The Paris and San Bernardino terrorist attacks have brought encryption to the top of the agenda in Washington. The debate has often been described in very general terms: Should we or should we not allow back-door access to encrypted data for government agencies? But this level of generality obscures the fact that we have not one policy question, but two: How should we deal with data at rest, and how should we deal with data in transit?
What we mean by encryption
Before delving into the duality of the debate, we should clarify the context and the question. Encryption is the process by which a message can be encoded in such a way that only a recipient with the appropriate key can read it. If the encryption system is properly designed, nobody can learn anything about the message without the key. Thanks in part to US government standardization efforts, there are several well-regarded and highly secure encryption systems available to the public. Encryption is legally required for a wide range of purposes, including online banking, credit-card processing, health records, and more. High-quality implementations are built into every web browser and operating system.
In general, everybody understands that as a general technique, encryption is essential. The recent debate focuses on two much narrower questions:
- If a user wants to encrypt the stored data on their computer, should the manufacturer (or some other party) have a key?
- When two individuals want to communicate securely over the Internet, should some third party have the keys to decrypt their conversation?
The former question affects the government’s ability to retrieve data from a seized cell phone or laptop. The latter affects the government’s ability to wiretap. These questions pose different issues both technically and in terms of policy, and they should be considered separately.
Encrypted storage
The debate around encrypted data storage is largely between a few large vendors and law enforcement. Stored-data encryption is relevant only in cases where the government has physical access to an encrypted device (like a cell phone or laptop), and is thus an issue for law enforcement, not intelligence gathering.
Before 2014, the major mobile operating systems (iOS and Android) were set up so that, by default, the software vendors (Apple and Google) held the keys used to store data on the device. The purpose of storing the key was so that a user who lost their key could recover their data. Since the keys are chosen when the device is first used, this meant keeping them for months or years. Both vendors have now switched to a model where users, and only users, have the keys. It is this switch from provider-held to user-held keys that has aroused the ire of law enforcement officials. The anti-encryption vision, roughly speaking, seems to be that government pressure (or a statute) can undo this shift and restore a world in which vendors design their products so that they have copies of the relevant keys.
In all probability, this vision cannot be brought about. Even if the major vendors buckled, there are many minor ones – including freely available community-developed operating systems like Linux, where there is no vendor to compel and where anybody is free to produce their own version with whatever features they like. There is also add-on software that users can install for encryption, regardless of their base operating system. It is conceivable (though unlikely) that Apple and Google could be pressured into revising their systems. It is inconceivable that the government could stop determined users from encrypting their data.
The government’s initiative will primarily affect those least motivated by security: that is, it will affect ordinary Americans more than criminals with something to hide. The effects on ordinary Americans will be negative. A system in which manufacturers as well as users hold keys is necessarily less secure than one in which only users hold keys. Adding additional trusted parties to a system invariably increases complexity. Complexity is the bane of security engineering; every detail is a place where a mistake could be made or a trust assumption could be flawed. Consequently, a policy requiring vendors to keep keys will reduce the security of American citizens and businesses more than that of criminal suspects.
Encrypted communications
Turning now to communications encryption, a somewhat different set of issues arise. Here the topic is not just law enforcement, but also intelligence gathering. Furthermore, for these systems, there is a sharp distinction between the most popular systems and the rest.
The most popular communication systems, such as email, Facebook, or Google Talk, relay all communication through a server. When Alice sends a Facebook message to Bob, the message goes from Alice, to Facebook, to Bob. Both legs of the transfer are encrypted, but, importantly, they are encrypted separately. Facebook has a copy of the unencrypted message; this is necessary so that users can re-read or search through their past conversations. Thus there is limited benefit from the government pushing the major vendors into a redesign: the salient policy question is not encryption, but whether companies should be compelled to turn user data over to the government.
The major vendors are not the whole story, however. There are many less-well-known messaging systems, such as Off the Record messaging, that do support privately-encrypted communication. With these systems, when Alice and Bob talk, no third party sees the contents of their communication; the encryption is “end to end.”
Writing an operating system is a major challenge; writing an end-to-end encrypted chat program is the sort of thing that universities routinely assign as a course project. Given that there are thousands of developers up to the job, both within and beyond the United States, it is hard to imagine preventing users who want encrypted messaging from getting it. Here again, any sort of intervention is likely to only affect the people we are least interested in surveilling.
The true nature of the encryption debate
The “encryption debate” turns out to be a misnomer. The real debate is about two specific use cases: device encryption and communications encryption. In both cases, the government’s law enforcement and intelligence operations are hindered by the way modern technology products are designed. However, in neither case is there any commercially or technically plausible redesign that would address the government’s concerns. For better or worse, we are going to be living in a world with ubiquitous access to encryption. Law enforcement would be better advised to find ways to adapt to new realities, rather than engaging in a quixotic struggle to restore a vanishing status quo.
from AEI » Latest Content http://ift.tt/1QhbWcB
0 التعليقات:
Post a Comment