For any pair of parties, there are four possible conditions of the
authentication clocks that could require correction:
1. The management station's notion of the value of the
authentication clock for agentParty exceeds the agent's
notion.
2. The management station's notion of the value of the
authentication clock for mgrParty exceeds the agent's
notion.
3. The agent's notion of the value of the authentication
clock for agentParty exceeds the management station's
notion.
4. The agent's notion of the value of the authentication
clock for mgrParty exceeds the management station's
notion.
The selective clock acceleration mechanism intrinsic to the protocol
corrects conditions 2 and 3 as part of the normal processing of an
authentic message. Therefore, the clock adjustment procedure below
does not provide for any adjustments in those cases. Rather, the
following sequence of steps specifies how the clocks may be
synchronized when condition 1, condition 4, or both of those
conditions are manifest.
1. The responsible management station saves its existing
notions of the authentication clocks for the two parties
agentParty and mgrParty.
2. The responsible management station retrieves the
authentication clock values for both agentParty and
mgrParty from the agent. This retrieval must be an
unauthenticated request, since the management station
does not know if the clocks are synchronized. If the
request fails, the clocks cannot be synchronized, and the
clock adjustment procedure is aborted without further
processing.
3. If the management station's notion of the authentication
clock for agentParty exceeds the notion just retrieved
from the agent by more than the amount of the
communications delay between the two protocol entities,
then condition 1 is manifest. The recommended estimate
of communication delay in this context is one half of the
lifetime value recorded for agentParty.
4. If the notion of the authentication clock for mgrParty
just retrieved from the agent exceeds the management
station's notion, then condition 4 is manifest, and the
responsible management station advances its notion of
the authentication clock for mgrParty to match the
agent's notion.
5. If condition 1 is manifest, then the responsible
management station sends an authenticated
management operation to the agent that advances the
agent's notion of the authentication clock for
agentParty to be equal to the management station's
notion. If this management operation fails, then the
management station restores its previously saved notions
of the clock values, and the clock adjustment procedure
is aborted without further processing.
6. The responsible management station retrieves the
authentication clock values for both agentParty and
mgrParty from the agent. This retrieval must be an
authenticated request, in order that the management
station may verify that the clock values are properly
synchronized. If this authenticated query fails, then the
management station restores its previously saved notions
of the clock values, and the clock adjustment procedure
is aborted without further processing. Otherwise, clock
synchronization has been successfully realized.
It is important to note step 4 above must be completed before
attempting step 5. Otherwise, the agent may evaluate the request in
step 5 as unauthentic. Similarly, step 5 above must be completed
before attempting step 6. Otherwise, the management station may
evaluate the query response in step 6 as unauthentic.
Administrative advancement of a clock as described above does not
introduce any new vulnerabilities, since the value of the clock is
intended to increase with the passage of time. A potential
operational problem is the rejection of management operations that
are authenticated using a previous value of the relevant party clock.
This possibility may be avoided if a management station suppresses
generation of management traffic between relevant parties while this
clock adjustment procedure is in progress.
6.4 Secret Distribution
This section describes one strategy by which a SNMP protocol entity
that supports both the Digest Authentication Protocol and the
Symmetric Privacy Protocol can change the secrets for a particular
SNMP party.
The frequency with which the secrets of a SNMP party should be
changed is a local administrative issue. However, the more frequently
a secret is used, the more frequently it should be changed. At a
minimum, the secrets must be changed whenever the associated
authentication clock approaches its maximal value (see Section 7).
Note that, owing to both administrative and automatic advances of the
authentication clock described in this memo, the authentication clock
for a SNMP party may well approach its maximal value sooner than
might otherwise be expected.
The following sequence of steps specifies how a responsible
management station alters a secret value (i.e., the private
authentication key or the private privacy key) for a particular SNMP
party.
1. The responsible management station generates a new
secret value.
2. The responsible management station encapsulates a
SNMP Set request in a SNMP private management
communication with at least the following properties.
o Its source supports the Digest Authentication
Protocol and the Symmetric Privacy Protocol.
o Its destination supports the Symmetric Privacy
Protocol and the Digest Authentication Protocol.
3. The SNMP private management communication is
transmitted to its destination.
4. Upon receiving the request, the recipient processes the
message according to [1] and [2].
5. The recipient encapsulates a SNMP Set response in a
SNMP private management communication with at least
the following properties.
o Its source supports the Digest Authentication
Protocol and the Symmetric Privacy Protocol.
o Its destination supports the Symmetric Privacy
Protocol and the Digest Authentication Protocol.
6. The SNMP private management communication is
transmitted to its destination.
7. Upon receiving the response, the responsible
management station updates its local database with the
new value.
If the responsible management station does not receive a response to
its request, there are two possible causes.
o The request may not have been delivered to the
destination.
o The response may not have been delivered to the
originator of the request.
In order to distinguish the two possible error conditions, a
responsible management station could check the destination to see if
the change has occurred. Unfortunately, since the secret values are
unreadable, this is not directly possible.
The recommended strategy for verifying key changes is to set the
public value corresponding to the secret being changed to a
recognizable, novel value: that is, alter the public authentication
key value for the relevant party when changing its private
authentication key, or alter its public privacy key value when
changing its private privacy key. In this way, the responsible
management station may retrieve the public value when a response is
not received, and verify whether or not the change has taken place.
(This strategy is available since the public values are not used by
the protocols defined in this memo. If this strategy is employed,
then the public values are significant in this context. Of course,
protocols using the public values may make use of this strategy
directly.)
One other scenario worthy of mention is using a SNMP party to change
its own secrets. In this case, the destination will change its local
database prior to generating a response. Thus, the response will be
constructed according to the new value. However, the responsible
management station will not update its local database until after the
response is received. This suggests the responsible management
station may receive a response which will be evaluated as
unauthentic, unless the correct secret is used. The responsible
management station may either account for this scenario as a special
case, or use an alteration of the relevant public values (as
described above) to verify the key change.
Note, during the period of time after the request has been sent and
before the response is received, the management station must keep
track of both the old and new secret values. Since the delay may be
the result of a network failure, the management station must be
prepared to retain both values for an extended period of time,
including across reboots.
6.5 Crash Recovery
This section describes the requirements for SNMP protocol entities in
connection with recovery from system crashes or other service
interruptions.
For each SNMP party in the local database for a particular SNMP
protocol entity, its identity, authentication clock, private
authentication key, and private privacy key must enjoy non-volatile,
incorruptible representations. If possible, lifetime should also
enjoy a non-volatile, incorruptible representation. If said protocol
entity supports other security protocols or algorithms in addition to
the two defined in this memo, then the authentication protocol and
the privacy protocol for each party also require non-volatile,
incorruptible representation.
The authentication clock of a SNMP party is a critical component of
the overall security of the protocols. The inclusion of a reliable
representation of a clock in a SNMP protocol entity enhances overall
security. A reliable clock representation continues to increase
according to the passage of time, even when the local SNMP protocol
entity -- due to power loss or other system failure -- may not be
operating. An example of a reliable clock representation is that
provided by battery-powered clock-calendar devices incorporated into
some contemporary systems. It is assumed that management stations
always support reliable clock representations, where clock adjustment
by a human operator during crash recovery may contribute to that
reliability.
If a managed agent crashes and does not reboot in time for its
responsible management station to prevent its authentication clock
from reaching its maximal value, upon reboot the clock must be halted
at its maximal value. The procedures specified in Section 6.3 would
then apply.
If a managed network element supports a reliable clock
representation, recovering from a crash requires few special actions.
Upon recovery, those attributes of each SNMP party that do not enjoy
non-volatile or reliable representation are initialized as follows.
o If the private authentication key is not the OCTET
STRING of zero length, the authentication protocol is
set to identify use of the Digest Authentication Protocol
in conjunction with the algorithm specified in
Section 2.4.1.
o The last-timestamp is initialized to the value of the
authentication clock.
o The nonce is initialized to zero.
o If the lifetime is not retained, it should be initialized to
zero.
o If the private privacy key is not the OCTET STRING
of zero length, the privacy protocol is set to identify use
of the Symmetric Privacy Protocol in conjunction with
the algorithm specified in Section 2.4.2.
Upon detecting that a managed agent has rebooted, a responsible
management station must reset all other party attributes, including
the lifetime if it was not retained. In order to reset the lifetime,
the responsible management station should set the authentication
timestamp in the message to the sum of the authentication clock and
desired lifetime. This is an artificial advancement of the
authentication timestamp in order to guarantee the message will be
authentic when received by the recipient.
If, alternatively, a managed network element does not support a
reliable clock representation, then those attributes of each SNMP
party that do not enjoy non-volatile representation are initialized
as follows.
o If the private authentication key is not the OCTET
STRING of zero length, the authentication protocol is
set to identify use of the Digest Authentication Protocol
in conjunction with the algorithm specified in
Section 2.4.1.
o The authentication clock is initialized to the maximal
time value.
o The last-timestamp is initialized to the maximal time
value.
o The nonce is initialized to zero.
o If the lifetime is not retained, it should be initialized to
zero.
o If the private privacy key is not the OCTET STRING
of zero length, the privacy protocol is set to identify use
of the Symmetric Privacy Protocol in conjunction with
the algorithm specified in Section 2.4.2.
The only authenticated request a management station should generate
for a party in this initial state is one that alters the value of at
least its authentication clock, private authentication key, and
lifetime (if that was not retained). In order to reset these values,
the responsible management station must set the authentication
timestamp in the message to the maximal time value. The nonce value
may be used to distinguish multiple messages.
7. Security Considerations
This section highlights security considerations relevant to the
protocols and procedures defined in this memo. Practices that
contribute to secure, effective operation of the mechanisms defined
here are described first. Constraints on implementation behavior that
are necessary to the security of the system are presented next.
Finally, an informal account of the contribution of each mechanism of
the protocols to the required goals is presented.
7.1 Recommended Practices
This section describes practices that contribute to the secure,
effective operation of the mechanisms defined in this memo.
o A management station should discard SNMP responses
for which neither the request-id component nor the
represented management information corresponds to any
currently outstanding request.
Although it would be typical for a management station
to do this as a matter of course, in the context of these
security protocols it is significant owing to the possibility
of message duplication (malicious or otherwise).
o A management station should not interpret an agent's
lack of response to an authenticated SNMP management
communication as a conclusive indication of agent or
network failure.
It is possible for authentication failure traps to be lost or
suppressed as a result of authentication clock skew or
inconsistent notions of shared secrets. In order either to
facilitate administration of such SNMP parties or to
provide for continued management in times of network
stress, a management station implementation may
provide for arbitrary, artificial advancement of the
timestamp or selection of shared secrets on locally
generated messages.
o The lifetime value for a SNMP party should be chosen
(by the local administration) to be as small as possible,
given the accuracy of clock devices available, relevant
round-trip communications delays, and the frequency
with which a responsible management station will be
able to verify all clock values.
A large lifetime increases the vulnerability to malicious
delays of SNMP messages. The implementation of a
management station may, when explicitly authorized,
provide for dynamic adjustment of the lifetime in order
to accommodate changing network conditions.
o When sending state altering messages to a managed
agent, a management station should delay sending
successive messages to the managed agent until a
positive acknowledgement is received for the previous
message or until the previous message expires.
When using the noAuth protocol, no message ordering
is imposed by the SNMP. Messages may be received in
any order relative to their time of generation and each
will be processed in the ordered received. In contrast,
the security protocols guarantee that received messages
are ordered insofar as each received message must have
been sent subsequent to the sending of a previously
received message.
When an authenticated message is sent to a managed
agent, it will be valid for a period of time that does not
exceed lifetime under normal circumstances. During the
period of time this message is valid, if the management
station sends another authenticated message to the
managed agent that is received and processed prior to
the first message, the first message will be considered
unauthentic when it is received by the managed agent.
Indeed, a management station must cope with the loss
and re-ordering of messages resulting from anomalies in
the network as a matter of course. A management
station implementation may choose to prevent the loss
of messages resulting from re-ordering when using the
security protocols defined in this memo by delaying
sending successive messages.
o The frequency with which the secrets of a SNMP party
should be changed is indirectly related to the frequency
of their use.
Protecting the secrets from disclosure is critical to the
overall security of the protocols. Frequent use of a secret
provides a continued source of data that may be useful
to a cryptanalyst in exploiting known or perceived
weaknesses in an algorithm. Frequent changes to the
secret avoid this vulnerability.
Changing a secret after each use is is generally regarded
as the most secure practice, but a significant amount of
overhead may be associated with that approach.
Note, too, in a local environment the threat of disclosure
may be insignificant, and as such the changing of secrets
may be less frequent. However, when public data
networks are the communication paths, more caution is
prudent.
o In order to foster the greatest degree of security, a
management station implementation must support
constrained, pairwise sharing of secrets among SNMP
entities as its default mode of operation.
Owing to the use of symmetric cryptography in the
protocols defined here, the secrets associated with a
particular SNMP party must be known to all other
SNMP parties with which that party may wish to
communicate. As the number of locations at which
secrets are known and used increases, the likelihood of
their disclosure also increases, as does the potential
impact of that disclosure. Moreover, if the set of SNMP
protocol entities with knowledge of a particular secret
numbers more than two, data origin cannot be reliably
authenticated because it is impossible to determine with
any assurance which entity of that set may be the
originator of a particular SNMP message. Thus, the
greatest degree of security is afforded by configurations
in which the secrets for each SNMP party are known to
at most two protocol entities.
7.2 Conformance
A SNMP protocol entity implementation that claims conformance to this
memo must satisfy the following requirements:
1. It must implement the noAuth and noPriv protocols
whose object identifiers are defined in [4].
noAuth This protocol signifies that messages generated
by a party using it are not protected as to origin or
integrity. It is required to ensure that a party's
authentication clock is always accessible.
noPriv This protocol signifies that messages received
by a party using it are not protected from
disclosure. It is required to ensure that a party's
authentication clock is always accessible.
2. It must implement the Digest Authentication Protocol in
conjunction with the algorithm defined in Section 2.4.1.
3. It must include in its local database at least one SNMP
party with the following parameters set as follows:
o partyAuthProtocol is set to noAuth and
o partyPrivProtocol is set to noPriv.
This party must have a MIB view [2] specified that
includes at least the authentication clock of all other
parties. Alternatively, the authentication clocks of the
other parties may be partitioned among several similarly
configured parties according to a local implementation
convention.
4. For each SNMP party about which it maintains
information in a local database, an implementation must
satisfy the following requirements:
(a) It must not allow a party's parameters to be set to
a value inconsistent with its expected syntax. In
particular, Section 2.4 specifies constraints for the
chosen mechanisms.
(b) It must, to the maximal extent possible, prohibit
read-access to the private authentication key and
private encryption key under all circumstances
except as required to generate and/or validate
SNMP messages with respect to that party. This
prohibition includes prevention of read-access by
the entity's human operators.
(c) It must allow the party's authentication clock to be
publicly accessible. The correct operation of the
Digest Authentication Protocol requires that it be
possible to determine this value at all times in
order to guarantee that skewed authentication
clocks can be resynchronized.
(d) It must prohibit alterations to its record of the
authentication clock for that party independently of
alterations to its record of the private
authentication key (unless the clock alteration is an
advancement).
(e) It must never allow its record of the authentication
clock for that party to be incremented beyond the
maximal time value and so "roll-over" to zero.
(f) It must never increase its record of the lifetime for
that party except as may be explicitly authorized
(via imperative command or securely represented
configuration information) by the responsible
network administrator.
(g) In the event that the non-volatile, incorruptible
representations of a party's parameters (in
particular, either the private authentication key or
private encryption key) are lost or destroyed, it
must alter its record of these quantities to random
values so subsequent interaction with that party
requires manual redistribution of new secrets and
other parameters.
5. If it selects new value(s) for a party's secret(s), it must
avoid bad or obvious choices for said secret(s). Choices
to be avoided are boundary values (such as all-zeros)
and predictable values (such as the same value as
previously or selecting from a predetermined set).
7.3 Protocol Correctness
The correctness of these SNMP security protocols with respect to the
stated goals depends on the following assumptions:
1. The chosen message digest algorithm satisfies its design
criteria. In particular, it must be computationally
infeasible to discover two messages that share the same
digest value.
2. It is computationally infeasible to determine the secret
used in calculating a digest on the concatenation of the
secret and a message when both the digest and the
message are known.
3. The chosen symmetric encryption algorithm satisfies its
design criteria. In particular, it must be computationally
infeasible to determine the cleartext message from the
ciphertext message without knowledge of the key used in
the transformation.
4. Local notions of a party's authentication clock while it is
associated with a specific private key value are
monotonically non-decreasing (i.e., they never run
backwards) in the absence of administrative
manipulations.
5. The secrets for a particular SNMP party are known only
to authorized SNMP protocol entities.
6. Local notions of the authentication clock for a particular
SNMP party are never altered such that the
authentication clock's new value is less than the current
value without also altering the private authentication
key.
For each mechanism of the protocol, an informal account of its
contribution to the required goals is presented below. Pseudocode
fragments are provided where appropriate to exemplify possible
implementations; they are intended to be self-explanatory.
7.3.1 Clock Monotonicity Mechanism
By pairing each sequence of a clock's values with a unique key, the
protocols partially realize goals 3 and 4, and the conjunction of
this property with assumption 6 above is sufficient for the claim
that, with respect to a specific private key value, all local notions
of a party's authentication clock are, in general, non-decreasing
with time.
7.3.2 Data Integrity Mechanism
The protocols require computation of a message digest computed over
the SNMP message prepended by the secret for the relevant party. By
virtue of this mechanism and assumptions 1 and 2, the protocols
realize goal 1.
Normally, the inclusion of the message digest value with the digested
message would not be sufficient to guarantee data integrity, since
the digest value can be modified in addition to the message while it
is enroute. However, since not all of the digested message is
included in the transmission to the destination, it is not possible
to substitute both a message and a digest value while enroute to a
destination.
Strictly speaking, the specified strategy for data integrity does not
detect a SNMP message modification which appends extraneous material
to the end of such messages. However, owing to the representation of
SNMP messages as ASN.1 values, such modifications cannot --
consistent with goal 1 -- result in unauthorized management
operations.
The data integrity mechanism specified in this memo protects only
against unauthorized modification of individual SNMP messages. A more
general data integrity service that affords protection against the
threat of message stream modification is not realized by this
mechanism, although limited protection against reordering, delay, and
duplication of messages within a message stream are provided by other
mechanisms of the protocol.
7.3.3 Data Origin Authentication Mechanism
The data integrity mechanism requires the use of a secret value known
only to communicating parties. By virtue of this mechanism and
assumptions 1 and 2, the protocols explicitly prevent unauthorized
modification of messages. Data origin authentication is implicit if
the message digest value can be verified. That is, the protocols
realize goal 2.
7.3.4 Restricted Administration Mechanism
This memo requires that implementations preclude administrative
alterations of the authentication clock for a particular party
independently from its private authentication key (unless that clock
alteration is an advancement). An example of an efficient
implementation of this restriction is provided in a pseudocode
fragment below. This pseudocode fragment meets the requirements of
assumption 6.
Pseudocode Fragment. Observe that the requirement is not for
simultaneous alteration but to preclude independent alteration. This
latter requirement is fairly easily realized in a way that is
consistent with the defined semantics of the SNMP Set operation.
Void partySetKey (party, newKeyValue)
{
if (party->clockAltered) {
party->clockAltered = FALSE;
party->keyAltered = FALSE;
party->keyInUse = newKeyValue;
party->clockInUse = party->clockCache;
}
else {
party->keyAltered = TRUE;
party->keyCache = newKeyValue;
}
}
Void partySetClock (party, newClockValue)
{
if (party->keyAltered) {
party->keyAltered = FALSE;
party->clockAltered = FALSE;
party->clockInUse = newClockValue;
party->keyInUse = party->keyCache;
}
else {
party->clockAltered = TRUE;
party->clockCache = newClockValue;
}
}
7.3.5 Ordered Delivery Mechanism
The definition of the Digest Authentication Protocol requires that,
if the timestamp value on a received message does not exceed the
timestamp of the most recent validated message locally delivered from
the originating party, then that message is not delivered. Otherwise,
the record of the timestamp for the most recent locally delivered
validated message is updated.
if (msgIsValidated) {
if (timestampOfReceivedMsg >
party->timestampOfLastDeliveredMsg) {
party->timestampOfLastDeliveredMsg =
timestampOfReceivedMsg;
}
else {
msgIsValidated = FALSE;
}
}
Although not explicitly represented in the pseudocode above, in the
Digest Authentication Protocol, the ordered delivery mechanism must
ensure that, when the authentication timestamp of the received
message is equal to the last-timestamp, received messages continue to
be delivered as long as their nonce values are monotonically
increasing. By virtue of this mechanism, the protocols realize goal
4.
7.3.6 Message Timeliness Mechanism
The definition of the SNMP security protocols requires that, if the
authentication timestamp value on a received message -- augmented by
an administratively chosen lifetime value -- is less than the local
notion of the clock for the originating SNMP party, the message is
not delivered.
if (timestampOfReceivedMsg +
party->administrativeLifetime <=
party->localNotionOfClock) {
msgIsValidated = FALSE;
}
By virtue of this mechanism, the protocols realize goal 3. In cases
in which the local notions of a particular SNMP party clock are
moderately well-synchronized, the timeliness mechanism effectively
limits the age of validly delivered messages. Thus, if an attacker
diverts all validated messages for replay much later, the delay
introduced by this attack is limited to a period that is proportional
to the skew among local notions of the party clock.
7.3.7 Selective Clock Acceleration Mechanism
The definition of the SNMP security protocols requires that, if the
timestamp value on a received, validated message exceeds the local
notion of the clock for the originating party, then that notion is
adjusted forward to correspond to said timestamp value. This
mechanism is neither strictly necessary nor sufficient to the
security of the protocol; rather, it fosters the clock
synchronization on which valid message delivery depends -- thereby
enhancing the effectiveness of the protocol in a management context.
if (msgIsValidated) {
if (timestampOfReceivedMsg >
party->localNotionOfClock) {
party->localNotionOfClock =
timestampOfReceivedMsg;
}
}
The effect of this mechanism is to synchronize local notions of the
party clock more closely in the case where a sender's notion is more
advanced than a receiver's. In the opposite case, this mechanism has
no effect on local notions of the party clock and either the received
message is validly delivered or not according to other mechanisms of
the protocol.
Operation of this mechanism does not, in general, improve the
probability of validated delivery for messages generated by party
participants whose local notion of the party clock is relatively less
advanced. In this case, queries from a management station may not be
validly delivered and the management station needs to react
appropriately (e.g., by administratively resynchronizing local
notions of the clock in conjunction with a key change). In contrast,
the delivery of SNMP trap messages generated by an agent that suffers
from a less advanced notion of a party clock is more problematic, for
an agent may lack the capacity to recognize and react to security
failures that prevent delivery of its messages. Thus, the inherently
unreliable character of trap messages is likely to be compounded by
attempts to provide for their validated delivery.
7.3.8 Confidentiality Mechanism
The protocols require the use of a symmetric encryption algorithm
when the data confidentiality service is required. By virtue of this
mechanism and assumption 3, the protocols realize goal 5.
8. Acknowledgements
The authors would like to thank the members of the SNMP Security
Working Group of the IETF for their patience and comments. Special
thanks go to Jeff Case who provided the first implementation of the
protocols. Dave Balenson, John Linn, Dan Nessett, and all the members
of the Privacy and Security Research Group provided many valuable and
detailed comments.
9. References
[1] Case, J., M. Fedor, M. Schoffstall, and J. Davin, The Simple
Network Management Protocol", RFC1157, University of Tennessee
at Knoxville, Performance Systems International, Performance
Systems International, and the MIT Laboratory for Computer
Science, May 1990. (Obsoletes RFC1098.)
[2] Davin, J., Galvin, J., and K. McCloghrie, "SNMP Administrative
Model", RFC1351, MIT Laboratory for Computer Science, Trusted
Information Systems, Inc., Hughes LAN Systems, Inc., July 1992.
[3] Rivest, R., "The MD5 Message-Digest Algorithm", RFC1321, MIT
Laboratory for Computer Science, April 1992.
[4] McCloghrie, K., Davin, J., and J. Galvin, "Definitions of Managed
Objects for Administration of SNMP Parties", RFC1353, Hughes LAN
Systems, Inc., MIT Laboratory for Computer Science, Trusted
Information Systems, Inc., July 1992.
[5] FIPS Publication 46-1, "Data Encryption Standard", National
Institute of Standards and Technology, Federal Information
Processing Standard (FIPS); Supersedes FIPS Publication 46,
January 15, 1977; Reaffirmed January 22, 1988.
[6] ANSI X3.92-1981, "Data Encryption Algorithm", American National
Standards Institute, December 30, 1980.
[7] FIPS Publication 81, "DES Modes of Operation", National Institute
of Standards and Technology, December 2, 1980, Federal
Information Processing Standard (FIPS).
[8] ANSI X3.106-1983, "Data Encryption Algorithm - Modes of
Operation", American National Standards Institute, May 16, 1983.
[9] FIPS Publication 74, "Guidelines for Implementing and Using the
NBS Data Encryption Standard", National Institute of Standards
and Technology, April 1, 1981. Federal Information Processing
Standard (FIPS).
[10] Special Publication 500-20, "Validating the Correctness of
Hardware Implementations of the NBS Data Encryption Standard",
National Institute of Standards and Technology.
[11] Special Publication 500-61, "Maintenance Testing for the Data
Encryption Standard", National Institute of Standards and
Technology, August 1980.
[12] Information Processing -- Open Systems Interconnection --
Specification of Basic Encoding Rules for Abstract Syntax
Notation One (ASN.1), International Organization for
Standardization/International Electrotechnical Institute, 1987,
International Standard 8825.
10. Authors' Addresses
James M. Galvin
Trusted Information Systems, Inc.
3060 Washington Road, Route 97
Glenwood, MD 21738
Phone: (301) 854-6889
EMail: galvin@tis.com
Keith McCloghrie
Hughes LAN Systems, Inc.
1225 Charleston Road
Mountain View, CA 94043
Phone: (415) 966-7934
EMail: kzm@hls.com
James R. Davin
MIT Laboratory for Computer Science
545 Technology Square
Cambridge, MA 02139
Phone: (617) 253-6020
EMail: jrd@ptt.lcs.mit.edu