4 ms·

Unfortunately, each key will be poorly protected, because each group will rationalize that "it would be impossible for someone to get all those keys" while some

Unfortunately, each key will be poorly protected, because each group will rationalize that "it would be impossible for someone to get all those keys" while someone gets all the keys. The US government couldn't even keep the most sensitive information about its personnel secret, see the OPM breach: https://en.wikipedia.org/wiki/Office_of_Personnel_Management_data_breach https://en.wikipedia.org/wiki/Office_of_Personnel_Management... The Inspector General had warned Congress of "persistent deficiencies in OPM's information system security program" and nothing was done. This resulted in some really sensitive data being exfiltrated, probably some good blackmail info. The US government has not done a great job even protecting its own employees. The biggest problem here, though, is incentives. A government cannot be sued, or suffer any other negative impact, if it accidentally loses keys that lead to others' death, blackmail, or job loss. Too bad, so sad. If tomorrow it becomes illegal or unacceptable to do something that today is legal, a government can retroactively punish those who engaged in it. In short, the government has no strong incentive to protect the keys, only you do.

slg 6y ago
It sounds like we have already moved from this is an impossible request to this is a difficult and impractical request. Setting up a government competition with a prize to solve a hard problem seems reasonable.
- unionpivo 6y ago
  > It sounds like we have already moved from this is an impossible request to this is a difficult and impractical request. I am not sure how you got this from conversation above. Still sounds impossible to me. Problem as stated is, that if skeleton keys are broken, then everything everywhere gets broken. The fact that government has bad initiatives to keep keys secret, just means it will be easier for bad actors to steal them. But that is just a side argument.
  - slg 6y ago
    >I am not sure how you got this from conversation above. Still sounds impossible to me. When discussing encryption, many people say that this is an impossible problem to solve because of math. This conversation has shifted to being a problem about incentives. Math is impossible to change. Incentives are changeable. If the only problems here are the specifics of the scheme and the incentives of people involved, those are solvable. >Problem as stated is, that if skeleton keys are broken, then everything everywhere gets broken. Which is why you build redundancy into the system that would require multiple independent breaches and a system to reencrypt data in the event of a breach.
    - nytgop77 6y ago
      Attacker can record encrypted data before getting the masterkey. During usage, all pieces of master key will have to be brought together into one place (phisical or virtual), so eventually still only one breach is needed. If I encypt data using encryption without backdoor, then to illegitimately get clear text requires effort X, and it gives access to 1 persons data. If whole country encrypts data using backdoored scheme, then effort Y will give access to data of 100 000 000 people. It would seem that security of the backdoor should be 100mil times stronger then that of usual crypto.
    - unionpivo 6y ago
      > When discussing encryption, many people say that this is an impossible problem to solve because of math. This conversation has shifted to being a problem about incentives. > Math is impossible to change. Incentives are changeable. If the only problems here are the specifics of the scheme and the incentives of people involved, those are solvable. It's still a math problem in the end. Today if you want to break into someones encrypted data, you either have to break encryption algorithm (which is hard and by hard I mean even stuff we consider insecure like 3ple DES are still not broken) or you have to break each data individually. With skeleton keys that changes. Suddenly all the world secrets are one* key away. That is the big difference. Doesn't matter how you do it, it fundamentally changes the equation And you open up HUGE attack vector, by enabling bad actors to bribe/coerce people with access to skeleton keys. * Doesn't matter if you split it up, have different org have parts of keys etc.
- dwheeler 6y ago
  In the real world, impractical is the same as impossible. It's easy to enable a third party to read encrypted data, simply give them the key. That is not and never has been the problem. The problem is ensuring that only authorized uses occur. There is no algorithm that can determine good guys from bad guys. Keys are notoriously hard to keep secret. Any mechanism that tries to transmit them now adds a third party, and we have enough problems trying to keep things secure when they're only two parties. Storing them securely is highly improbable to occur in practice. We cannot even store simple data securely. The OMB breach showed that employees were unwilling to apply basic security practices to protect data about themselves... why would they securely store data when they won't be hurt? Splitting them helps a little, but it doesn't solve the fundamental problem. And this ignores the fundamental problem that groups like terrorists can simply use a different algorithm. When somebody says please think of the terrorist children, they assume that somehow they will use the bad encryption everyone else is stuck with. If someone is willing to violate one law, they will be willing to violate another.

slg 6y ago

It sounds like we have already moved from this is an impossible request to this is a difficult and impractical request. Setting up a government competition with a prize to solve a hard problem seems reasonable.

unionpivo 6y ago
> It sounds like we have already moved from this is an impossible request to this is a difficult and impractical request. I am not sure how you got this from conversation above. Still sounds impossible to me. Problem as stated is, that if skeleton keys are broken, then everything everywhere gets broken. The fact that government has bad initiatives to keep keys secret, just means it will be easier for bad actors to steal them. But that is just a side argument.
- slg 6y ago
  >I am not sure how you got this from conversation above. Still sounds impossible to me. When discussing encryption, many people say that this is an impossible problem to solve because of math. This conversation has shifted to being a problem about incentives. Math is impossible to change. Incentives are changeable. If the only problems here are the specifics of the scheme and the incentives of people involved, those are solvable. >Problem as stated is, that if skeleton keys are broken, then everything everywhere gets broken. Which is why you build redundancy into the system that would require multiple independent breaches and a system to reencrypt data in the event of a breach.
  - nytgop77 6y ago
    Attacker can record encrypted data before getting the masterkey. During usage, all pieces of master key will have to be brought together into one place (phisical or virtual), so eventually still only one breach is needed. If I encypt data using encryption without backdoor, then to illegitimately get clear text requires effort X, and it gives access to 1 persons data. If whole country encrypts data using backdoored scheme, then effort Y will give access to data of 100 000 000 people. It would seem that security of the backdoor should be 100mil times stronger then that of usual crypto.
  - unionpivo 6y ago
    > When discussing encryption, many people say that this is an impossible problem to solve because of math. This conversation has shifted to being a problem about incentives. > Math is impossible to change. Incentives are changeable. If the only problems here are the specifics of the scheme and the incentives of people involved, those are solvable. It's still a math problem in the end. Today if you want to break into someones encrypted data, you either have to break encryption algorithm (which is hard and by hard I mean even stuff we consider insecure like 3ple DES are still not broken) or you have to break each data individually. With skeleton keys that changes. Suddenly all the world secrets are one* key away. That is the big difference. Doesn't matter how you do it, it fundamentally changes the equation And you open up HUGE attack vector, by enabling bad actors to bribe/coerce people with access to skeleton keys. * Doesn't matter if you split it up, have different org have parts of keys etc.
dwheeler 6y ago
In the real world, impractical is the same as impossible. It's easy to enable a third party to read encrypted data, simply give them the key. That is not and never has been the problem. The problem is ensuring that only authorized uses occur. There is no algorithm that can determine good guys from bad guys. Keys are notoriously hard to keep secret. Any mechanism that tries to transmit them now adds a third party, and we have enough problems trying to keep things secure when they're only two parties. Storing them securely is highly improbable to occur in practice. We cannot even store simple data securely. The OMB breach showed that employees were unwilling to apply basic security practices to protect data about themselves... why would they securely store data when they won't be hurt? Splitting them helps a little, but it doesn't solve the fundamental problem. And this ignores the fundamental problem that groups like terrorists can simply use a different algorithm. When somebody says please think of the terrorist children, they assume that somehow they will use the bad encryption everyone else is stuck with. If someone is willing to violate one law, they will be willing to violate another.