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:
EES involves five distinct government actions. First, the government launched the program by making the classified SKIPJACK algorithm available to a manufacturer of EES-compliant products. Second, the government announced FIPS 185.{348} Third, it is purchasing large numbers of EES-compliant products for its own use. Fourth, it is encouraging others to use EES products. Fifth, it is setting up the two escrow agents who will hold the keys. As a group, these five actions amount to attempting to create a voluntary national key escrow system. Individually and collectively these activities are constitutional.
The NSA controls access to the SKIPJACK algorithm and the details
of the LEAF.{349} To
date it has made the design of the chips available to one
manufacturer, Mykotronx, Inc.{350} FIPS 185 indicates that only organizations
already holding security clearances need apply for access to the
classified specifications for SKIPJACK. A party lacking such a
clearance might have a legitimate grievance if she were unable to
obtain such clearance for the purpose of [Page 794]manufacturing EES-compliant
microcircuitry.{351}
Indeed, if potential competitors to the NSA's chosen manufacturer
were denied access to the information they needed to compete with
Mykotronx, they could plausibly allege an equal protection
violation or a violation of procedural due process. The government
has no obligation, however, to make the algorithm available to
anyone who asks.{352}
The government is free to purchase goods and services to meet its needs.{353} Choosing to purchase EES-compliant devices does not, in itself, create any constitutional issues. Such purchases are constitutional even if they work as an indirect subsidy to producers who are able to lower their unit costs. The government could constitutionally provide direct subsidies if Congress chose to do so.{354} Nor is the denial of market share to non-EES products unconstitutional, even if it has the effect of raising their costs.
The government's cheerleading for EES is also constitutionally permissible. So long as no one is threatened with sanctions for failing to adhere to EES, the government is entitled to make its case to the nation for why we would all benefit if we accepted a limit on our privacy.{355}
[Page 795]The government has
the authority to act as an escrow agent,{356} although there is some question from where the
money to pay for the escrow agents would come. Preliminary
estimates put the cost of the escrow agents' activities at $16
million per year.{357}
These expenses may require a separate appropriation by Congress,
although both NIST and the Justice Department have funds which
arguably might be tapped for this purpose.{358}
Nor is the program as a whole unconstitutional. Even if EES
becomes widespread, everyone in the U.S. remains free to use any
alternative, subject only to restrictions on his or her ability to
export the cryptosystem to foreign correspondents.{359} It remains feasible
and legal to preencrypt a message with an ordinary, non-escrowed
cipher, feed it to an EES-compliant device, and make even EES
communications potentially unintelligible to eavesdroppers armed
with the chip unique key.{360} Indeed, the very ease with which EES [Page 796]can be circumvented raises
the possibility that the government might some day require key
escrow as the price of using strong cryptography.
D. Voluntary EES Is Unlikely to Displace Un-Escrowed
Cryptography
As we have seen, the Administration's stated motives for EES are not
entirely consistent. The government's "hard sell" depicts
non-EES encryption as a threat that needs to be avoided.{361} By contrast, the
"soft sell" treats EES as part of a package deal that the
government offers to those who desire government-certified
encryption.{362}
EES is officially voluntary, yet has been introduced in a manner which
the government hopes will induce, even coerce, the public to choose an
EES system over any alternative.{363} In the Administration's view, it is
unreasonable to object to a plan that protects users from
communications interception by everyone except the government. At
worst, the Administration argues, under EES the user bears no greater
risk of government interception (authorized or not) than do
unencrypted callers.{364} Supporters also point to the need to help law
enforcement in the fight against dangers such as terrorism.{365}
Perhaps the most often repeated objection to EES is that because
people remain free to use alternatives, EES can never achieve its
stated objective of maintaining law enforcement access to private
encrypted communications. Clipper's critics suggest that it can
catch only stupid criminals. The government has had three
responses to this argument. The least subtle response has been
that [Page 797]criminals are
often dumber than one thinks.{366} A more subtle response is that Clipper may at
least postpone the perhaps inevitable adoption of an alternative
cryptosystem that the government cannot easily decrypt.{367} The most subtle
response notes that a secure communication requires compatible
equipment on both ends of the line.{368} If Clipper becomes the de facto standard, the
existence of a few other devices on the margin will have a
negligible effect on the government's ability to monitor electronic
communication when it feels required to do so.
The government's policy centers on its hope that EES will become the
market standard. Yet EES will not likely triumph in the marketplace,
even with the advantage of massive government orders, because many
people find something deeply distasteful about being asked to buy a
product that comes ready-made to be wiretapped, even if the
wiretapping is designed to be conducted only in limited circumstances
by duly authorized bodies. In light of likely technical developments,
a "threat assessment" of the government's potential
surveillance capabilities makes the thought of wiretap-ready
communications even more disturbing. This is especially true
considering the history of government abuse of civil rights and the
possibility, however remote, that government policy might change even
as escrowed chip keys remain fixed. In any case, for e-mail,
alternatives to EES already exist which are cheaper, more flexible,
and appear to offer more complete privacy.{369} Non-EES [Page 798]voice products are also becoming
available.{370}
1. Why EES Worries People
In addition to the fundamental objection that the government
should not expect Americans to facilitate the decryption of their
private communications, opponents of EES have raised numerous
technical and practical objections to the plan. Critics of EES
take what appears to the government to be an absolutist stand,
refusing to trust anyone with the key needed to decrypt their
communications.{371} To
these critics, the government's protestation that EES adds nothing
to current authority because federal law enforcement agencies need
the same court order to obtain a wiretap on an EES-equipped phone
as on an ordinary telephone, makes no impression. The critics
believe either that current rules provide insufficient privacy or
that the government cannot be trusted to follow the rules.
a. Preserving the Status Quo Prevents a Return to the
Status Quo Ante
The status quo that EES seeks to preserve was not always the status
quo. At the time Americans adopted the Bill of Rights, private
communications were far more secure than they are today. Before the
invention of the telephone, the radio, and the long-distance
microphone, one could have a secure conversation by going for a quiet
walk in an open field. Correspondents could encrypt letters in
ciphers that no government could break.{372} ModernWho Needs
[Page 799]communications have
expanded the circle of people to whom we speak, but this fact alone
does not mean that communications should necessarily be more
vulnerable. Only recently, it was difficult for the government to
trace incoming calls, even pursuant to a court order, because the
telephone company used slow mechanical tracing devices. Having
overcome that problem, the FBI now seeks legislation to keep it from
becoming difficult again.{373} Nor does the possibility that more criminals
will avoid detection if the privacy available to individuals were to
be increased necessarily mean that choosing to increase privacy is
unwise. The Bill of Rights already includes many provisions that
prefer to provide protections to all citizens at the cost of providing
benefits to the guilty.{374} What this means is that some value judgments
must be made, and that someone will have to make them.
Where once people only had to worry about eavesdroppers they could see, today an eavesdropper could be anywhere that a telephone signal happens to reach. Modern encryption seems [Page 800]poised to re-create the functional equivalent of the privacy available in the late 1790s and to apply it to devices like telephones and modems, which are increasingly replacing face-to-face contact and letter writing.{375} EES would prevent this return to the status quo ante, at least when the government is the eavesdropper.
Widespread adoption of Clipper and massive wiretapping ability
would make traffic analysis more feasible for a hypothetical
government oblivious to the need to obtain warrants. If Clipper is
widely used, communications encrypted by other means signal that
the user may have something to hide. Indeed, for this reason some
privacy advocates encourage the routine use of strong cryptography
in all communications in order to provide a cloaking effect for all
personal communications. If everyone makes a habit of using strong
cryptography, the presence of an encrypted message will never be
probative of a guilty conscience or a need for secrecy.{376}
b. EES Does Not Preserve the Status Quo
EES is designed to be inflexible, and this inflexibility will impose
costs on some users. Each chip's unique key is permanently branded
onto it. If for some reason that key should be compromised, the user
has no choice but to throw away the chip and buy a new one. This
inflexibility is designed to make it impossible for users to select
keys that are not held by the government.{377} Under Title III,
the government must notify persons who were the subject of an
authorized wiretap.{378} This duty is unaffected by EES, but [Page 801]the consequences change.
Previously there was little a citizen needed to do after receiving
notice that her phone had been tapped, but now she must consider
whether the disclosure to law enforcement officials of the chip unique
key in her telephone means that she should replace it, at a cost,{379} or whether she
should trust government assurances that all records of the key kept
outside the escrow agents have been destroyed.{380}
Two telephones communicating via Clipper Chips use the same session key; thus, when Alice and Bob are talking, a public servant with a warrant for Alice's telephone does not need to know Bob's chip key to decrypt the conversation. Knowing Alice's chip key will suffice because Alice's LEAF will provide all the information needed. Except for the fact that he is overheard talking to Alice, Bob's security is unaffected by a wiretap of Alice's line.
But if Alice and Bob are using e-mail to communicate and Capstone Chips{381} to do their encryption, both Bob and the public servant are in a different position. Capstone is designed to allow Alice and Bob to use public key encryption for their session keys.{382} Bob's Fortezza card knows Alice's public key, but not her private key or her chip key, so the only LEAF it is able to generate is one that relies on Bob's own chip key. This creates a lot of work for a public servant tapping Alice's line. Every time she gets an e-mail from a new correspondent, the public servant must decrypt its LEAF with the family key and then go to the escrow agents and request the chip unique key for the new person. If Alice communicates with many people who use Fortezza cards, the public servant may wind up holding a large, and rather valuable, collection of chip keys.
Because the wiretap order mentions only Alice, the court that
issued the order has discretion to decide whether each of the
people whose session keys were disclosed should be notified of that
[Page 802]fact.{383} Although nothing in
Title III or the Attorney General's rules requires it, Bob deserves
to be told.
Bob's Fortezza card will provide his digital signature as well as
encryption for his e-mail. Disclosure of the digital signature key to
anyone who might even be tempted to sell or make use of it would
represent an enormous risk to Bob. Anyone holding Bob's key to his
digital signature could masquerade as him and authenticate any
transaction or correspondence (for example, in a raid on Bob's
electronic bank account) with a digital signature that Bob would be
powerless to disavow. Fortunately, current plans for Fortezza call
for separate keys for message encryption and for digital signatures.{384} Furthermore,
although Bob is powerless to change the chip unique key used to encode
his e-mail's LEAF, Fortezza will allow him to change the key to his
digital signature. Thus, Bob's ability to uniquely identify himself
remains secure.
c. The Status Quo May Not Be Stable
The biggest divide between the two sides to the EES debate
concerns what they consider relevant. The Clinton Administration,
as one would expect, operates on the assumption that government
officials can be trusted to act legally.{385} The government therefore measures the social
consequences of its proposals by the effect on the government's
lawful powers and the citizen's lawful rights. Critics of EES,
however, tend to discount this approach. Instead, they undertake
a threat analysis of the EES proposal.{386} It may seem a little silly to conduct a
threat analysis of a cryptographic proposal by a government that
has the raw physical power to do far worse things than spying on
its citizens, but in fact threat assessment enjoys a grand
tradition. The Framers of the Constitution did [Page 803]not assume that "men
were Angels."{387}
They conducted a kind of threat analysis of government and decided
that it could only be trusted if centralized power were divided in
a manner that set interest against interest so as to protect the
governed.{388} The
impulse to rely as much as possible on structures that force proper
behavior by government officials, and as little as possible on
simple trust, is as old as the nation.{389}
Some of these threats to the status quo are political. For example, one glaring risk in the current EES proposal is that the escrow procedures exist entirely within the purview of the Attorney General, and could be changed at any time without any warning.{390}
[Page 804]Some threats
consist of individual or official malefaction. In this age of spy
scandals, it is always possible that the escrow agents, through
negligence or corruption, may allow someone to acquire the full
list of key segments.{391} The method by which keys are generated for
the EES chips may lend itself to subversion of the escrow scheme
from the moment the keys are generated. Although hedged with
elaborate safeguards, all keys are generated by a single computer
in a secure facility closed to public inspection. Because users
are not in a position to monitor the key-generation procedure, they
must trust that the published safeguards are being observed. Even
if the risk of surreptitious subversion of the generation process
were small, the risk to communications security would be greater
than if the keys had never been escrowed.
Some threats to the status quo are mathematical. Critics argue that a classified algorithm such as SKIPJACK--one that has not been exposed to merciless attack by academic cryptologists--is less likely to be secure than one subject to full peer review and thus might contain an intentional, or even unintentional, "back door" that would make it vulnerable to sophisticated mathematical attack.{392} The government's response is that SKIPJACK's security is certified by the NSA{393} and by independent outside experts.{394} The government classified SKIPJACK not out of fear that publicity might expose the algorithm to attack, but to prevent users from enjoying the fruits of its research and development while at the same time avoiding participation in its key escrow system. The Administration argues that SKIPJACK is so strong that, were people able to use it without escrowing their keys, they would undermine the goal of easy government access to encrypted messages that EES is designed to achieve.{395} Some critics remain unsatisfied by this explanation. They argue that because EES is voluntary, the government should not attempt to require compliance with the escrow procedure as a condition of using SKIPJACK.{396} The Administration's response is, in effect, that if users wish to use a government-certified algorithm, they should be prepared to take the bitter with the sweet.
Some threats, perhaps the most realistic, are technological. Changes in technology are likely to make electronic eavesdropping easier, more effective, and cheaper for the government.{397} All other things being equal, a rational government would react to these changes by increasing the use of electronic eavesdropping. As government eavesdropping becomes more affordable, the reasonable citizen's desire for countermeasures ought to become greater as well.
[Page 805]The technological
threat appears more ominous if one tries to forecast what the
government may be able to do a decade from now. Currently, all the
wiretapping technology in the world is useless if there is no one to
listen to the conversations. The physical and economic limit of what
is currently achievable is demonstrated by the East German Ministry
for State Security, the Staatsicherheit or Stasi, which at its
peak was probably the most sophisticated and farreaching internal
surveillance organization ever created. Out of a population of 17
million, the Stasi had 34,000 officers, including 2100 agents reading
mail and 6000 operatives listening to private telephone conversations,
plus 150,000 active informers and up to 2 million part-time
informers.{398}
Together they produced dossiers on more than one out of three East
Germans, amounting to one billion pages of files.{399} There are
fifty-nine times more telephones in the United States than there were
in East Germany and about fifteen times as many people.{400} The people (and
machines) in the United States make about 3.5 trillion calls per
year.{401} Even if
every telephone service provider in the United States were to record
every conversation in the country, the government could not make use
of the tapes because it lacks the human resources necessary to listen
to them. Even if political constraints could not prevent the growth
of an American Stasi, the financial constraints are currently
insurmountable.{402}
The cost may soon shrink dramatically. EES, the Digital
Telephony initiative,{403} and advances in computer power, combined with
the increasing links among federal databases{404} and [Page 806]advances in voice recognition
protocols, suggest that soon the physical constraints on
widespread, government-sponsored eavesdropping may disappear.
Voice recognition already allows computers to pick out a particular
speaker's voice from the babble of communications;{405} combined with the
power to search for particular words in all messages, this advance
in technology will provide a powerful surveillance tool to any
government willing to use it. Computers can monitor communications
twenty-four hours per day, and they do not collect overtime. In
the absence of physical and economic constraints, the only
constrictions on omnipresent automated telephone monitoring will be
legal and political.{406}
2. Spoofing EES: The LEAF-Blower
EES suffered a glancing blow when a researcher at AT&T
discovered that it could be "spoofed," albeit with some
effort.{407} The
protocol that produces the spoofs quickly became popularly known as
the "LEAF-blower."{408} The process is too slow to be of[Page 807]much practical value in
Clipper-telephone communications, but might be applied by patient
e-mail users of Capstone.{409}
Recall that an EES-compliant device will only decrypt a message that comes headed by what appears to be a valid LEAF. A "spoof" replaces the real LEAF with a simulacrum, which appears valid to the decrypting chip, and even an eavesdropper armed with the family key, but is in fact meaningless. Because the actual session key is negotiated before the LEAF is generated, the absence of the true session key in the LEAF does not affect communications so long as the LEAF passes the validity check. Because the decrypting chip checks the LEAF against a 16-bit checksum,{410} which uses the actual session key as one of its inputs, a spoof requires more than just copying a LEAF off a previous transmission. A spoof is computationally complex because the spoofer must use trial and error to generate a LEAF with a phony session key whose checksum equals that of the real session key. Each time the LEAF-blower is used, an average of 32,768 LEAFs must be tried before one works. Tests at AT&T on a prototype Capstone-based PCMCIA card showed that, on average, more than forty minutes would be needed to produce a valid-looking spoof.{411}
A LEAF-blower allows a "rogue" EES device to communicate with all other EES devices, without the recipient even knowing that the sender has spoofed the chip. Because it can take up to forty-two minutes to counterfeit the LEAF, however, the technique is likely to remain primarily of interest only to very patient people. Interestingly, NIST claims it was always aware that a LEAF-blower device could be constructed. It found the risk acceptable, however, because the technique was too slow to be of practical value.{412} Furthermore, because the chip serial number contains a field identifying the manufacturer as well as the chip, anyone who decrypts a rogue LEAF with the family key will be able to recognize a bogus chip serial number without having to consult the escrow agents.{413}
[Page 808]Thus, the way to
feign compliance with EES remains preencrypting the message with
some other system before using the EES device. Preencryption is
undetectable with the family key alone, but is discernable only
after the escrow agents have released the chip unique key.
Preencryption is relatively easy for e-mail, but it is difficult to
achieve for real-time voice communication. As a result, an eaves-
dropper armed with the family key should be in a good position to
monitor compliance with EES even if she cannot decrypt the
conversation.{414}
E. What Happens If EES Fails?
The large number of government orders and the attraction of
SKIPJACK for those who need the security of a government-certified
cryptosystem means that EES is unlikely to disappear, especially in
its incarnation as the Fortezza PCMCIA card.{415} It has, however,
engendered enough opposition to put its future in doubt.{416} The existence of
other well-regarded ciphers such as triple-DES{417} and IDEA,{418} combined with public
distaste for wiretap-ready telephones, the many unanswered
questions about the proposal, the cost premium for a hardware (as
opposed to a software) cryptosystem, the inflexibility of EES, and
the lack of interoperability with foreign cryptosystems will likely
combine to render EES if not stillborn, then at least stunted.
It seems reasonable, therefore, to speculate as to how the
government will react if EES fails to become the standard.
Assuming the government does not come up with a wholly new system
to replace EES, two options exist:{419} (1) do nothing; or (2) [Page 809]forbid the use of unescrowed
cryptography. The former option is implicit in the "soft
sell" policy that describes EES as the price the private
sector must pay for using SKIPJACK. If the private sector refuses
EES, it forgoes SKIPJACK. That is its privilege, and no further
government action would be needed.
The latter of the two approaches is implicit in the "hard sell" for EES. If widespread unregistered encryption can be used by "drug dealers, terrorists, and other criminals," to quote the White House,{420} then the country cannot afford to do nothing. But with unregistered cryptography already widely available, the only option may be a "Digital Volstead Act."{421}
The Clinton Administration considered banning unescrowed
encryption,{422}
but then concluded that it would "not propose new legislation to
limit use of encryption technology."{423} A future
administration might, however, reverse this decision, particularly if
an investigation into a high-profile crime, such as the terrorist
bombing of a major building or the management of a child pornography
ring, was found to have been seriously hampered by the use of advanced
cryptography. The current Administration has carefully left that
option open for its successors, noting that by forgoing a ban on
unescrowed encryption it is not "saying that [Page 810]`every American, as a matter of
right, is entitled to an unbreakable commercial encryption
product.'"{424}
The government is clearly willing to require that communications be made wiretap-ready, at least when it knows that its dictates can be enforced.{425} It is also "apparent that the law enforcement community is still looking for a way to meet its surveillance needs in the age of digital communications."{426} If EES fails, the law enforcement and intelligence communities, at least, will seek to preserve their capabilities. Legislation requiring that all strong cryptographic programs use key escrow may be the only remaining solution. As FBI Director Freeh commented, "If five years from now . . . what we are hearing is all encrypted" material that the FBI is unable to decipher, then the policy of relying on voluntary compliance with EES will have to change.{427} "The objective is for us to get those conversations whether they are . . . ones and zeros [or] wherever they are, whatever they are, I need them."{428} As a result, Part III examines the legal problems that would flow from hypothetical legislation making key escrow mandatory.