Electronic access control system and method

Technical Field

0001This invention relates to an electronic lock-and-key access control system and discloses a novel and very flexible approach for managing the physical security of such a system. Such systems are extensively used in industry, in the hotel business, in banks, and in other environments where it is desirable to control or restrict access to rooms or equipment. Many of these systems are using cards, sometimes even smartcards with built-in processing power. Such cards are generally magnetically or in other ways readable by a reader associated with each lock. To enable or block the use of certain keys, to restrict the access timely, or to change the security system in any other way, access information or messages have to be transmitted to - and sometimes from - the locks. This is often done via a cabling system connecting the locks to a central security management device, but other transmission means are also used. However, the more or less direct connection between management device and locks is characteristic for prior art approaches.

Background

0002Each and every one of us relies daily on keys to gain access to our cars, homes, and offices, to snap open the lock on our bicycles, or unlatch our postal boxes. Hence, keys are undeniably an integral part of modern society. Today, most of these keys are still of the mechanical type, as are the associated locks.

0003Although widely spread, these traditional, mechanical keys suffer from a number of shortcomings. First, in case of security breach, such as loss of a key, unauthorized duplication, or deprecation of trust assumptions, the lock must be replaced. Second, a key is valid as long as the corresponding lock remains in place. For example, once an access right is given to an individual by handing him/her a key, that right cannot be revoked unless the key is returned or the lock changed. As a consequence of these limitations, traditional lock-and-key systems do not allow to institute a time-based access policy, for example allowing access during office hours only, or protecting certain areas during certain times of the day. In the many environments where such requirements exist, traditional lock-and-key systems are no more sufficient.

0004Clearly, traditional keys offer only inflexible and rather limited management possibilities, but they are widely available, very reliable, and reasonably cheap. Still, replacement solutions with higher flexibility and extensive management possibilities have been invented and found their way to the market. For example in hotels and enterprises, electronic locks and appropriate electronic keys are widely used, providing very flexible and rather quick management of access rights.

0005An example of a system using smartcards of different kinds is shown in Lee US Patent 5 204 663 to Applied Systems Institute, Inc., which discloses an access control system with integrated-circuit cards, i.e. smartcards, some of which have a memory to store key access information and so-called transaction information. Whereas the key access information here relates to the immediate use of the key, e.g. the present access code of the card, relates the transaction information to other activity, e.g. a log of previous use of the lock as it is recorded in the lock's memory, specific new access codes to be uploaded into specific locks, or a log of failed entry attempts of the card user. Some of this transaction information is transferred from the lock into the card's memory, some from the card into the lock, and some just stored on the card for later readout.

0006One example of a lock which may be used in such an environment is the SaFixx smartlock, described in the Internet on http://acola.com, a lock made by ACOLA GmbH in Villingen-Schwenningen, Germany. The Safixx smartlock even exhibits some specific properties (discussed below) that may make it quite useful in connection with the present invention.

0007Other examples can be found in the literature. All of them have one disadvantage. They require a kind of "direct" access to each and every electronic lock in the system, be it that the locks are connected by a cabling or radio network to a management center, or that one has to walk up to each lock and "reprogram" it manually or electronically, as with the Safixx smartlock mentioned above or as with the system disclosed in the above-cited Lee US Patent 5 204 663. Either of these methods is burdensome: A network of cables to all locks in a hotel or an plant will usually cost more than the electronic locks themselves, a radio transmission system requires a transceiver and power in each and every lock and may thus be even more expensive (and failure-prone) than a cabling network, and walking to each and every lock may be simply impossible in a reasonable time frame.

0008US Patent 4,800,255 relates to an electronic access card comprising a printed circuit board. A microprocessor is carried by the printed circuit board. A keyboard is carried by the printed circuit board and is connected to the microprocessor for inserting information into the microprocessor. A coil is carried by the printed circuit board and is coupled to the microprocessor for receiving information to be conveyed to the microprocessor and for transmitting information from the microprocessor. Visual display means is coupled to the microprocessor for visually displaying information which is carried in the microprocessor.

0009US Patent 4,887,292 relates to an electronic lock system with a "lockout list" that identifies keys that are to be prevented from opening system locks. Date coded lockout lists are carried in the memories of both the system's locks and keys. When a lock and key engage, the date codes of the lock and key lists are compared to determine which list is that most up-to-date. This list is then transferred into the other unit, overwirting that unit's stale list. Subsequent interactions of either of these units with other locks and keys will further spread the most timely lockout list data. By this arrangement, new lockout lists can be disseminated throughout a lock system simply by loading them into one or more keys. As these keys are used with locks, and those locks used with other keys, the list will spread throughout the system.

0010Here, the invention intends to provide solutions. To summarize, it is an object of the present invention to avoid the above-described disadvantages and devise a reliable and flexible electronic lock-and-key system which has no need for "direct" (in the above sense) connections between locks and an associated management center controlling security within the system.

0011Another object is to simplify expansion of a given system when installing additional locks or other access or security devices, or to facilitate changes within a given system by installing new locks and/or exchanging existing ones.

0012A further object is to devise an architecture that allows a practically unlimited flexibility with regard to protocol changes or security updates between management center and locks. Such updates are necessary after a "successful" attack, or after card keys have been lost or stolen, or when security aspects of the system are changed, e.g. certain physical areas in a plant or lab become "restricted access" areas.

0013A specific object is to devise an architecture which allows an "on-the-fly" expansion of such a system.

The Invention

0014In essence, the novel approach according to the invention concentrates on using means, especially hardware, already existing in a usual electronic lock-and-key system. Access control and other information which need to be updated or changed is disseminated through existing smartcard or similar keys and the existing locks without any need for a connection between the latter and a central or distributed management center controlling this information. For this, a suitably adapted networking protocol is being used and, wherever appropriate, cryptography to protect the system against possible attacks.

0015In the following, the primary innovation, namely the propagation of access control information in cable-free environments, is discussed in more detail. It is also described how cost effective and easily manageable electronic locks can be generally implemented. Thereafter, attacks and implied security assumptions of such a system are discussed.

0016Contrary to most current electronic security systems, the system disclosed in this document does not require fixed network connectivity of any kind, be it wire-based or radio-based. Consequently, cost associated with cabling or radio transmission equipment is eliminated. Experience shows that this cost can be an order of magnitude higher than the cost of the locks themselves. It should be mentioned here that in this document the term "key" does not designate the traditional metallic key but rather any arbitrary carrier of information, e.g. smartcards or IBM's JavaCard. Similarly, the term "lock" will usually designate an electronic lock.

0017Instead of the locks receiving electrical power via a fixed cable, the required energy to operate the lock can be delivered either through the user's key or through an battery embedded in the lock or door. Clearly, in such a construction, power consumption must be kept to a strict minimum so that batteries last at least a few years. Electronic locks exhibiting the desired physical properties such as tamper resistance, operational reliability and long battery life, already exist on the market. The above mentioned SaFixx smartlock is an example of a lock exhibiting these properties. The present invention emphasizes the logical aspects of cable-free lock constructions, not so much the physical design.

Description of an Embodiment

0018In the following, the invention will be described in more detail in connection with a drawing which shows a block diagram of a typical access control system according to the invention, in particular the principle of message or information propagation in such a system.

0019As shown in the Figure, a <i>Door Domain Access Manager</i> (DDAM) 1 is the entity holding authority over an ensemble of locks 2. For the sake of simplicity, only some of them carry reference numbers 2a, 2b, etc. DDAM 1 plus the set of locks 2 constitute an <i>administrative domain</i> 4. It is assumed that DDAM 1 is composed of a single entity, as shown. This restriction may be removed by applying threshold signatures as described by Douglas R. Stinson in "Cryptography - Theory and Practice", CRC Press, 1996, and some more or less straightforward extensions to the communication protocol presented in this document.

0020Also shown is a plurality of keys 3, of which just a few have reference numbers 3a, 3b, etc., again to keep the drawing clean.

0021Each lock 2, usually associated with a door, as mentioned above, has a key (or card) reader of conventional type and a memory. The memory is organized to store two different sets of data or information, one first set consisting of one or more tokens (the number depends on the overall organization of the lock-and-key system), the second set comprising a message of so-called designated data, i.e. data designated for <u>other</u> keys or locks. The latter is a transient message which is only held for some time in the lock's transient memory part.

0022The keys 3 are of a very similar design as far as their memory is concerned. Their memory is also able to store two different sets of data or information. One first set consists again of one or more tokens (the number depending on the overall design of the lock-and-key system) and may be called a "key-activating" set. The second set consists again of a message of designated data, i.e. a transient message designated for <u>other</u> keys or locks, which message is held only for some time in the key's transient memory part. Simply said is this latter part of the key's memory the information transport path within the whole system.

0023Looking at the Figure, one recognizes that, using the keys as transport means for the designated data, there is usually a plurality of ways to each lock and key. A rather simple example shall clarify this.

0024Assume DDAM 1 issues a message for lock 2e (at the bottom), e.g. that a certain key, say key 3d, is not allowed to enter the room behind the door of lock 2e. This message designated for lock 2e is "entered" into the network via line 5, which latter may be a cable connection, any mobile data carrier, a wireless connection of any kind, etc. As soon as lock 2a has received the message, each and every key used at this lock 2a receives and stores in its transient memory part a copy of said message. Key 3a will transport it to lock 2b, key 3b will read and store it, transporting it to lock 2c (and others), from there, the set will travel via key 3c, lock 2d, key 3d and/or keys 3e to its designation, lock 2e. Note that even key 3d, who would be negatively affected by the transported message, may convey it to the designated lock 2e.

0025As soon as lock 2e receives the designated message, it issues an acknowledgement which now travels back through the system in the same fashion as the original message until it reaches DDAM 1, which closes the loop. The acknowledgement, since it recognizably refers to the message, erases the still stored (original) message in the both the keys and the locks through which it travels. Again, this is the overall function of the electronic lock system according to the invention in great simplification; details will be apparent from the subsequent description, which also addresses the theory behind some solutions.

0026As mentioned above, a user is granted access through a lock 2 (or door, which terms are used interchangeably in this document) only if his/her key exhibits an appropriate token issued by the DDAM. All tokens are minted with an expiration date. The DDAM may choose to revoke the access rights of any user at any time.

0027Keys are built so that they may hold multiple tokens as well as other types of data. Whether tokens are based on public key cryptography, shared secrets or some other type of security primitive is irrelevant at this stage.

0028In addition to storage capacity, doors may have computational capabilities allowing them to process tokens issued by the DDAM. The relation between DDAM and doors/locks 2 is that of master and slave. It is the DDAM's reserved prerogative to determine who may traverse which door and until what time and date. Consequently, the doors within an administrative domain 4 must completely trust the DDAM and blindly obey its orders. However, doors do not honor orders given by a DDAM outside their domain. There is no direct cabling linking the doors among themselves nor with the DDAM required. This mandates the alternative communication mechanism addressed above so that the DDAM and the locks can still exchange information. But note that the possibility of doors exchanging messages with one another is not necessarily precluded; also, there may be single selective "direct" connections between the DDAM and one or more selected locks.

Suitability Considerations

0029As users are required to present a valid token to access a door (by construction), the dissemination of positive access rights is not an issue. The DDAM can pass a token to a given user through a (secure) terminal or the token can simply be dropped at a door, e.g. lock 2a, until the user unwittingly picks it up. Renewing tokens that are about to expire can be accomplished in a similar fashion.

0030On the other hand, how can one ensure that a negative token (i.e. access revocation) reaches the relevant lock without excessive delay?

0031The solution for this problem is particularly appropriate in cases where all users happen to go through certain central doors, such as a building entrance, an elevator, or garage barriers. In another favorable scenario which is typical of business environments, rapid propagation of information is guaranteed by workers who arrive at work in the morning or by the maintenance/security personnel who frequently go through many doors. Looking at the Figure, door/lock 2a may be particularly suitable for that purpose.

0032As shall be explained later, if control information does not reach the destination node, i.e. the destined lock 2, the source, usually the DDAM, will not receive an acknowledgment. The DDAM can thus detect failure. It follows that in case there is no privileged central door and the target door is infrequently visited, the operator of the DDAM is alerted and he/she can always walk to the door in question. Critical doors requiring immediate information propagation could also be connected to the DDAM with a dedicated cable or by other "direct" means.

Network Model

0033As explained, to make certain that information is propagated in a timely fashion, user's physical keys 3 carry messages to and from between locks 2 and DDAM 1. Moreover, locks 2 act as a temporary repositories for messages designated for other locks. Doors and/or locks are modeled as nodes in a network and users as channels linking these nodes. For example, a temporary channel between lock 2a and lock 2b is established when a user travels through them, using key 3a. To improve connectivity, intermediary locks and user keys must act as repositories. As explained above, a user Alice may pick up a message from door 2a and leave it at door 2b. Then, user Bob with key 3e happens to go through door 2a and carries the message to its final destination, say door 2e.

0034More generally, both keys and doors contain a snapshot view of all current pending traffic, that is, all sent but not yet acknowledged packets. This, for any destination within the administrative domain. When a key is inserted in a door, both key and door/lock update their respective view of the network. Naturally, this update need not be interactive as it can be efficiently performed off-line. This way the user shall not be delayed waiting for the key and the lock to synchronize when going through a door. Off-line updates mandate that each entity possesses its own power source.

The Networking Protocol

0035The well-known TCP/IP protocol, as described by W. Richard Stevens in "TCP/IP Illustrated", Vol. I, Addison Wesley, 1994, serves as a basis and is simplified and somewhat optimized for the novel application. In an abstract sense, the problem can be stated as the building of a transmission control protocol (e.g. TCP) on top of a high-loss, varying-topology LAN.

0036First, it is assumed that each node of the network, that is a key, a door/lock or the DDAM, can be addressed individually. Hereafter, the term node may refer to a door/lock, a key, or the DDAM. This can be accomplished by using the addressing scheme of any networking layer protocol, in particular the Internet Protocol, as described by J. Postel: "Internet Protocol" in the publication of the Internet Engineering Task Force, September 1981, RFC 791.

0037As mentioned earlier, each node is responsible for propagating messages or packets, even if it is not the intended recipient, until the packet is acknowledged, at which time the packet is erased from the node's memory.

0038Like in TCP, described by J. Postel: "Transmission Control Protocol" in Internet Engineering Task Force, September 1981. RFC 793, each packet requiring acknowledgment is sent with a monotonically increasing sequence number. TCP deals with packet fragmentation caused by the IP layer, in the present homogeneous setting, packets are not liable to be fragmented. Hence the sequence numbers correspond to packets and not the number of bytes sent. This is unlike TCP, but has the advantage of being both simpler and more space efficient.

0039Once a message is received, the recipient issues an acknowledgment which is sent back to the sender, acknowledging the received packet. This acknowledgment consists of the usual addressing information and at least two other fields: a cumulative sequence number, denoted <i>n,</i> and a bit field of w bits length.

0040The actual value of <i>n</i> indicates that packets bearing a lower sequence number (inclusive) have been correctly received without gaps. The bit field provides information on out-of-order packets. The bit field loosely corresponds to the so-called "sliding window" of TCP.

0041As in TCP, packets outside of the sliding window, that is packets bearing a sequence number outside the range <i>n</i>+1 to <i>n</i>+<i>w</i>, are dropped. However, if an incoming packet completes to a sequence with no gaps, then <i>n</i> is incremented and an acknowledgment sent back to the source. Each packet received out of order but within the sliding window is tallied by turning on a corresponding bit in the bit field.

0042Thus, after the acknowledgment propagates back the sender of the packet, the source can selectively retransmit packets that have been lost. Packets that have correctly arrived, albeit out of order, need not be retransmitted. Also, packets that have been selectively acknowledged are dropped by intermediary nodes regardless of whether the source has received the acknowledgment or not, thus preserving precious memory space at each intermediary node.

0043It is important to note that acknowledgments presented here have an absolute ordering much like cumulative acknowledgments. Thus, they may be compared with one another. Consequently, only the newest acknowledgment need to be stored by the intermediate nodes as the newest acknowledgment overrides all previous ones.

Window Size

0044It is reasonable to assume that in most cases the frequency of revocation messages sent to a specific door or lock is low. Thus, very small window sizes (say 2) may amply suffice to satisfy the needs of even large organizations.

0045However, for exceptional cases, the window size can be increased at will. This can be accomplished by the DDAM sending out a broadcast message instructing all doors to adjust the window size for a given source/destination pair. Broadcast messages bear a fixed destination address recognized by all.

Clocking information

0046The DDAM could keep time on behalf of all other entities in the administrative domain. The DDAM sends timing updates as frequently as it can (i.e. each time it has contact with a user or door). As with acknowledgments, only the latest clock information need be stored by intermediate nodes, again saving precious memory space. In this way, there is no need for doors to have individual clocks.

Cryptographic Aspects

0047The cryptographic representation of a token can have overreaching effects on security management. The security implications of a secret key based architecture shall first be discussed. Next, the implications of a public key architecture are described.

Shared Secret

0048Tokens can be based on a shared secret between the DDAM and the locks. Let us assume the DDAM and lock <i>l</i>_j share the secret <i>s</i>_j. To grant user <i>u</i>_i access through <i>l</i>_i, the DDAM would authenticate the access string <i>a</i>_ij, where<maths id="math0001" num="(1)"><math display="block"><mrow><msub><mrow><mtext mathvariant="italic">a</mtext></mrow><mrow><mtext>ij</mtext></mrow></msub><mtext> = "Let </mtext><msub><mrow><mtext mathvariant="italic">u</mtext></mrow><mrow><mtext>i</mtext></mrow></msub><mtext> through </mtext><msub><mrow><mtext mathvariant="italic">l</mtext></mrow><mrow><mtext>j</mtext></mrow></msub><mtext> until 1/1/2001".</mtext></mrow></math><img file="EP1024239B1_D0001.tif" /></maths>

0049The authentication could be based on any secure keyed message authentication function, such as HMAC as described by R. Canetti et al in "HMAC: Keyed-Hashing for Message Authentication", Engineering Task Force, February 1997, RFC 2104. We have<maths id="math0002" num=""><math display="block"><mrow><msub><mrow><mtext mathvariant="italic">t</mtext></mrow><mrow><mtext>ij</mtext></mrow></msub><mtext> = HMAC(</mtext><msub><mrow><mtext mathvariant="italic">s</mtext></mrow><mrow><mtext>j</mtext></mrow></msub><mtext>, </mtext><msub><mrow><mtext mathvariant="italic">a</mtext></mrow><mrow><mtext>ij</mtext></mrow></msub><mtext>) ∥ </mtext><msub><mrow><mtext mathvariant="italic">a</mtext></mrow><mrow><mtext>ij</mtext></mrow></msub><mtext>,</mtext></mrow></math><img file="EP1024239B1_D0002.tif" /></maths> were <i>t</i>_ij is the access token and ∥ denotes string concatenation. Note that tokens could have other representations based on different types of authentication. User <i>u</i>_i would be granted access through <i>l</i>_j by presenting the token <i>t</i>_ij until 1 January 2001.

0050As in any digital representation of information, a token may be easily duplicated. In the shared secret setting, if a token can be "seen" by a rogue lock or any other unauthorized party, then the token has been effectively stolen. Indeed, the attacker could present the stolen token to the corresponding lock and obtain illegitimate access.

0051To mitigate the menace, security must be augmented by other means. For example, one may impose a universally unique serial number to be burned into each physical key, i.e. smartcard or the like, and issue tokens for a particular serial number. This may provide enough security to repel unsophisticated attackers. However, a determined adversary may print his/her own physical key with fake serial numbers.

0052For high security applications, it is crucial that the physical key allows only selective access to information. As a physical key is inserted in a lock, it should be able to verify whether the user has a specific token allowing passage through that lock. However, the particular lock should not be able to read tokens (on the key) relevant to other locks, possibly in other administrative domains. But even if a physical key allows selective access to information, lock <i>l</i>_x can pretend to be lock <i>l</i>_y and steal the token (for <i>l</i>_y).

0053The token duplication attack constitutes a severe threat, particularly if electronic locks become truly ubiquitous. Universal deployment means that physical keys will be used in mutually untrusting domains.

0054One can protect against this attack by requiring the lock to identify itself to the physical key; only then will the key reveal the token for the identified lock.

0055If this identification is based on a shared secret between the doors and the users, the number of shared secrets will explode for even medium sized organizations because each lock will have to share a secret with each user that goes through it. Thus, lock-user shared secrets may be suitable for small settings only.

0056Alternatively, a KryptoKnight type third party authentication technique may be used as described by P. Janson et al "Scalability and flexibility in authentication services: The KryptoKnight Approach" in IEEE INFOCOM '97, Tokyo, Japan, April 1997, and R. Bird et al "The KryptoKnight family of light-weight protocols for authentication and key distribution" in IEEE Transactions on Networking, 1995. Also, a Kerberos type third party authentication techniques could be employed, described by J. Kohl et al. "The Kerberos Network Authentication Service", V5, Internet Engineering Task Force, September 1993, RFC 1510 and by B. Clifford Neuman et al in "Kerberos: An authentication service for computer networks", IEEE Communications Magazine, 32(9): pp.33--38, September 1994. In such a scheme, all users and locks would share a secret with the DDAM. Since the DDAM shares a secret with all nodes, it can convince any node of the identity of another.

0057Let <i>s</i>_i represent the long term secret that user <i>u</i>_i shares with the DDAM, and <i>s</i>_j the secret that lock <i>l</i>_j shares with the DDAM. The DDAM would randomly choose an encryption key <i>k.</i> The access string <i>a</i>_ij would become:<maths id="math0003" num="(2)"><math display="block"><mrow><msub><mrow><mtext mathvariant="italic">a</mtext></mrow><mrow><mtext>ij</mtext></mrow></msub><mtext> ="Let through </mtext><msub><mrow><mtext mathvariant="italic">l</mtext></mrow><mrow><mtext>j</mtext></mrow></msub><mtext> user knowing </mtext><mtext mathvariant="italic">k</mtext><mtext>, until 1/1/2001"</mtext></mrow></math><img file="EP1024239B1_D0003.tif" /></maths>

0058The DDAM would issue the ticket <i>T</i>_ij defined as<maths id="math0004" num="(3)"><math display="block"><mrow><msub><mrow><mtext mathvariant="italic">T</mtext></mrow><mrow><mtext>ij</mtext></mrow></msub><mtext> = {HMAC(</mtext><msub><mrow><mtext mathvariant="italic">s</mtext></mrow><mrow><mtext>j</mtext></mrow></msub><mtext>, </mtext><msub><mrow><mtext mathvariant="italic">a</mtext></mrow><mrow><mtext>ij</mtext></mrow></msub><mtext>)∥ </mtext><msub><mrow><mtext mathvariant="italic">a</mtext></mrow><mrow><mtext>ij</mtext></mrow></msub><mtext>}</mtext><msub><mrow><mtext mathvariant="italic">s</mtext></mrow><mrow><mtext>j</mtext></mrow></msub><mtext> ∥ {k}</mtext><msub><mrow><mtext mathvariant="italic">s</mtext></mrow><mrow><mtext>i</mtext></mrow></msub></mrow></math><img file="EP1024239B1_D0004.tif" /></maths> where {m}_x denotes symmetric encryption of message <i>m</i> using encryption key <i>x</i> and <i>a</i>_ij is given by equation (2).

0059User <i>u</i>_i would strip the encryption {k}<i>s</i>_i from the ticket <i>T</i>_ij to retrieve the token <i>t</i>_ij. It would then obtain <i>k</i> from {k}<i>s</i>_i by decrypting it with its long term shared encryption key with the DDAM, that is <i>s</i>_i.

0060To pass tough door <i>l</i>_j, the user would present <i>t</i>_ij. The lock <i>l</i>_j would retrieve the encryption key <i>k</i> contained in a_ij by decrypting <i>t</i>_ij with its long term shared secret key, <i>s</i>_j. The door would then challenge the user for the knowledge of encryption key <i>k.</i> If the response checks with the issued challenge, then the user would be granted access.

0061Note that the DDAM has to precompute <i>T</i>_ij for given message a_ij and given nodes <i>u</i>_i and <i>l</i>_j. Remember that there is no direct channel (a connecting cable or the like) to the DDAM. Consequently, the Kerberos setting is unsuitable for situations where either the message being authenticated, or the intervening parties, are not known in advance. This is an important limitation if the application requires that disruption attacks be thwarted.

0062The verbosity of the access string a_ij should indicate to the reader that the protocol just described is not optimized, but is essentially intended for illustration purposes.

Public Key Setting

0063An alternative setting would have the DDAM and all users within the domain hold a public/secret key pair. Each lock within an administrative domain would be installed with the DDAM's public key. The DDAM would then certify the public keys of all users within its administrative domain.

0064The access token for user <i>u</i>_i through lock <i>l</i>_j would be the DDAM's signature on the message <i>a</i>_ij, as defined in equation (1). Whenever a user wishes to cross a door, it would present the token to the lock, which would verify DDAM's signature. To prevent impersonation, the door would require the user to prove knowledge of the secret key corresponding to the user's (DDAM certified) public key.

0065This proof can be based on any challenge response type protocol, such as described by 10 by C. P. Schnorr in "Efficient signature generation by smart cards", Journal of Cryptology, 4(3): pp.161--174, 1991, or by Uriel Feige et al in "Zero-knowledge proofs of identity", Journal of Cryptology: the journal of the International Association for Cryptologic Research, 1(2): pp. 77-94, 1988, or by Jean-Jacques Quisquater and Louis Guillou in "A practical zero-knowledge protocol fitted to security microprocessor minimizing both transmission and memory" in Christoph G. Gunther, editor, <i>Advances in Cryptology---EUROCRYPT88,</i> Volume 330 of Lecture Notes in Computer Science, pp. 123 -128, Springer-Verlag, May 1988.

0066The solution just presented separates the user's credentials from his/her public key. Thus, Alice's credentials may be transported to the relevant door without Alice having to intervene. To benefit from new credentials, Alice just has to prove her identity. Alice does not have to obtain a new secret from the DDAM nor change her public key. The new credentials are transparent to Alice. On the other hand, if one had tied each of Alice's credentials to a corresponding secret held by Alice, then Alice would have to communicate to the DDAM on a secure channel each time her credentials changed.

0067For example, in the Quisquater-Guillou (above) protocol, the DDAM would send Alice her new credentials, denoted by the string <i>J</i>, and a corresponding secret <i>B,</i> such that<maths id="math0005" num="(4)"><math display="block"><mrow><msup><mrow><mtext mathvariant="italic">JB</mtext></mrow><mrow><mtext>v</mtext></mrow></msup><mtext> = 1 mod </mtext><mtext mathvariant="italic">n</mtext></mrow></math><img file="EP1024239B1_D0005.tif" /></maths> where <i>n</i> is a public modulus <i>and v</i> is a public exponent. Let ϕ denote Euler's ϕ function. Only the DDAM knows the inverse of <i>v</i> mod <i>ϕ</i>(<i>n</i>), and hence can efficiently compute <i>B,</i> verifying equation (4).

0068To allow the access rights granted by <i>J,</i> the relevant lock would challenge Alice with a random number and check that Alice knows the secret corresponding to <i>J,</i> that is <i>B.</i> As usual, Alice does not reveal <i>B</i> as a result of the challenger-response protocol. Note that the secret <i>B</i> has to be sent by the DDAM to Alice on a secure channel.

False Acknowledgment Injections

0069In case an attacker is able to inject an acknowledgment message bearing a high sequence number, all intermediary hops will drop packets bearing a lower sequence number. Although this disruption attack will eventually be detected, it will momentarily prevent all communications between the DDAM and the locks.

0070In a variant attack, assuming that newer messages have higher priority, an attacker may flood locks by sending fake packets bearing high sequence numbers. The intermediate nodes will regard such packets as being newer and drop lower sequence numbered but genuine messages. This will also disrupt communications.

0071High security applications, especially those slated for universal deployment, mandate that all messages, including acknowledgments, be authenticated. Thus, all nodes, that is all doors/locks, users and the DDAM would have a public key certified by the DDAM.

0072The Kerberos type shared secret setting does not allow for messages to be authenticated for two dynamically chosen parties. Thus disruption attacks can only be prevented in the public key setting.

0073From the above discussion of the security implications of shared key versus public key based architecture it should have become clear that, because of the key explosion problem, a shared key architecture is suitable only for low security applications or for small scale deployment, whereas high security applications or universal deployment both mandate public key cryptography.

0074It should also have become clear that the invented flexible architecture for managing physical security using electronic locks and physical keys described hereinbefore does not require a permanent channel between the locks and a security management center. The question of timely propagation of access control information raised by the missing permanent connection is solved by having users act as transmission channels and locks as message repositories.