Encryption — the cryptographic transformation of data (called “plaintext”) into a form (called “cipher text”) that conceals the data’s original meaning to prevent it from being known or used
Strong encryption — encryption that cannot be decrypted through reasonably accessible computational methods or algorithmic flaws
Weak encryption — encryption that can be decrypted through reasonably accessible computational methods or algorithmic flaws; additionally, also considered weak is strong encryption that has a built-in bypass capability (commonly referred to as a “backdoor”)
End-to-end encryption — a system of communication in which the only people who can read the messages are those who are communicating; no eavesdropper can access the cryptographic keys needed to decrypt the conversation — not even a company that runs the messaging service.38
Encryption is a fundamental technology for security. The key policy question for understanding how to treat encryption is: who should be able to access sensitive data and communications? Encryption is necessary to ensure that sensitive data and communication are not accessed by bad actors. However, encryption can also be used by bad actors to shield communications from law enforcement. In the last few years, encryption has become increasingly salient as the private sector has invested in differentiation on the basis of user-friendly encryption to secure increasing amounts of personal (sensitive) data. On the other hand, some policy-makers increasingly insist on weaker encryption. On this particular policy topic, minimal opportunity for a middle ground exists. An encryption algorithm either obfuscates data or it does not. And algorithms cannot divine the intentions of those seeking to circumvent them.
To help frame the policy discussion for encryption, it is helpful to think about policy on two (related) axes: who has access to encryption (e.g. the public or private sector) and what type of encryption do they have access to?
- Two analytically helpful (though not necessarily technological) types of encryption exist: weak and strong. In some circumstances, one entity’s strong encryption might be considered weak by another. For example, encryption that may be used by consumers may not be sufficiently strong for defence ministries. However, the policy-relevant deliberation is not about technological standards but about the choice of whether to allow access to data by an entity other than the user regardless of the underlying technology.
- For the purposes of encryption, two entities are fundamental: the public sector and the private sector. Again, within the public and private sectors, different subgroups might utilize unique technologies around encryption but the question is about access and its purpose.
Requiring differential encryption (particularly for the private sector) has significant risks and benefits:
- If the private sector is required to use weak encryption, then bad actors may potentially obtain access to customer data. The greater the amount of customer data that companies are allowed to capture, the greater the potential damage associated with bad actors accessing this data.
- Alternatively, if the private sector is allowed to use strong encryption, law enforcement may be hampered in its efforts to access data relevant to preventing crime or investigating its aftermath. Strong encryption also stymies intelligence agencies in the collection of data.
- A policy of weak encryption for the private sector may be unstable in the long run when coupled with allowing private-sector access to greater amounts of personalized data as the cost of bad actors defeating encryption may become greater than the value of thwarted criminal activity. Requiring private-sector vendors to develop encryption workarounds may also impose non-trivial costs, not only in terms of customer risk but in terms of software development and engineering costs.
To understand the concerns and risks around weak encryption, it may be helpful to reason through a commonly debated example: if policy-makers insist on a way to bypass a popular messaging application — regardless of whether that application can currently support such a bypass — bad actors will try to move to other applications to mask communications. If policy-makers then insist on achieving access to communication at a more fundamental technical level (e.g. at the level of an operating system), so that bad actors have no choice but to (at least in the short run) risk exposing their communications, other bad actors in cyberspace will have even greater incentives to pierce that encryption. After all, operating-system-level privileges are valuable to law enforcement/intelligence for the same reason as they are to adversaries. This logic can continue down the technical stack but, in general, the more inescapable the bypass capability sought, the more attractive that bypass becomes to adversaries.
Policy model: Encryption
Key values trade-offs created by encryption policy choices
Increasing adoption of encryption for internet traffic
Case study: Encryption and business model disruption
Some commentators have encouraged the adoption of end-to-end encryption to ensure that data-at-rest and data-in-transit remain secure from unauthorized access and disruption.39 Responding to some policy impetus (e.g. General Data Protection Regulation in Europe), companies are increasingly implementing end-to-end encryption. However, it is important to acknowledge that end-to-end encryption threatens business models premised on monetizing individual-specific attributes or using individual-specific data for advanced analytics (including personalized AI and machine learning). In adopting end-to-end encryption, companies limit the ability to inspect communications. In so doing, companies limit the inferences they are capable of making regarding an individual — whether that individual is likely young or old, male or female, etc.
As a consequence, the ability to then sell an adjacent service (e.g. advertising or an AI-based service) targeting individuals based on revealed attributes is greatly diminished. In the case of advertising, the explanation is reasonably straightforward: the ability to target and measure the impact of an ad is paramount for marketing teams to articulate a value proposition to negotiate for budgetary authority. In the case of advanced analytics, the impact of encryption of those business models is a bit more subtle. In general, advanced analytics require the aggregation of both data and computing typically limited to accessing a cloud resource. However, some companies have experimented with using mathematical models capable of inference that never leaves an endpoint or is obfuscated when in transit to cloud resources, such that data remains anonymized and encrypted. Nonetheless, the trade-off is clear: encryption obfuscates data that could otherwise form the basis of the richer inference underlying personalized AI/machine learning-based services.
Case study: Quantum computing and encryption
Recently, companies have begun to commercialize access to quantum computing. In light of this access, some commentators have raised concerns about the ability to encrypt data in light of these new computational techniques. These concerns are somewhat alarmist; while it is true that the current mathematical algorithms underlying much of the encryption used by the public and private sectors would be vulnerable to these new computing techniques, already efforts are in place to develop new algorithms to thwart quantum computing. The U.S. National Institute of Standards and Technology (NIST) has begun developing so-called “post-quantum” cryptographic techniques.40
Connecting policy to values
Encryption policy choices sharply implicate a number of values and, depending on policy choice, create key trade-offs between these values. The value trade-offs surfaced by encryption policy are very similar to those raised by zero-day policy. It would be inconsistent, for example, to argue for pervasive stockpiling of zero-days while insisting on strong encryption for the private sector — vulnerabilities in encryption make it weaker:
- Security may theoretically be improved in two scenarios: one in which the private sector has weak encryption or strong encryption. One line of thinking, more associated with law enforcement, is that governments can provide greater security for citizens and firms by accessing communications that may be used by criminal elements in a weak encryption policy.
- Another line of thinking suggests that strong encryption is likely to promote greater security, such that bad actors do not discover and exploit backdoors to encryption against a country’s citizens and firms. A weak encryption policy is more likely to mitigate the risk of coordinated and broad threats, as law enforcement access will presumably deter would-be conspirators and facilitate rapid criminal response. A strong encryption policy is more likely to mitigate the risk of bad actors exploiting sensitive information.
- The economic value associated with different encryption policy scenarios is a function of a few effects. For instance, greater security achieved through weak encryption and strong encryption is associated with fewer damages arising from cyberincidents. However, in the case of weak encryption, this must be weighed against the costs of these same backdoors being used against a given country’s citizens and firms, as well. Additionally, some observers have noted that actions deteriorating trust in ICT create substantial intangible costs in terms of diminished ICT adoption.
- Privacy is also impacted by choices in encryption policy. In a weak encryption policy, the improvement in security is premised on decreased privacy. While, in most cases, presumably decreased privacy will be limited to suspected criminals, the risk is that the confidentiality of non-adversaries will also be compromised.
- Encryption policy impacts the accountability of both the public and private sectors. It is incumbent on the private sector to adopt sufficiently strong encryption to thwart adversaries. However, if the private sector is mandated to use weak encryption, the public sector has greater accountability to ensure that backdoors remain undiscovered and that increased access to communications is closely monitored and also productively used by law enforcement.