Let me be clear first and foremost: I do not think installing a backdoor in security algorithms is a good idea. They undermine the trust in the software and in the company that provides the service. That being said, I do agree that encryption provides a certain measure of protections for criminal...
@A.Darwin the secure channel is HTTPS or equivalent security, with only the most recent algorithms supported. This scheme is for encryption other than HTTPS.
1: Political: you want an NGO whose sole reason for existing is to provide information to the government; is it a multi-national NGO? If so how does it determine which countries' law enforcement agencies to accept requests from? I certainly wouldn't use that service if every law enforcement officer in the world had access. If the NGO is within the jurisdiction of a single country, then how will it keep its independence / avoid court-orders?
2: Technical: Why do you need some crazy new encyption scheme with two decryption keys? Why do you need to delete the second key from the client? What's wrong with regular AES and just send a copy of your key to the NGO like a regular key escrow?
There's a major flaw in this scheme. Assume for a second that you're a terrorist. You have a bunch of encryption cyphers that you can choose from. You know that this one sends keys to the authorities, and this other one doesn't. Which will you pick?
This proposal seems almost identical to the infamous "Clipper chip" (except that had a single backdoor key, if I remember correctly). There's plenty of discussion online, showing why people were up in arms about it.
You're glossing over the big gaping security hole -- how to do you effectively airgap a database that accepts billions of real-time key updates daily, and has to be able to respond to hundreds, if not thousands of daily escrowed key requests every day? That database will become the target of very capable hackers, including well funded government agencies. Even if you use people as the air-gap, it's easy to compromise a person through threats or money.
@Agent_L - you can still have an "air gapped" computer that accepts automated, electronic input. As an extreme example, you could have the public system print encryption keys in barcodes on paper, then have the protected database computer scan in the barcodes -- no less secure (and much more accurate) than having a human type in the keys.
If this algorithm becomes well adopted, you end up creating one of the, if not the, most valuable databases in the world. Now you are going to have to protect it from everyone. We are talking on the level that there are groups, government and otherwise, who will threaten the NGOs employees and families to get a copy of that database. This weakens security by creating a single target where resources will be focused.
@Johnny another solution is a Data Diode (yes those exist) which uses fiber optics and some protocol fudging, to allow only one-way communication, typically into a highly secured system designed specifically to gather data.
If you wanted to do this, you would just use something like PGP where the secret never needs to be securely transferred to anyone. Everything is open so that would eliminate an attack vector. The master key used for escrow can be airgapped before the public portion of the key is distributed. The problem is that anyone can write a non-compliant implementation. Sure, Apple or Google can use such things for their own on-device encryption, but that can be hacked to simply fail to use the escrow-key and the device will still function. I'd probably hack it if given the chance to limit intrusion.
Could you guarantee the incorruptibility of employees at this NGO? There'd be a lot of people, criminals, corporations, and governments alike who would pay very high prices for this data.
Discussion on question by Nzall: Woul…
Discussion on question by Nzall: Woul…