.gwern.net/Timestamping (One of the few things blockchain stuff is useful for)
- cm3 10y ago
  Yes, but how do you conclude without a doubt that what was hashed hasn't been tampered, leading a site to believe it's legal or malware-free content, while it actually isn't, and you've just been fed a shadow version of the real thing. We can reach for consensus algorithms, but ultimately without something like Van Jacobson's Content Centric Networking, it will be very hard to prove with enough confidence for a judge to make a ruling.
  - im4w1l 10y ago
    1. Because the hash is in the blockchain, certificates cannot be created retroactively. 2. The notary could make their archive public, so that any interested party could go through and verify that it is legit. 3. By using many different notaries the chance of any one of them being bad would be reduced.
    - cm3 10y ago
      Sure, and if the quorum consists of 6 members, and you managed to undermind 3 of them by making sure they're fed the same custom shadow content, you'll have a split.
      - iopq 10y ago
        You could literally put it in the Bitcoin blockchain. This will be spread among thousands of machines with cryptographic proof.
nickpsecurity 10y ago
The answer is mostly in the link I gave you. Here's an implementation: http://www.surety.com/digital-copyright-protection/prove-ownership.aspx http://www.surety.com/digital-copyright-protection/prove-own... You publish the hash chains with signature in places you can't retract. They use New York Times for instance. Far as their sigs, they can use one or more HSM's. One or more atomic clocks plus high-quality NTP for timing. Open-source client checks all of it right after it happens, maybe keeping local copy. Every other aspect is just standard INFOSEC. I dont have time to answer the last question right now. Off to work. :(
- cm3 10y ago
  But the last one is the most interesting to finally be able to trust the results. It's the same problem as faced in distributed systems, but in, say, NoSQL databases I haven't heard anyone consider the possibility of 4 of 6 members of a quorum to be undermined, and how to detect that.
pbhjpbhj 10y ago
>how do you prove that the archive (snapshot) is what was actually at the target at creation time // It's not what was at the sender though, surely, it's what gets sent to the client. Each client receives different content (at some level). If I paste a 'forbidden' link to my blog that I've never followed, and no-one ever views the page, surely that's not a crime??
- cm3 10y ago
  > Each client receives different content (at some level). Exactly, and now imagine how easy it is for someone motivated and with the right resources to meddle with that. Similarly, it's hard to prove a machine was used by its user to consciously navigate to an illegal link. There's too much automatism on the web and on networked machines, plus if we consider link shorteners, there are too many explanations against that would not allow admitting evidence like machine-x-owned-by-mike-miller-accessed-mp3forfree-artist-album-2016.pdf. It's the equivalent of me walking in a street where I didn't know drugs or weapons or humans were sold, and being arrested for just walking through there. In real life you have some clues where not to go, especially if you're from the town, but on the internet it's too easy to have your machine load a random illegal link or search alarming terms on Google without your consent.

by cm3 10y ago

sonoffett 10y ago

TLSNotary is one approach: https://tlsnotary.org https://tlsnotary.org, although I'm not certain it provides trusted timestamps.

Natanael_L 10y ago
It doesn't, but you can feed the content through a digital timestamping system

sbierwagen 10y ago

Hash it, store the hash in a blockchain: https://www.gwern.net/Timestamping https://www.gwern.net/Timestamping (One of the few things blockchain stuff is useful for)

cm3 10y ago
Yes, but how do you conclude without a doubt that what was hashed hasn't been tampered, leading a site to believe it's legal or malware-free content, while it actually isn't, and you've just been fed a shadow version of the real thing. We can reach for consensus algorithms, but ultimately without something like Van Jacobson's Content Centric Networking, it will be very hard to prove with enough confidence for a judge to make a ruling.
- im4w1l 10y ago
  1. Because the hash is in the blockchain, certificates cannot be created retroactively. 2. The notary could make their archive public, so that any interested party could go through and verify that it is legit. 3. By using many different notaries the chance of any one of them being bad would be reduced.
  - cm3 10y ago
    Sure, and if the quorum consists of 6 members, and you managed to undermind 3 of them by making sure they're fed the same custom shadow content, you'll have a split.
    - iopq 10y ago
      You could literally put it in the Bitcoin blockchain. This will be spread among thousands of machines with cryptographic proof.

nickpsecurity 10y ago

The answer is mostly in the link I gave you. Here's an implementation: http://www.surety.com/digital-copyright-protection/prove-ownership.aspx http://www.surety.com/digital-copyright-protection/prove-own... You publish the hash chains with signature in places you can't retract. They use New York Times for instance. Far as their sigs, they can use one or more HSM's. One or more atomic clocks plus high-quality NTP for timing. Open-source client checks all of it right after it happens, maybe keeping local copy. Every other aspect is just standard INFOSEC. I dont have time to answer the last question right now. Off to work. :(

cm3 10y ago
But the last one is the most interesting to finally be able to trust the results. It's the same problem as faced in distributed systems, but in, say, NoSQL databases I haven't heard anyone consider the possibility of 4 of 6 members of a quorum to be undermined, and how to detect that.

pbhjpbhj 10y ago

>how do you prove that the archive (snapshot) is what was actually at the target at creation time // It's not what was at the sender though, surely, it's what gets sent to the client. Each client receives different content (at some level). If I paste a 'forbidden' link to my blog that I've never followed, and no-one ever views the page, surely that's not a crime??

cm3 10y ago
> Each client receives different content (at some level). Exactly, and now imagine how easy it is for someone motivated and with the right resources to meddle with that. Similarly, it's hard to prove a machine was used by its user to consciously navigate to an illegal link. There's too much automatism on the web and on networked machines, plus if we consider link shorteners, there are too many explanations against that would not allow admitting evidence like machine-x-owned-by-mike-miller-accessed-mp3forfree-artist-album-2016.pdf. It's the equivalent of me walking in a street where I didn't know drugs or weapons or humans were sold, and being arrested for just walking through there. In real life you have some clues where not to go, especially if you're from the town, but on the internet it's too easy to have your machine load a random illegal link or search alarming terms on Google without your consent.