THE METAPHOR IS THE KEY: CRYPTOGRAPHY,
THE CLIPPER CHIP, AND THE CONSTITUTION

A. Michael Froomkin

Document information and copyright notice

[Page n] references relate to the pagination of the printed version.

Click here to jump to a specific page:

To table of contents

It is old news that common-law legal reasoning is both analogical and taxonomical,{663} and that metaphor is a powerful tool for both.{664} Nevertheless, the observation that "[t]he power of a metaphor is that it colors and controls our subsequent thinking about its subject"{665} is particularly relevant and powerful when the law encounters a new technology.{666} The law's first reaction to a[Page 861]new technology is to reach for analogies and to explain why the new technology can be treated identically to an earlier technology. Railroads, for example, could be slotted into the existing legal categories created to deal with highways, collisions, and freight tariffs.{667} In contrast, airplanes--a technological advance on the same order as the railroad--required a significant change in the law because to be useful the airplane must fly over land, a classical trespass, without a right of way.{668}

In the case of cryptography, as with other new technologies, the dominant mode of judicial and perhaps political judgment is likely to be classification by analogy and argument by metaphor. Given that in recent years Big Brother seems to have inspired less fear in any branch of government than has Big Drugs, the selection of the right metaphor is critical.

Four metaphors seem particularly likely to appeal to the courts, which can be summarized under the rubrics of "car," "language," "house," and "safe." These four metaphors reflect two fundamentally different characterizations of cryptography. Both "car" and "language" characterize cryptography as part of the means used to transmit the message. In this view, an encrypted message is simply another communication, one which can best be understood as a special case of the general rules regulating communications. In contrast, "house" and "safe" treat cryptography as something that happens before the message leaves the sender. Both "house" and "safe" suggest that the proper approach is to start with the sender's decision to encipher the message in order to exclude unwanted recipients, and then explore the implications of this choice for the government's ability to monitor communications. The differences[Page 862]among these metaphors go beyond putting a positive or negative gloss on encryption; they amount to different definitions of the nature of the thing itself. Interestingly, both the general metaphor of communication and the metaphors of exclusion are sufficiently indeterminate to permit the introduction of more specific, and contradictory, metaphors that support both sides in the mandatory EES debate.

1. Focus on Communication

The ubiquity of the conduit metaphor suggests that encryption, if understood as a communications conduit, stands the best chance of being accepted by the courts, Congress, and the public. The same ubiquity, however, also means that different analogic embodiments of the same general metaphor can lead to diametrically opposite views of the constitutionality of a hypothetical ban on unescrowed cryptography. Indeed, if the encrypted message is seen[Page 863]as a mobile message unit--a "car" on the information superhighway--then mandatory EES appears far less troubling than if the encrypted messages are analogized to language itself.

a. "Car"--How Messages Travel

The LEAF feature in Clipper and Capstone makes checkpoints for compliance with mandatory key escrow particularly easy to implement, in a way that other forms of escrowed encryption might not. Telephone and e-mail messages can be intercepted and copied at random, or according to some pattern, then decrypted with the family key. If there is no LEAF at all, the message is clearly in violation of the mandatory escrow rule. If there is a LEAF, although the text of the message remains inaccessible without recourse to the escrow agents,{672} public servants can check whether the message has been encrypted with an EES-compliant device because the chip serial number is supposed to contain a recognizable string identifying the manufacturer.{673} If this string is instead random, law enforcement knows it has encountered a spoofed LEAF.{674} The beauty of this system from a constitutional perspective is that the intrusion on privacy rights is relatively small. Law enforcement does not need to decrypt actual messages without a search warrant.

The car metaphor also provides law enforcement with a solution to a timing problem. The information highway patrol may not want to wait to obtain a warrant to find out whether a message's cryptography is registered. Just as police who stop cars on the highway have authority to conduct a search without a warrant,{675}[Page 864]the same might be done on the information highway, and for similar reasons. Waiting for a warrant takes too long--the car is gone; the message is gone.{676}

The car metaphor leads naturally, at least from the law enforcement perspective, to random traffic stops. If the analogy is to vehicular checkpoints, the examination of the LEAF can easily be characterized as a minimally intrusive investigation in a way that it cannot be if the focus is on the act of encryption as something occurring in the home or office. In the case of real cars, the Supreme Court applies a reasonableness analysis to vehicle stops, weighing the gravity of the public concern, the degree to which the seizure advances that concern, and the interference with the individual's liberty. In United States v. Martinez-Fuerte,{677} the Court held that limited seizures of vehicles for questioning of the occupants could be constitutional even absent a particularized suspicion due to the government's compelling interest in reducing the influx of illegal aliens, the Border Patrol's inability to do so at the border, and the minimal harm to the traveller.{678} In Michigan Department of State Police v. Sitz,{679} the Court found that even if only 1.5% of the drivers stopped were drinking, the governmental interest was great enough to justify the intrusion of a highway checkpoint.{680} Sitz used a three-prong test{681} requiring (1) a[Page 865] weighing of the gravity of the public concerns served by the seizure, (2) some empirical evidence that the program was effective,{682} and (3) a weighing of the severity of the interference with individual liberty. {683} The car metaphor will no doubt have appeal beyond law enforcement if only because it is already commonplace to talk about the Internet as a highway, complete with "on-ramps," "fast lanes," and "maps." Perhaps it is time to abandon these convenient phrases. If the car metaphor prevails, there will be far fewer constitutional rights in cyberspace than if any other metaphor comes to dominate.{684}

b. "Language"

Although a cipher resembles a foreign language, it is not[Page 867]identical. No one can speak in DES without mechanical aids, and no one can understand a DES-encrypted message if they do not understand the language of the plaintext. Cryptologist Dorothy Denning argues that these differences are so great that in some important sense encrypted speech "is not speech."{689} As Denning notes, languages have semantic blocks such as words, phrases, sentences, or ideograms that "carry" meaning and that can be manipulated, but ciphertext has no such blocks.{690} Also, Denning argues, all languages share the property that thoughts, emotions, beliefs, requests, offers, and concepts can be expressed without knowledge of any other language.{691} In contrast, ciphertext not only needs to be decrypted to be understood, but the recipient must understand the language of the plaintext in order to comprehend the message.{692}

Existing First Amendment jurisprudence provides guidance that helps determine whether these differences between enciphered communications and ordinary language speech should be considered legally significant.{693} The First Amendment protects communicative acts, not specific modes of communication.{694} The First[Page 868]Amendment protects many communicative acts that do not use words including photographs, pictures, nude dances,{695} and silent protest.{696} Although they do not use words, these protected communications do have messages. They have semantic units and referents of the type described by Roland Barthes and others.{697} Indeed, the Supreme Court recently reaffirmed the importance of the connection between an unspoken assertion of the speaker's identity and the communicative content of the message. In City of Ladue v. Gilleo,{698} the Court held that an ordinance prohibiting the display of signs in the front yard of a house violated the resident's right to free speech as "the identity of the speaker is an important component of many attempts to persuade."{699} Similarly, there is no reason to believe that the First Amendment is concerned with the degree to which a communicative system depends upon another or stands alone.{700} If the First Amendment protects works of art so obscure that they can only be understood by their creator,{701} it can equally well be applied to protect encrypted speech. Thus, from a First Amendment standpoint, the differences between ciphertext and ordinary language identified by Denning are irrelevant to[Page 869]whether a communication is protected speech.

From the viewpoint of the language metaphor, the most troubling difference between ciphertext and classic protected speech is that, in the case of ciphertext, a mechanical aid is required to both create and comprehend the message. This difficulty should not be exaggerated--metaphors, after all, are invoked when things are not identical, not when they are precisely the same. It is unlikely that the mechanical aid would appear as troublesome if, instead of a cryptographic system, the device in question were a prosthetic device, such as a voice synthesizer or an eye-movement operated computer.{702} In the case of these prosthetic devices, one would not expect to find arguments that the communication emanating from the device, or the signals used to operate the device, were any less entitled to First Amendment protection than ordinary speech. Again, the ciphertext example is not identical because the parties to the communication presumably have alternate forms of communication available to them, but this just underlines the fact that it is not the mechanization of the communication that ought to be the stumbling block. As new technologies such as voice recognition become commonplace, one would not expect to see arguments that the speech is somehow less protected while in the binary form that intermediates between sound waves and text, even if the author could have used a pencil instead of speaking to a computer. In any event, because a work of art requiring computer animation or computer-aided virtual reality would depend critically on the assistance of a machine, it would clearly be entitled to the same First Amendment protections as a painting.

Encrypted speech is not exactly an ordinary language, but it is similar to what we ordinarily mean by language. Moreover, most of the differences are not the ones that usually matter in the First Amendment context. The most significant difference between encrypted speech and ordinary speech is the role of a machine. Indeed, the encrypted communication much resembles the telephone call, as voice is translated into a ciphertext (an electrical or fiber-optical signal) that is transmitted to the recipient who then[Page 870]decrypts it (plays it on a speaker). Telephone calls are subject to wiretaps not because a machine is involved in the communication, but rather because once public servants have obtained the appropriate warrant, the signals are in the same position as unmechanized speech inside the home.

Rejection of the language metaphor might lead to undesirable consequences. If the government were able to require that users of strong cryptography ensure the government's ability to decrypt their messages, it might be only a small step to imposing limits on the use of languages other than English.{703} Would the last two speakers of a dying language be required to provide translations of their conversations if the government charged that they were conspiring in it? Such a rule would be as difficult to apply as it might be difficult to distinguish among impenetrable slang or strange accents, a language, and a code.

2. Focus on Exclusion

Simply putting something into a safe does not, however, ensure that it is beyond the law's reach. It is settled law that a criminal defendant can be forced to surrender the physical key to a physical safe, so long as the act of production is not testimonial.{706} Presumably a similar rule compelling production would apply to a criminal defendant who has written down the combination to a safe on a piece of paper. There appears to be no authority on whether a criminal defendant can be compelled to disclose the combination to a safe that the defendant has prudently refrained from committing to writing, and in Fisher v. United States,{707} the Supreme Court hinted that compelling the disclosure of documents similar to a safe's combination might raise Fifth Amendment problems.{708} Perhaps the combination lock problem does not arise because the police are able to get the information from the manufacturer or are simply able to cut into the safe. These options do not exist when the safe is replaced by the right algorithm. Although brute-force cryptography is a theoretical possibility,{709} neither safe cracking, nor number crunching, nor an appeal to the manufacturer is a practical option when the armor is an advanced cipher. The recently released Federal Guidelines for Searching and Seizing[Page 872]Computers{710} suggest that "[i]n some cases, it might be appropriate to compel a third party who may know the password (or even the suspect) to disclose it by subpoena (with limited immunity, if appropriate)."{711}

Even if ciphertext is analogous to a document in an uncrackable safe whose combination has never been written down, there are important differences between a paper in a vault and an encrypted e-mail. A safe is a container into which people put things and take them out again, preserving the contents over time by keeping unauthorized persons from having access to them.{712} Ordinarily, a safe stays put. E-mails usually move around.

Current law on moving containers is not very friendly towards privacy. The Supreme Court has stated that "some containers (for example, a kit of burglar tools or a gun case) by their very nature cannot support any reasonable expectation of privacy because their contents can be inferred from their outward appearance."{713} That, at least, can never be the case with an encrypted message, because the external appearance of ciphertext gives no clue as to its content.

The moving safe begins to look a little like luggage.{714} Intuitively, the privacy interest in a safe seems greater than the privacy interest in luggage, which is perhaps fortunate because the privacy interest in luggage has been shrinking towards the vanishing point.{715} Ordinary luggage can be searched without a warrant upon reasonable suspicion.{716} There appear to be no recent[Page 873]reported cases of police forcing a combination lock on luggage without either consent,{717} a warrant, or an alert from a drug-sniffing dog.

The privacy interest in locked luggage is insufficient to protect the owner against brief detentions of her property, in order to permit a dog sniff.{718} According to the Supreme Court, the sniff is not a Fourth Amendment "search" because the suitcase remains closed during the test, the dog discloses only the presence or absence of narcotics and cannot reveal the contents of the suitcase, and the "canine sniff" is the least intrusive method of ascertaining the presence of narcotics in the baggage.{719}

A sniff has some similarities to the investigation of a LEAF--low intrusion, bag/message remains closed/encrypted during investigation, and the investigation discloses nothing else about the contents of the bag/message. Like a dog alert, detection of an invalid LEAF could tell police that the message has been encrypted with unescrowed cryptography. Unlike dog alerts, LEAF-sniffing will prove generally unreliable because preencrypting a message with another cipher will hide the contents while presenting a valid LEAF for the world to see.{720}

Overall, the safe analogy is appealing. Unfortunately, it either[Page 874]maps the problem onto an equally unsettled area of law or collapses back to another form of the conduit metaphor.{721} It also has the moderate defect of being vulnerable to technical changes, although for the foreseeable future high-level encryption will remain far, far easier than brute-force decryption. It may be, however, that despite the potential for instability after the next technological or cryptographic revolution, the absence of law relating to combination locks without written combinations (the case most analogous to a strong cipher with a secret passphrase) creates an opportunity to make new law unencumbered by the baggage of the Supreme Court's luggage precedents. There is no obvious reason why a person's privacy interest in the contents of a safe, or a ciphertext, should decrease sharply because the object is in transit, and it would not be difficult to have the law reflect that reasonable expectation.

b. "House"--Where Messages Come From

The Supreme Court's first encounter with wiretapping produced a five-to-four decision holding that a wiretap was neither a search nor a seizure because it took place outside the home and did not interfere with the transmission of the message. "The reasonable view," Chief Justice Taft announced, "is that one who installs in his house a telephone instrument with connecting wires intends to project his voice to those quite outside"; having left the sanctity of the home, those messages and those wires "are not within the[Page 875]protection of the Fourth Amendment."{724} The majority's first encounter with the law enforcement aspects of the telephone treated it as a familiar thing: an instrument to send a message out into the world, to meet whatever fate might befall it once it was outside the constitutionally protected zone of the home--a protection that relied in large part on the homeowner's property interest in the residence.{725} Justice Brandeis's dissent relied instead on Boyd's holding that the Fourth Amendment prevents the government from forcing a person to produce an incriminating document.

Katz abandoned the idea that the test for Fourth Amendment protection rested on location. "[T]he Fourth Amendment," Justice Stewart wrote, "protects people, not places,"{726} and it thus protected a conversation originating in an enclosed public telephone booth. Or rather, as Justice Harlan put it in his concurrence, the Fourth Amendment protects a person who has "exhibited an actual (subjective) expectation of privacy" and the expectation is "one that society is prepared to recognize as `reasonable.'"{727} Justice Stewart cautioned that "[t]o read the Constitution more narrowly is to ignore the vital role that the public telephone has come to play in private communication."{728} Dissenting in Katz, Justice Black accused the majority of choosing "to rely on their limited understanding of modern scientific subjects in order to fit the Constitution to the times and give its language a meaning that it will not tolerate."{729} His complaint was that Justice Harlan's concurrence, and by implication the majority's opinion, argued that it was "`bad physics'"{730} to maintain the rule originating in Olmstead v. United States that electronic eavesdropping was not a search.{731} Justice Black believed that the Fourth Amendment's protection of the "right of the people to be secure in their persons, houses, papers, and effects, against unreasonable searches and seizures" connoted the idea of "tangible things with size, form, and weight, things[Page 876]capable of being searched, seized, or both" and that an overheard conversation, even on a wire, was none of these.{732}

In fact, it was Justice Black's physics that were faulty. Electrons have size, form, and mass, as do digitized and encrypted messages. Yet despite its questionable physics, Justice Black's legal conclusion appears to be gaining support: since, Katz, the Fourth Amendment, and its emanations have been read more and more narrowly. The Court has examined expectations of privacy that often seem greater, and more objectively reasonable, than those of a telephoner in a public phone booth, but has nonetheless found those expectations--when held by guilty parties--to be unreasonable.{733} Similarly, Boyd has been whittled away to the point that what vitality it retains is limited to personal, noncommercial papers, and even that is now in doubt. The rationale for Boyd's original holding has been effectively abandoned.{734}

In the place of Katz and Boyd, the Supreme Court has substituted an anemic form of the property theory of the Fourth Amendment that animated the majority in Olmstead and Justice Black's dissent in Katz, a theory that in its new form rarely seems to extend outside the curtilage of the home.{735} Although not stated as a general principle, as a practical matter the Katz test has come to depend on the objective reasonableness of an expectation of privacy,{736} and the Court has routinely turned to legal or administrative sources to define the parameters of reasonableness{737}-[Page 877]except when it ignores them.{738} Thus, for example, once trash is placed on the curb to be picked up, the property interest in it is gone, and the trash is up for grabs.{739} If it is lawful to fly over a property, it is objectively unreasonable for the owner to expect that the property was safe from aerial inspection regardless of the frequency of such flights or community standards of reasonableness, whatever those may be.{740} But despite the trespass by the observer, there is no reasonable expectation of privacy against ambulatory police intrusion into an "open" field, though surrounded by woods, behind a locked gate with a "No Trespassing" sign, because the field is open to view.{741}

Katz still applies to its facts, which involved a wiretap of a telephone call from an enclosed area. And the home clearly retains a special status, because in United States v. Karo{742} warrantless monitoring of a beeper placed by a government informant became unconstitutional at the point where it "reveal[ed] a critical fact about the interior of the premises that the Government . . . could not have otherwise obtained without a warrant."{743} But given the Court's record with reasonable expectations, the reason that Katz is still good law seems to have more to do with the existence of Title III than any constitutional principle. It has come to the point where the citizen's privacy interests might be better protected if the[Page 878] discredited property-based theory of Olmstead were revived instead of reviled.{744} Coincidentally, the development of mass-market strong cryptography means that the property theory, combined with a computer or a scrambler telephone, produces a level of constitutionally protected communications privacy comparable to the original Katz standard.{745} Today, with the right cryptography, it no longer matters as much if Katz is narrowed or overturned, at least as far as communications privacy is concerned.{746} Now there is something that the citizen can do, inside the home, before a message is sent out on its vulnerable journey throughout the world.

The value of the legal protections that can pertain to the home if based on a property theory like that in Olmstead should not be overstated. So long as we hold to current conceptions of the home, as a physical structure with walls and sometimes a curtilage too, the interconnections between homes will continue to be classified as "outside" the "house." As a result, regulations, including a ban on the use of unescrowed strong cryptography in communications that leave the house, remain a real possibility.

The "house" metaphor may provide some protection against the complete reversal of Boyd, depending on whether a court could compel the production of a key that had not been committed to paper. If the court were unwilling to do this, say on Fifth Amend[Page 879]ment grounds, strong cryptography would provide a nearly unbreakable means of protecting one's private papers stored in the home computer.{747}

The decision to classify the cryptographic key as akin to a private paper located in the home also may have interesting legal consequences as the idea of the home evolves. Assumptions about space, about what constitutes the "inside" of a home or an office, may need to be reexamined:

[I]n the era where people work for "virtual corporations" and conduct personal and political lives in "cyberspace," the distinction between communication of information and storage of information is increasingly vague. The organization in which one works may constitute a single virtual space, but be physically dispersed. So, the papers and files of the organization or individual may be moved within the organization by means of telecommunications technology. Instantaneous access to encryption keys, without prior notice to the communicating parties, may well constitute a secret search, if the target is a virtual corporation or an individual whose "papers" are physically dispersed.{748}

C. The Power to Choose

A court, or any other interested party, that might be called upon[Page 880]to select among competing metaphors will naturally be concerned about where they might lead. Widespread cryptography may have social implications that are difficult to predict. Strong privacy may not be to everyone's taste. Secret identities will protect the anonymous hate mailer, the drug dealer, the purchaser of seedy movies, the congenitally shy, and the electronic political pamphleteer with equal efficacy. The implications of anonymous transactions for taxes, product liability, and copyright, to name only a few, remain to be worked out.

Sissela Bok hypothesizes a society in which "everyone can keep secrets impenetrable at will. All can conceal innocuous as well as lethal plans, the noblest as well as the most shameful acts, and hatreds and conspiracies as much as generosity and self-sacrifice. Faces reveal nothing out of turn; secret codes remain unbroken."{750} Although Bok recognizes that some version of such a society "might develop precisely in response to the felt threat from increased [information] transparency,"{751} she views such a society as clearly undesirable because "[i]t would force us to disregard the legitimate claims of those persons who might be injured, betrayed, or ignored as the result of secrets inappropriately kept."{752} Strong protection of cryptography may lead exactly there.

However, the absence of the refuge of digital anonymity may be worse. As identifying data on each of us becomes more voluminous and more easily accessible to government and to private parties,{753} our lives are changed, and not necessarily for the better. Indeed, although law enforcement agencies believe they benefit greatly from their electronic eavesdropping capabilities, it is unclear whether society as a whole enjoys a net benefit when one considers both past abuses and the possibilities for future abuses. To foreclose an option that would give the lie to Justice Douglas's dystopian warning that "[w]e are rapidly entering the age of no privacy, where everyone is open to surveillance at all times; where there are no secrets from the government"{754} would be a communal confession[Page 881]of lack of trust in our fellows and in ourselves. It is chilling to think we are fully capable of making this confession, and that we may even deserve it.

In making legal judgments about the Constitution and cryptography, one should keep in mind what is possible and what is not. This means that one should consider both the extent to which long-held ideas about the meaning of the Bill of Rights need to be rethought in light of technological changes{755} and that some prohibitions simply are unenforceable. Just as the ITAR have failed to stop the spread of strong cryptography abroad (and the development of indigenous substitutes abroad), so too would any attempt to ban unescrowed cryptography be doomed to failure.{756}

Privacy, constitutional law, and law enforcement are not games. It is unsettling to think that one's rights may turn on the extent to which people are able to find a technological means to defeat what would otherwise be legitimate government action. The good news is that technological change can provide an opportunity to rethink, and perhaps increase the coherence of, some constitutional doctrines. When technology changes social realities, interpretations [Page 882]of the Constitution should recognize the change to the maximum extent the text permits. Thus, although we might be better off today with the Olmstead standard than into what the Supreme Court has turned Katz, we should not lose sight of the fact that Olmstead was wrongly reasoned. Chief Justice Taft's formalistic unwillingness to recognize that a telephone call was different from a message on a billboard or a shouted conversation on a busy street is not an attitude that deserves emulation.

The bad news is that sometimes a technological development is unstoppable. It may be that the Administration intends Clipper only as a delaying action.{757} It may be that a future administration, acting for what seem to be good reasons at the time, will attempt a form of cryptological Prohibition. If so, it will fail as Prohibition failed, as the War on Drugs is failing, and as the ITAR are failing with respect to cryptography.{758}

Conclusion

The executive branch's primary concern has been to accommodate the interests of banks and others who require strong cryptography, while also preserving to the greatest extent possible law enforcement and intelligence capabilities. Noncommercial social implications of cryptography have received relatively little attention. The private sector's motives are more difficult to summarize, but[Page 883]there has clearly been a demand for cryptographic products, and this demand is expected to grow rapidly.{760}

The executive branch's desire to maintain its ability to eavesdrop on electronic communications at will has driven it to abuse the technical standard-setting process. By manipulating the FIPS procedure, the Clinton Administration has achieved its initial objective of promulgating a standard that is insulated from any meaningful public comment and immune from judicial review. The Administration thus hopes to create a de facto rule where it lacks the statutory authority to create a rule de jure. Despite the seemingly underhanded aspects of the executive branch's behavior, there is no clear evidence that it has failed to comply with existing laws or the Constitution. There is, however, room for doubt, as some of the critical information regarding whether NIST retains its statutorily mandated independent judgment is classified. Congress would be well advised to reassure itself and the public that NIST has complied with the Computer Security Act's requirement that it not delegate decision-making to the NSA. If the NSA is calling the shots, firm persuasion, and perhaps corrective legislation, will be required.

The Administration hopes to coerce acceptance of an escrowed encryption product through its vast purchasing power, but whether this hope can be realized remains unclear. If this attempt fails, the next step may be to seek legislation requiring users of strong cryptography to allow the government some form of access to their cryptographic keys. If, despite what currently seems to be a prevailing opposition to even voluntary key escrow, such a bill were nonetheless to become law, mandatory key escrow would create serious constitutional problems that the courts would have to resolve.

Under current law, the judicial reaction to a hypothetical mandatory key escrow statute would be limited primarily to a balancing test analysis, although private noncommercial users would have a particularly strong Fourth Amendment argument on their side, and a good First Amendment argument as well. Recent history suggests, however, that the government's interest in national security or law enforcement often outweighs the citizen's right to privacy.

By their nature, balancing tests almost demand that courts give[Page 884]some play to the judge's hopes and, especially, fears. A mandatory key escrow statute would evoke two conflicting sets of fears, one over control and the other over lawlessness, symbolized by the archetypes of Big Brother and the criminal cabal. In the end, the conflict may be decided by the way the courts characterize cryptography. Just as the cryptographic "key" is a metaphor, so too may the choice among possible metaphors determine how much constitutional protection an encrypted message gets. If the courts treat a ciphertext as if it had been written in a foreign language, it will trigger a First Amendment analysis that will result in giving cryptography more protection than if the courts focus on the place where the message is encrypted. If encryption is considered no more than the outer envelope in a message transmission system--essentially a "car" on the information superhighway--it is likely to receive the lowest level of protection.

Encryption has much to offer the commercial, professional, and personal users of telephones, computers, and computer networks. As these and other uses grow, they will breed conflict, some of which will inevitably be brought to the courts. The legal ecology of cyberspace is currently underpopulated, but not for long. Clipper and Capstone are only the first of many attempts by the government, and no doubt others, to protect the status quo from changes that upset long-established power relationships. The choices made in the next few years will shape the evolution of electronic communication, and society in general, for decades to come. It would be sad if cyberspace became the first place that the government required civilians, in peacetime, to structure their private communications to make hypothetical future eavesdropping by law enforcement easier.