Computer system security via dynamic encryption

[0001] This application is a continuation-in-part of U.S. patent application Ser. No. 10/387,711, entitled Computer System Security via Dynamic Encryption, filed on Mar. 13, 2003, and claims the benefit of the filing date thereof. The entire specification of the parent application is incorporated herein by reference.

[0002] 1. Field of the Invention (Technical Field)

[0003] The present invention relates to the field of computer system security, more particularly to a dynamic data encryption and node authentication method and system that distributes the complexity of the encryption algorithm over the dynamics of data exchange and involves full synchronization of encryption key regeneration at system nodes, independent of the node clocks.

[0004] 2. Background Art

[0005] The fundamental objective of cryptography is to enable users to communicate securely via an insecure shared data communication channel or system environment, maintaining data integrity, privacy, and user authentication. Over the past century, various cryptography systems have been developed which require a great deal of time to break even with large computational power. However, if an intruder obtains the encryption key, the encryption mechanism, and probably the entire system security, is compromised and a new key is required.

[0006] In order to make an encryption system nearly impenetrable to an intruder, two strategies are commonly used: 1) a long encryption key, and/or 2) a complex encryption function. A key of length n bits has a 2ⁿ search space. Therefore, for large values of n an intruder needs to spend more than a lifetime to break the cipher. Also, simpler encryption functions provide a less secure encryption system. For instance, an encryption code that applies the logic XOR function is easy to decipher no matter how long the key length is. This is because the XOR operation is performed on one bit of data and its corresponding bit from the encryption key, one bit at a time. The deciphering approach of such simple encryption functions by an intruder is based on the divide-and-conquer mechanism. The intruder first deciphers individual key fragments, which is relatively uncomplicated to accomplish due to the simple linearity of the XOR function, then reconstructs the entire key once all of the individual fragments are obtained. It is more difficult to apply such a divide-and-conquer approach to break the key of a nonlinear exponential encryption function, such as used in the Rivest-Shamir-Adelman (RSA) system.

[0007] At present, there are two major cryptography system philosophies: 1) symmetric systems (static or semi-dynamic key), and 2) public key systems (static key). In symmetric systems, e.g., DES, AES, etc., a key is exchanged between the users, the sender and receiver, and is used to encrypt and decrypt the data. There are three major problems with symmetric systems. First, exchanging the key between users introduces a security loophole. In order to alleviate such a problem, the exchanged key is encrypted via a secure public key cryptography system. Second, the use of only one static encryption key makes it easier for an intruder to have an ample amount of time to break the key. This issue is addressed by the use of multiple session keys that are exchanged periodically. Third, and more importantly is the susceptibility to an insider attack on the key. This is referred to as the super user spying on the setting duck static key inside the system, where the time window between exchanging keys might be long enough for a super user, who has a super user privilege, to break in and steal the key.

[0008] In the RSA public key cryptography system, a user (U) generates two related keys, one is revealed to the public, deemed the public key, to be used to encrypt any data to be sent to U. The second key is private to U, called the private key, and is used to decrypt any data received at U, which was encrypted with the corresponding public key. The RSA cryptography system generates large random primes and multiplies them to get the public key. It also uses a complex encryption function such as mod and exponential operations. As a result, this technique is unbreakable in the lifetime of a human being for large keys, e.g., higher than 256 bits, and also eliminates the problem of the insecure exchange of symmetric keys, as in a DES system. However, the huge computational time required by RSA encryption and decryption, in addition to the time required to generate the keys, is not appealing to the Internet user community. Thus, RSA cryptography is mainly used as one shot solid protection of the symmetric cryptography key exchange.

[0009] In the RSA public key system, if a first user (U_A) requests a secure communication with a second user (U_B), the latter will generate a pair of encryption keys: public E_B and private D_B. An internal super user spy (S), with a helper (H) intruding on the communication line externally, can easily generate its own pair of keys, a public E_S and private D_S, and pass D_S and E_B to H. Then S can replace the public key E_S with its own public key E_S. Thus, all data moving from U_A to U_B will be encrypted using E_S instead of E_B. Now H can decrypt the cipher text moving between U_A and U_B using the private key D_S, store it, and re-encrypt it using the original E_B, in order for U_B to receive and decrypt it without any knowledge of the break that occurred in the middle. Such an attack is typically called the super-user-in-the-middle attack.

[0010] Even though they are secure against outsider attack, both the symmetric and public key cryptography systems are still vulnerable to insider attacks. By obtaining the key at any time of a secure session, an intruder can decipher the entire exchanged data set, past and future. Further, a super user can easily steal a static symmetric key and send it to an outside intruder to sniff and decrypt the cipher text, particularly in the DES and AES systems.

[0011] A common way to protect a static encryption key is to save it under a file with restricted access. This restriction is not enough, however, to prevent a person with super-user privilege from accessing the static key in the host file. Even when keys are changed for each communication session, for example in the Diffie-Hufman system, there is a time window enough for the super-user to obtain the semi-static key. In most crypto systems, once the key is found the previous and future communicated data are no longer secure.

[0012] Various other attempts have been made to circumvent intrusion by outside users through encryption of communicated data. Examples of such methods include that described in U.S. Pat. No. 6,105,133 to Fielder, et al., entitled, Bilateral Authentication and Encryption System; U.S. Pat. No. 6,049,612 also to Fielder, et al., entitled, File Encryption Method and System; and U.S. Pat. No. 6,070,198 to Krause, et al., entitled, Encryption with a Streams-Based Protocol Stack. While the techniques described in these patents may be useful in preventing unwanted intrusion by outsiders, they are still prone to attack by the super-user-in-the-middle.

[0013] The present invention alleviates the problems encountered in the prior art, providing continuous encryption key modification. A new key is generated from a previous key as well as from a previously encrypted data record. The regenerated key is then used to encrypt the subsequent data record. The key lifetime is equal to the time span of record encryption, reducing the possibility that an intruder will break the key or that a super-user will copy the key. The present invention also alleviates the super-user-in-the-middle attack. An intruder must obtain an entire set of keys, at the right moment, without being noticed, in order to decrypt the entire ciphered message.

[0014] The present invention also reduces computational overhead by breaking the complexity of the encryption function and shifting it over the dynamics of data exchange. Speed is also improved through the use of a simple XOR logic encryption function. A shuffling mechanism based on a dynamic permutation table, which is generated from the current session key, is coupled with the XOR logic operation, to strengthen the encryption function. Encryption is fully automated and all parties, the source user, destination user, and central authority, are clock-free synchronized, and securely authenticated, at all times. The dynamic key encryption system of the present invention is deployable at any level of a system, as there is complete synchronization between parties.

[0015] The present invention further eliminates the possibility of an intruder obtaining additional keys, and therefore additional data, in the circumstance where an intruder is able to break one of the keys.

[0016] A previously encrypted data record is combined with a previous key to create a new key for encrypting a subsequent data record. Hence, if an intruder were to break one key, the intruder could only decrypt its corresponding data record and nothing more, past or future ciphered data.

[0017] The present invention is a method of providing a secure data stream between system nodes. The method includes encrypting data at a node with an encryption key, or dynamic session key (DSK), selecting encrypted data, and regenerating a new DSK with an encryption key and selected encrypted data. Selecting encrypted data includes selecting encrypted data using a byte from a previous encryption key as a seed of random generation.

[0018] The new DSK is generated by performing a logic operation on a previous DSK and selected previously encrypted data record. Preferably, the logic operation is an XOR operation. The previously encrypted data record and previous DSK are XORed to form an expanded key, ExpK. Bytes are randomly selected from the ExpK to generate the new DSK, using a byte from a previous DSK as a seed of random generation.

[0019] A data record is encrypted with a DSK by performing a logic XOR operation on the data and DSK to form a temporary cipher. Portions of the cipher are then permuted to form another cipher, which is then transmitted over a data stream to the destination user node.

[0020] Blocks of n previously encrypted data records of a time increment (T), each record of m bytes, and a corresponding number of n DSKs, each of m bytes, are combined to form n new DSKs. The n new DSKs are then used to encrypt the subsequent block of n data records. The process continues until all data records are encrypted and transmitted from the source user node.

[0021] The method further comprises the step of receiving encrypted data at a destination user node and decrypting the received encrypted data with a DSK. Once the encrypted data is decrypted, new DSKs are regenerated at the destination user node using selected decrypted data and a previous DSK. A central authority node is used to assure that both the source and destination nodes begin with the same initial DSK so that the destination node can properly decrypt the received encrypted data.

[0022] The present invention is further a system for providing a secure data stream between a source programmable apparatus and a destination programmable apparatus. The system comprises: a source programmable apparatus; a data stream created by the source programmable apparatus; means for encrypting data of the data stream with a DSK; and means for regenerating a new DSK using selected previously encrypted data. The system also includes a destination programmable apparatus in electrical communication with the source programmable apparatus; means for transmitting encrypted data to the destination programmable apparatus; means for decrypting the encrypted data received at the destination programmable apparatus with a DSK; and means for regenerating a new DSK using selected previously decrypted data.

[0023] A primary object of the present invention is to provide a dynamic encryption method and system having no static keys, public or private, that are susceptible to a security breach. Another primary object of the invention is to provide improved security to a data stream between a source and destination user, in particular to provide improved security against a super-user having insider privileges. Another primary object of the present invention is to provide such improved security at a high speed, in a secure system environment, via mutually authenticated users and CAs. Yet another primary object of the present invention is to provide a dynamic encryption method and system wherein the breaking of one key by an intruder will aid only in decrypting its corresponding data record and nothing more, past or future ciphered data.

[0024] A primary advantage of the present invention is that it is fully automated, with all system nodes synchronized and mutually authenticated, to ensure security. Another primary advantage of the invention is that it is simple and fast, yet secure against spying by an internal super-user or outside intruder due to the large number of dynamic keys (n) that would need to be broken or compromisedone key per ciphered data record, (n) parallel sessions of encryption, and zero entropy of the ciphered text. Yet another primary advantage of the invention is that it minimizes the exchange of keys between users and/or the CA. Still yet another advantage of the invention is that an initial DAK is securely exchanged between a user and CA which is continuously regenerated during the entire life of the user, allowing the user and CA to synchronize (realign DAKs) and authenticate to one another as needed for a communication session or when there is a disaster misalignment between a user and CA.

[0025] Other objects, advantages and novel features, and further scope of applicability of the present invention will be set forth in part in the detailed description to follow, taken in conjunction with the accompanying drawings, and in part will become apparent to those skilled in the art upon examination of the following, or may be learned by practice of the invention. The objects and advantages of the invention may be realized and attained by means of the instrumentalities and combinations particularly pointed out in the appended claims.

[0026] The accompanying drawings, which are incorporated into and form a part of the specification, illustrate a preferred embodiment of the present invention and, together with the description, serve to explain the principles of the invention. The drawings are not to be construed as limiting the invention.

[0027] FIG. 1a is a diagrammatic overview of a central authority (CA) generating daemons to manage users' dynamic authentication keys (DAKs) in accordance with the present invention;

[0028] FIG. 1b is a diagrammatic overview of secure communication between users in accordance with the present invention;

[0029] FIG. 2 is a diagrammatic illustration of a user registration request to a CA in accordance with the present invention;

[0030] FIG. 3 is a diagrammatic overview of synchronization and authentication between users and a CA, and initial generation of the DSK by the CA in accordance with the present invention;

[0031] FIG. 4 is a diagrammatic illustration detailing the method of FIG. 3;

[0032] FIG. 5 is a diagrammatic illustration of synchronization of DAKs between a CA and a user in accordance with the present invention;

[0033] FIG. 6a is a diagrammatic illustration of the method whereby a CA authenticates a user in accordance with the present invention;

[0034] FIG. 6b is a diagrammatic illustration of the method whereby a user authenticates a CA in accordance with the present invention;

[0035] FIG. 7a is a diagrammatic illustration of the method whereby a CA freezes and resumes regeneration of users' DAKs upon occurrence of a CA shutdown event, in accordance with the present invention;

[0036] FIG. 7b is a diagrammatic illustration of the method whereby a user freezes and resumes regeneration of its DAK upon occurrence of a user shutdown event, in accordance with the present invention;

[0037] FIG. 8 is a diagrammatic illustration of secure communication establishment at the source user node in accordance with the present invention;

[0038] FIG. 9 is a diagrammatic illustration of secure communication establishment at the destination user node in accordance with the present invention;

[0039] FIG. 10 is a diagrammatic illustration of a user handshaking method with a CA in accordance with the present invention;

[0040] FIG. 11 is a diagrammatic illustration of the dynamic encryption and permutation method of the present invention;

[0041] FIG. 12 is a diagrammatic illustration of the dynamic decryption and permutation method of the present invention;

[0042] FIG. 13a is a diagrammatic illustration of the DSK regeneration method of the present invention;

[0043] FIG. 13b is a diagrammatic illustration of the DSK regeneration method of FIG. 13a demonstrating the utilization of previously encrypted data records from different time interval blocks in the regeneration process;

[0044] FIG. 14 is a diagrammatic illustration of the DAK regeneration method of the present invention;

[0045] FIG. 15a is a diagrammatic illustration of the formation of the auxiliary static key using the initial DAK in accordance with the present invention; and

[0046] FIG. 15b is a diagrammatic illustration of the formation of the auxiliary static key using the DAK and previous DSK in accordance with the present invention.

[0047] The present invention is a dynamic symmetric key encryption method and system for network security. The invention is implemented between end-users (U) and central authentication authorities (CAs) for user authentication purposes, and between end-users for secure exchange of digital data. The users and CA reside at their respective nodes, or programmable apparatuses, such as at one or more computers, and are in electrical communication, such as via a computer network. Communicated data flows over a data stream between the user and CA programmable apparatuses. Computer-readable memory provides storage for the data, dynamically changing keys, and other variables, as needed to allow for the regeneration of subsequent dynamic keys, and to carry out other necessary processes within the computer. Means are provided on the programmable apparatuses for performing all of the various methods involved in the dynamic encryption method. Such means include primarily computer-readable means, such as software, and the necessary related hardware.

[0048] The encryption method of the present invention distributes the complexity of the encryption algorithm over the dynamics of the data exchange, involving previously encrypted data as well as the previous key in the process of regenerating a new dynamic encryption key. Thus, there is a newly generated key for the encryption of every data record, yielding zero entropy between the cipher and the plain data. There are no static keys, public or private, that are susceptible to a security breach. In order to guarantee security, the encryption method of the present invention is preferably deployed in a system of registered end-users with a CA, whereby the CA maintains user connections' authentication and secures distribution of symmetric encryption session keys.

[0049] The invention employs two types of dynamic keys, namely dynamic authentication keys (DAKs) and dynamic session keys (DSKs). The former is used to mutually authenticate users and CAs; it is continuously regenerated throughout the existence of the user and the CA. The latter exists when the need arises for a secure data exchange session between users; its regeneration is maintained only through the life cycle of such a session. The placement of the initial DSK at users' nodes is securely carried out by the CA, and encrypted using the users' DAKs. The CA maintains an array of DAKs, one per user.

[0050] The invention further employs an auxiliary static key K, which is formed based on the DSK and the DAK, given that the user has established a session; otherwise, it is formed from the initial DAK only. This static key is continuously involved in the regeneration of the DAK. The auxiliary static key adds another dimension of security to the process against insider attacks, as it involves more dynamics to the regeneration of the DAK by allowing the contribution of the DSK to the process, every new communication session between users. The static nature of K is not to be exploited by the attacker since its exploitation does not lead to any important information; its relation to the DSK and the DAK is not reversible, i.e., K is manufactured from DSK and DAK, yet neither DSK nor DAK can be obtained from K.

[0051] The major role of the CA is to maintain the registered users' DAK generation and the secure delivery of symmetric session keys (DSKs) between the users. The CA also authenticates the source user to the destination user and vice versa, while authenticating itself to both users. Unless a user is pre-registered with the CA, it is nearly impossible for an intruder to have the same synchronized dynamic key with the CA, as a registered user. The only explicit user identity verification is carried out once, at the time the user first registers with the CA. Any consequent authentication is implicitly processed, without the need to exchange any plain keys, which would require a secure channel. The authentication method involves the three parties to any connection: source user, destination user, and CA, authenticating each other via their corresponding synchronized dynamic key.

[0052] Referring to FIG. 1a, a diagrammatic overview of a CA generating daemons to manage users' dynamic authentication keys (DAKs) in accordance with the present invention is shown. Registration of trusted users, users that have for example, provided a valid certificate of authentication or who are referred by a previously-registered third party user, occurs over a secure channel, for example, using RSA encryption, where the CA provides the user with an initial DAK. Referring to FIG. 1b, a diagrammatic overview of user communication encrypted with a dynamic session key (DSK) initially generated and sent by a CA is shown. Upon any user's connection request, a communication child process is forked at the CA, as well as at the user node. Each child process runs on the behalf of its parent during the whole communication session in order to avoid disturbance of the DAK generation process in the event of synchronization or authentication failure, for example, due to a false connection request, hardware failure, etc.

[0053] The CA is responsible for secure user authentication and generation of the initial trusted user dynamic authentication key, DAK. Every user must register with the CA in order to obtain its initial DAK, which is exchanged via a secure channel, e.g., using RSA. Upon the initiation of any new user's secure data exchange session, both end-users and the CA freeze their DAK generation and synchronize by establishing the same dynamic change. After the user and the CA are synchronized (their DAKs are aligned), they both yield the same randomly generated DAK at both nodes. This DAK is used in the authentication method as the encryption and decryption key, because it is maintained only by the CA and the user.

[0054] Referring to FIG. 2, a diagram further illustrates the user registration method with a CA. The user starts the registration process by sending a request 10, effectively requesting the generation of an initial value of its DAK, to the CA including authenticated documentation of its identity and purpose of registration, i.e. revealing it is a trusted user. Upon approval of the user's credentials by the CA, the CA starts a daemon related to the requesting user, and randomly selects an initial DAK, sending a copy to the user via a secure channel 12, for example, using the well-known RSA technique where the CA uses the user's public key to encrypt the newly generated DAK. Then the CA starts a daemon that permanently regenerates the DAK. Upon the reception of the initial DAK, the user starts a daemon in order to permanently regenerate the DAK. From that point forward, the user and CA randomly regenerate the next DAK every t period, 14, 15, based on the previous generated DAK and the auxiliary static key K. The user and CA also maintain a number-regeneration-counter (NRC). This method of regenerating new DAKs is depicted in greater detail in FIGS. 14 and 15.

[0055] Attention is now turned to FIG. 14. The diagram of FIG. 14 illustrates the continuous DAK regeneration method, where DAKj represents the j^th byte of the dynamic authentication key, Kj represents the j^th byte of the auxiliary static key, ExpKh represents the h^th byte of an expanded key, 1jm, 1h2 m. Each unit R represents the random selection of one byte among the ExpK's 2 m bytes. An expanded key, ExpK 186 of twice the size of the DAK 182 is generated as follows. Each of the DAK, K, and ExpK are divided into (m/2) regions, indexed from 1 to (m/2). Each region of DAK, 182 and DSK, 184 is of length 2 bytes, and each region of ExpK, 186 is 4 bytes. The indices of the four consecutive bytes of any region r in the ExpK are indexed with the values: 4r-3, 4r-2, 4r-1, and 4r. The indices of the two consecutive bytes of any region r, in the DAK and K, are indexed with the values: 2r-1 and 2r. The four bytes of region r in the ExpK are filled from DAK and K as follows:

[0056] ExpK4r-3DAK2r-1

[0057] ExpK4r-2DAK2r

[0058] ExpK4r-1(DAK2r-1) XOR (DAK2r)

[0059] ExpK4r(DAK2r-1) XOR (DAK2r) XOR (K2r-1) XOR (K2r)

[0060] It will be understood by those of skill in the art that, ExpK can alternatively be comprised of any number of bytes, e.g., 2 m, 3 m, 8 m, etc., and the invention is not limited to any particular size for ExpK. If ExpK were of a different, greater number of bytes than 2 m, then the logic operation would be altered as needed to fill the bytes of ExpK. It will also be understood that the byte positions chosen to be XORed together to fill the positions of ExpK could similarly be altered, and still remain within the inventive scope of the dynamic encryption and authentication method. Similarly, alternative logic operations could be substituted for the XOR operation.

[0061] Once ExpK is created, a random selection of m bytes from the 2 m bytes of ExpK is taken 190 based on the first byte of the DAK (DAK1) as the random function seed 188. Alternatively, a randomly selected byte of the DAK can be used as the random function seed. This function generates a sequence of m random numbers, in the range between 1 and 2 m, each of which represents the index of the ExpK byte to be placed as the next byte of the regenerated DAK 192. The operation depicted in FIG. 14 is performed at both the user and the CA nodes. In order to maintain synchronization control, a number-regeneration-count for the dynamic authentication key (DAK_NRC) 194 is maintained and incremented after each DAK regeneration. This method can be performed periodically or aperiodically (with mutual consent of CA and the user) according to the implementation of the invention.

[0062] Continuing on to FIGS. 15a and 15b, a diagram illustrates the method of forming the auxiliary static key K. FIG. 15b illustrates the formation of K using the CA-user aligned DAK, where DAKj represents the j^th byte of the dynamic authentication key, DSKj represents the j^th byte of the last dynamic session key, Kj represents the j^th byte of the auxiliary static key K, and ExpKh represents the h^th byte of an expanded key, 1jm, 1h2 m. Each unit R represents the random selection of one byte among the ExpK's 2 m bytes. Each static key K is created using the CA-user aligned DAK, 208 and the last user session DSK, 210. First, an expanded key (ExpK), 212, of twice the size of K is generated. Each of the DAK and the DSK is divided into (m/2) regions, indexed from 1 to (m/2). Each region of the DAK and the DSK is of length 2 bytes, and each region of ExpK is 4 bytes. The indexes of the four consecutive bytes of any region r, in the ExpK, are indexed with the values: 4r-3, 4r-2, 4r-1, and 4r. The indexes of the two consecutive bytes of any region r, in the DAK and the DSK, are indexed with the values: 2r-1 and 2r. The four bytes of region r in the ExpK are filled from DAK and DSK as follows:

[0063] ExpK4r-3DAK2r-1

[0064] ExpK4r-2DAK2r

[0065] ExpK4r-1(DAK2r-1) XOR (DAK2r)

[0066] ExpK4r(DAK2r-1) XOR (DAK2r) XOR (DSK2r-1) XOR (DSK2r)

[0067] Then, a random selection of m bytes from the 2 m bytes of ExpK is performed, 216, based on the DAK1 byte as the random function seed, 214. Alternatively, any randomly selected byte can serve as the random function seed. This function generates a sequence of m random numbers, in the range between 1 and 2 m, each of which represents the index of the byte to be placed as the next byte of K, 218. This operation is performed at both the source and destination user nodes. This method is used to form K when a user communication session is taking place and DSKs are being regenerated.

[0068] FIG. 15a illustrates the formation of K, 206 using the two copies of the initial CA-user aligned DAK, 196, 198. This method is used to form K when a communication session is not taking place, i.e., the user DSK does not exist. Thus, the mechanism of FIG. 15a follows the same pattern as FIG. 15b, except replacing the non-existing DSK by another copy of the initial DAK.

[0069] Returning to FIG. 3, a diagrammatic overview of synchronization and authentication between users and a CA, and initial generation of the DSK by the CA is shown. DAK regeneration starts at both the user and the CA nodes, upon user registration, where each generates the same sequence of random DAKs, with the initial DAK as a seed, even after the user-CA connection is terminated. Thus, key exchange between users, as well as between users and the CA, is minimized to maintain a high level of security. Users and the CA instead remain synchronized at all times with respect to DAK regeneration. When the user-CA connection is terminated after initial request for user registration with the CA, synchronized DAK regeneration continues off-line. Permanent regeneration of the DAK is maintained via a daemon running at each registered user, and a corresponding daemon running at the CA.

[0070] With continuing reference to the center portion of the diagram of FIG. 3, in order to establish a connection between a source user U_s and a destination user U_d, U_s's DAK regeneration daemon forks a child communication process U_sCOM, which freezes its version of the DAK and its NRC, and sends a connection request to the CA including synchronization information. Upon the reception of such request, the CA's communication server forks a communication child process CA_COM, which will notify U_d of the connection with U_s. Upon the acceptance by U_d to the connection, U_d's DAK regeneration daemon forks a child communication process U_dCOM, which freezes its version of the DAK and its NRC, and sends synchronization information back to the CA_COM. Then, the CA_COM snapshots both users' DAKs/NRCs from their corresponding CA's DAK daemons, and starts the synchronization and authentication processes with both users' communication child processes.

[0071] Upon successful mutual synchronization and authentication involving the three session parties, the CA_COM generates a random dynamic session key (DSK) and encrypts it using the aligned DAK of each user, and sends it to both users. After receiving the encrypted DSK, each of the two users' child communication process decrypts it using its respective aligned DAK, and starts a secure communication session with the other child communication process, via the decrypted DSK.

[0072] In case of failure, the child processes are terminated without interruption to the parent processes and/or daemons. This feature provides protection against a synchronization disturbance attack. In such an attack, an intruder imitating a registered user or a CA might force the three DAK number-regeneration-counters (NRCs) to freeze, which will create a chaotic state, but only in the child processes. The counters are continuously counting, i.e., DAKs are continuously regenerated, in the daemons without stopping, except when rebooting. The child processes snapshot, or freeze, the DAK/NRC for use in the synchronization, authentication, and secure DSK exchange. Thus, the continuously running DAK daemons are unaffected in the parent processes in the event of a failure to establish a connection, or in the event of a synchronization disturbance.

[0073] However, in the event of successful synchronization and authentication, the child processes at the users' nodes return the aligned DAK and the newly generated DSK to their respective parent daemons in order to initialize a new state for DAK regeneration, forcing the daemon to discard the current value of DAK and DSK (if exists), and consider the newly returned versions in the DAK regeneration process. Also, the CA_COM return the two aligned DAKs and the newly generated DSK to their respective DAK daemons at the CA in order to initialize a new state for DAK regeneration, forcing both local users' daemons to discard their current DAK and DSK (if exists) values, and consider the newly returned versions in the DAK regeneration process. Then, the CA_COM terminates successfully, whereas the U_sCOM and U_dCOM start a secure communication.

[0074] Referring to FIG. 4, a diagrammatic overview of synchronization, authentication, and generation of DSK by a CA in response to a request from a source user (U_s) to communicate with a destination user (U_d) is provided. (See also FIG. 3.) Source user U_s requests a DSK generation from CA to communicate with destination user U_d, and sends its frozen DAK_NRC along with the request, 16. The CA forks a CA_COM, which snapshots the two users DAKs, namely CA_DAKU_s and CA_DAKUd, and requests U_d to send its DAK_NRC, 18. This request notifies U_d that U_s is trying to establish a secure communication with it. Once the CA has received the U_d DAK_NRC, 20, the synchronization process is initiated, 22.

[0075] Synchronization ensures that the CA has DAK values identical to the users' DAK values, despite message propagation delay. The CA ensures that its locally snapshot DAK for U_s and the corresponding frozen DAK at U_s's node are aligned, and also ensures that its locally snapshot DAK for U_d and the corresponding frozen DAK at U_d's node, are aligned. To compensate for propagation delay and align the corresponding CA's DAK with each of the parties' DAKs, the difference (x) in the number of key regeneration counts (NRCs), between CA and each user, will be considered by the lagging party, which will regenerate its DAK an extra x times (See FIG. 5.)

[0076] If alignment is not achieved with both users, 26, CA ignores the synchronization effects from the non-synchronized user(s), sends an abort message to both users before killing its communication child process, and cancels the communication due to lack of synchronization.

[0077] In the event of successful synchronization with both users, 24, the CA launches an authentication method, 28, to certify their identity. (FIGS. 6a and 6b.) If successful authentication of both users is not achieved, 32, CA ignores any synchronization effects of the non-authenticated user(s), sends an abort message to both users before killing its communication child process, and cancels the communication due to lack of authentication. If both users are fully authenticated and synchronized with CA, 30, the CA randomly generates an initial DSK and sends it to both users, encrypted with each user's corresponding aligned DAK, to begin data exchange, 34. This encryption process, 34, is identical to the process described in FIG. 11, except that DSK_i is replaced by DAK, and D_i is replaced by the initial DSK. After the entire process is completed, successful or unsuccessful, the CA_COM terminates and returns its status to the parent process.

[0078] Referring to FIG. 5, a diagram illustrates synchronization of DAKs between a CA and a user, based on the number-regeneration-count for the DAK at each node. Initially, the number-regeneration-count of the DAK for user (U) at the CA, (CA_DAK_NRCU), is compared to the number-regeneration-count of the DAK at the user's node (U_DAK_NRC), 36. If the two NRCs are equal, then the CA and user are synchronized. If the comparison of the NRCs is outside of a predetermined acceptable range, a failure-to-synchronize message is reported, 40. This occurs when CA_DAK_NRC UU_DAK_NRC, 38, is larger than a predetermined value, and the communication is terminated.

[0079] If the comparison of the NRCs is within the predetermined acceptable range, then the lagging party performs a number of DAK regenerations equal to the calculated difference in order to synchronize (i.e., align the DAKs) with the other party. For example, if the CA NRC lags behind that of the user, 42, then the CA performs (U_DAK_NRCCA_DAK_NRC U) regenerations of its DAK in order to align with that of U, 44. If the user NRC lags behind that of the CA, the CA sends a synchronize message, including the calculated NRC difference, 46, so that the user can perform the appropriate number of regenerations to synchronize with the CA. Once the user performs the regenerations, it signifies that it has done so to the CA, 48.

[0080] Once the parties are synchronized, mutual authentication of DAKs is performed to ensure the parties indeed share the same DAK. FIGS. 6a and 6b illustrate the mutual authentication method. FIG. 6a illustrates authentication of a user by a CA, and FIG. 6b illustrates authentication of a CA by a user. The process begins by CA generating a random number, or nonce, N. CA sends N and E(N) to the user, where E(N) is the encrypted version of N using the CA_DAKU, the shared and aligned DAK, as well as an authenticate message, 50. The user decrypts E(N) using its frozen DAK, which should be identical to CA_DAKU, that of the CA, 64 and verifies that D(E(N))N, 66. The user thus authenticates the CA, 68. If D(E(N)) does not equal N, then the user failed to authenticate the CA and the connection is aborted, 70.

[0081] When the user has successfully authenticated the CA, 68, the user encrypts N² with its DAK, (E(N²)), and sends E(N²) back to the CA, as part of the authentication acknowledgment to complete the authentication process, 68. Upon receiving the authentication acknowledgment back from the user, 52, CA decrypts the received ciphered number, again using the aligned CA_DAK U, 54 and compares the decrypted value with the value of N², 56. If they are not equal, the CA reports a failure to authenticate the user, 60, and aborts the establishment of a connection. If they are equal, the user is successfully authenticated by the CA, i.e., has been registered with the CA, 58. CA performs the same mutual authentication method with the second user before user-to-user communication takes place. When all parties have been mutually authenticated, CA proceeds to DSK generation, 62, the last step of FIG. 4.

[0082] As systems are prone to unpredicted shutdown events, such as power loss, the dynamic encryption method and system automatically freezes and resumes key regeneration upon occurrence of such events. Referring to FIG. 7a, a diagram illustrating freezing and resuming regeneration of DAKs when a CA experiences a shutdown event, is shown. FIG. 7b shows freezing and resuming regeneration of DAKs when a user experiences a shutdown event. Referring to FIG. 7a, a CA is shown to experience a shutdown event, 72. The CA immediately sends a freezeDAK-regenerating message to all previously registered users, 74, as part of its shutdown handler routine. Meanwhile, the CA saves all users' DAKs into a temporary file, 76. After shutting down, 78, for a time period , the CA reboots and reloads all the previously saved DAKs. The CA then requests the DAK_NRC from all users to ensure validity of the current DAKs, 80. Then, the CA starts the synchronization process, 82 as described with respect to FIG. 5. The CA then initiates the mutual authentication process with all successfully synchronized users, 84. (FIGS. 6a and 6b.) Finally, the CA sends a resume-DAK-regenerating message to its registered and successfully synchronized and authenticated users, in order to resume realigned regeneration of the DAKs, 85. A similar method is performed by the user in the event of a user's system shutdown, as depicted in FIG. 7b.

[0083] After a user is registered with a CA, the user may request a secure communication with another user. Referring to FIG. 8, a diagram illustrates secure communication establishment at the source user (U_s) side. The source user first determines whether it is registered to the CA, 86. If not, the user performs the registration procedure and receives its initial DAK via a common secure channel, 88. (See also FIG. 2.) Then, the source user's DAK daemon forks a child communication process U_sCOM, 90, which freezes its DAK generation and runs on behalf of the user until the end of the users' session communication. U_sCOM requests a secure connection establishment with the destination user (U_d) from the CA, and sends its frozen DAK_NRC for synchronization purposes. In the event of successful handshaking with the CA, 92 (FIG. 10), the source user receives an initial dynamic session key DSK, from the CA, 94.

[0084] The U_s-U_d data exchange session uses the shared symmetric DSK sent by the CA to both users. However, to increase the level of dynamic encryption to n crypto parallel streams, a set of n updated DSKs is derived randomly from the initial DSK, used as a seed, sent by CA. The source user message is stepped through n records, or a block, at a time, and the first n DSKs generated are used to encrypt the first block of n records. After that, n DSKs are regenerated for each consecutive block as a function of previously generated DSKs and previously encrypted data records. (See FIGS. 13a and 13b.)

[0085] This process is depicted in FIG. 8 after the source user has had successful handshaking, 94. The source user first generates randomly n different DSKs (DSK_i, where 1in), of the same size as the initial DSK, 96. Then, n data records of the same size as the size of DSK (Record_i, where 1in) are extracted from the input data, 98, and encrypted 100. Next, for every Record_i a new DSK_i is regenerated, 102, which is depicted in greater detail in FIGS. 13a and 13b below on a byte-by-byte basis. Finally, the n ciphered records are transmitted to U_d, 104. The encryption method described at 100 of FIG. 8 is illustrated in greater detail in FIG. 11.

[0086] Turning to FIG. 11, a diagram illustrates the encryption method described at 100 of FIG. 8. In FIG. 11: D_ij is the j^th byte of the i^th data record, DSK_ij is the j^th byte of the corresponding i^th DSK, C_i^tmpj is the j^th byte of the th temporary cipher record, C_ij is the j^th byte of the produced i^th cipher record, and PT_ij is the j^th byte of the i^th permutation table, 1in and 1jm, for n records, each of which is of m bytes.

[0087] The data record D_i and the DSK_i are used to produce C_i^tmp, which is permuted based on PT_i to produce cipher record C_i, 1in, producing n cipher records. The encryption method uses a simple XOR logic operation as an encryption function between a DSK_i and its corresponding data record D_i. The data byte D_ij, 146, is XORed with the corresponding key byte DSK_ij, 148, resulting in a temporary cipher byte C_i^tmpj, 150. To shuffle the C_i^tmp bytes around, a permutation table PT_i, 152 is generated as a copy of the corresponding DSK_i. The PT_i bytes are scanned from index 1 to index m, using each entry's index and its associated entry value, as indices of two bytes in the temporary cipher record to be swapped. The permutation is performed as follows: permute C_i^tmpj with C_i^tmpPT_ij, for 1jm, which results in the final cipher record C_i, 156. After n data records are encrypted with n DSKs as aforementioned, the final cipher block, made up of n cipher records (C_i, where 1in), is available for transmission, and a set of n new DSKs is regenerated for encryption of the next data block, as discussed below with reference to FIGS. 13a and 13b.

[0088] Returning to FIG. 8, the final cipher block is transmitted to U_d, 104 after having generated the n updated DSKs, 102, for encrypting the next data block of n records. The method of data encryption and transmission is repeated, 106, until the entire data volume has been encrypted and transmitted.

[0089] The decryption method is depicted in FIG. 9 at the destination side, U_d. Initially, U_d receives a communication request from the CA, 108. Then its DAK daemon forks a child communication process U_dCOM in order to freeze the generation of its version of DAK and sends the DAK_NRC to the CA for synchronization purposes, 110. U_dCOM will run on behalf of the user until the end of the users' session communication. In the event of successful handshaking with the CA, 112 (FIG. 10), the destination user receives an initial dynamic session key DSK, from the CA, 114.

[0090] After successful handshaking, U_d uses the initial DSK to randomly generate n DSKs (DSK^t_i, 1in) of the same size as the initial DSK (used as a seed), 116. Beginning with the same initial DSK at both U_s and U_d sites, U_d randomly derives the same set of n regenerated DSKs as the source user U_s, and parallels the encryption method described at the source user side in reverse, to decrypt the transmitted data. U_d receives a block of n cipher records (Cipher_i, 1in) 118. All cipher records are decrypted using each record's corresponding DSK, 120. Concatenation of the decrypted records provides the original message data to the destination user, 122. U_d then regenerates new DSKs from the recovered block of data records and the current DSKs to be used for decryption of the next block of cipher records, 124, as illustrated in FIGS. 13a and 13b. The process is repeated, 126, until all the transmitted cipher data has been decrypted.

[0091] Referring to FIG. 12, a diagram illustrates the decryption method at 120 of FIG. 9 in greater detail. To shuffle the m bytes of each of the received n cipher records, 158, back to the correct order, a permutation table PT_i is generated for each cipher record C_i, 160. The permutation table is generated, as at the source side, as a copy of the corresponding DSK. Permutation is performed on the received cipher record bytes using each of the table entry's index and its associated entry value, as indices of the two bytes in the cipher record to be swapped, but in bottom-up order, 162: permute C_ij with C_iPT_ij, for j from m down to 1. The result is a temporary cipher record C_i^tmp, 164. The decryption method continues by performing the XOR operation on C_i^tmpj, 164 and the corresponding DSK_ij, for 1jm, 166. This results in the original data record D_i, for 1in, 168. The process continues as shown in FIG. 9, at 122.

[0092] As discussed in FIGS. 8 and 9, upon request for a secure communication, both the source user and destination user must accomplish a successful handshaking process with the CA. FIG. 10 illustrates a user handshaking method with the CA. Upon receiving a message from the CA, 130, the user responds accordingly, 132. If an abort message is received, then the user terminates the communication child process in charge of the connection, 144. If authentication is requested, 138, the user mutually authenticates with the CA as described in FIG. 6(b). If a synchronize message is received including a number (x), the user performs x regenerations of its DAK, in order to be synchronized with the CA, i.e., aligning to the same DAK 134, as described in FIG. 5. An acknowledgment message is then sent back to the CA, 136.

[0093] After either synchronization or authentication, the user waits for a session key message including an encrypted DSK, (E(DSK)). The encrypted DSK is decrypted, 140. Then, the communication child process returns the aligned DAK and the new DSK to its parent daemon in order to initialize a new state for the DAK regeneration state, 141, and the communication between the source and destination is securely established, 142. The process continues at 96 of FIGS. 8 and 116 of FIG. 9.

[0094] Attention is now turned to FIGS. 13a and 13b, where regeneration of n DSKs is shown. FIG. 13a is a diagram illustrating the dynamic session key (DSK) regeneration method of FIG. 8, 102, and FIG. 9, 124. FIG. 13b is a diagram illustrating n tracks or streams of parallel DSK regeneration for a plurality of data record blocks. Each block of data records includes those data records encrypted within a predefined time increment (T). FIG. 13b illustrates blocks of data records from the range of time increments of T1 through Tt1. A track is defined as those data records in alignment within their respective data blocks, hence a track consists of a vertical column of data records as depicted in FIG. 13b.

[0095] In FIG. 13a DSK^t_ij represents the j^th byte of the last i^th dynamic session key at time t, D^toldj represents the j^th byte of the last i^th data record at time t_old, and ExpK^t_ih represents the h^th byte of the i^th expanded key at time t, 1in and 1jm, 1h2 m. A data record D^told_i is selected randomly from the previously encrypted data records D_i¹, D_i², . . . , D_i^t1 in the i^th track. This is accomplished by randomly selecting the index t_old from the range of 1, t1 using a predetermined byte of the current key DSK^t_ij as the seed of random generation, 176. For example, the first byte of the current key, DSK^t_i1, may be used as the seed of random generation. However, when generating DSK²_i, the only available data record is D¹_i, therefore, t_old1. Once the index t_old is selected from the range 1, t1, the corresponding previously encrypted data record D^told_i 172 is XORed with the DSK as shown.

[0096] Next, an expanded key (ExpK^t_i), 174, of twice the size of DSK^t_i is generated. Each of DSK^t_i, D^told_i, and ExpK^t_i is divided into (m/2) regions, indexed from 1 to (m/2). Each region of DSK^t_i and D^told_i is of length 2 bytes, and each region of ExpK^t_i is 4 bytes. The indexes of the four consecutive bytes of any region r, in the ExpK^t_i, are indexed with the values: 4r-3, 4r-2, 4r-1, and 4r. The indexes of the two consecutive bytes of any region r, in the DSK^t_i and D^told_i, are indexed with the values: 2r-1 and 2r. Filling the four bytes of region r in the ExpK^t_i, from DSK^t_i and D^told_i is defined as follows:

[0097] ExpK_i^t4r-3DSK^t_i2r-1

[0098] ExpK^t_i4r-2DSK^t_i2r

[0099] ExpK^t_i4r-1(DSK^t_i2r-1) XOR (DSK^t_i2r)

[0100] ExpK^t_i4r(DSK^t_i2r-1) XOR (DSK^t_i2r) XOR (D^told_i2r-1) XOR (D^told_i2r)

[0101] Then, a random selection of m bytes from the 2 m bytes of ExpK^t_i is performed, 178, using the first byte DSK^t_i1 as the random function seed, 176. Alternatively, any byte DSK^t_ij, can serve as the random function seed. This function generates a sequence of m random numbers, in the range between 1 and 2 m, each of which represents the index of the byte to be placed as the next byte of the newly generated DSK^t1_i, 180. This operation is performed at both the source and destination user nodes. The method can be performed periodically or aperiodically (with mutual consent of source and destination), depending on the implementation of the invention.

[0102] Referring to FIG. 13b, the initial DSK is first received from the CA in order to start a secure source-destination communication with n tracks of parallel encryption, each track having its own stream of DSK keys. The initial DSK_CA received from the CA is used to randomly generate the initial n DSKs, one for each of the n tracks of data records, (D₁¹, D₂¹, . . . , D_n¹), where D_i^t represents the data record of the i^th track at time t, and DSK_i^t represents the DSK of the i^th track at time t. (See also FIG. 8, 96 and FIG. 9, 116.) At time t1, the DSK_i^t1 is regenerated based on DSK_i^t and a randomly selected data record from the previously encrypted data record set D_i¹, D_i², . . . , D_i^t1 in the i^th track. As stated, the data record is randomly selected by randomly selecting the index t_old from the range of 1, t1 using a predetermined byte of the current key DSK^t_ij as the seed of random generation. In FIG. 13b, R represents a random selection operation while RF represents the regeneration function of FIG. 13a.

[0103] A randomly selected encrypted data record from the data record set D_i¹, D_i², . . . , D_i^t1 is used in the regeneration of DSK_i^t1 to render the latter obsolete in the event that a DSK is compromised by unauthorized intrusion. In the unlikely event that DSK_i^t1, in the i^th track, is obtained by an intruder, then the intruder can only obtain the data record D_i^t1 because the intruder has the cipher C_i^t1. However, the intruder cannot obtain more than that. The intruder must obtain the entire set of previously encrypted data records D_i¹, D_i², . . . ,D_i^t in order to generate DSK_i^t2, or any successor DSK. Hence, breaking one DSK will aid only in decrypting its corresponding data record and nothing more, past or future ciphered data.

[0104] It will be understood by those of skill in the art that ExpK can alternatively be comprised of any number of bytes, and the invention is not limited to any particular size for ExpK. It will also be understood that the byte positions chosen to be XORed together to fill the positions of the ExpK of FIGS. 13 and 14 could similarly be altered, and still remain within the inventive scope of the dynamic encryption and authentication method. It will also be understood that the size of the regions of the keys used to regenerate new DSKs and DAKs, need not be 2 or 4 bytes, but could be of any size within the size of the respective keys. Similarly, alternative logic operations could be substituted for the XOR operation.

[0105] Although the invention has been described in detail with reference to this preferred embodiment, other embodiments can achieve the same results. Variations and modifications of the present invention will be obvious to those skilled in the art and it is intended to cover in the appended claims all such modifications and equivalents. The entire disclosures of all references, applications, patents, and publications cited above are hereby incorporated by reference.