Employing synthetic genes in genetic algorithms, information encoding and non-replicative encryption

Description of equivalent WO 0139004 A1

TITLE OF INVENTION: Employing Synthetic Genes in Genetic Algorithms, Information Encoding and Non-Replicative Encryption

INVENTOR: William L. Crowley, 1211 Happy Drive, Boise, Idaho 83706

RELATED APPLICATIONS: This application claims the benefit under 35 USC 1 19(e) of United States provisional application Serial No. 60/099,713, filed September 10, 1998 and entitled Information Encoding and Retrieval Through Synthetic Genes, which is herein incorporated by reference.

D E S C R I P T I O N

BACKGROUND OF THE INVENTION

Technical Field. This invention generally relates to genetic algorithms, information encoding and information encryption. Specifically, this invention relates to the structure of genetic algorithms and how that structure can be made to more accurately parallel the natural genetic paradigm; the structure of complex data set encoding, in particular, digitally volume rendered data sets; and absolute secure information encryption methods. The invention is particularly applicable to the range of problems wherein a more natural genetic-like implementation of the conventional genetic algorithm, the increased efficiency of information encoding in the form of synthetic genes and/or encryption of information utilizing synthetic genes may be found as a solution to these problems.

Background. Genetic algorithms have been used for decades for solving problems in many fields of research, development and implementation, most notably, the sciences and engineering. A central appeal of genetic algorithms has been the notion of novel solutions to otherwise intractable problems. For many of the same reasons that great fascination and often amazement exists for natural biological genetics, researchers have often found unique and unexpected solutions to various problem sets through the use of genetic algorithms.

In large part, genetic algorithms possess their power and effectiveness precisely because they mimic a process that, by the reckoning of current science, has been replaying itself successfully for billions of years.

Therefor, it would be desirable to provide a higher degree of transcription of natural genetic principles and apply this to the field of genetic algorithms. As technological advances create increased opportunities for data utilization and visualization, so too are there increased strains on the prior art of encoding and retrieving such information. In particular, the field of three dimensionally viewed objects and their manipulation in digital environments have become increasingly important in medicine, education and science. Virtual reality is fast becoming an accepted, and in some cases the only, means of achieving effective communication, education and simulation. Unfortunately, information current methods of information encoding are largely inefficient and cumbersome.

Therefor, it would be desirable to provide more efficient and compact method of information encoding.

Because of exponentially increasing digital traffic, both on and off the internet, the need for totally secure forms of information exchange has become paramount. With the advent of ever more powerful computers available to almost everyone, computing intensive algorithms for data encryption, as well as the decoding of such encryption, have now become practically feasible.

The rapid rise of technical tools, more common place expertise, and the sheer power of readily available computers have spawned a new class of users whose motives are not always honorable. "Code cracking" and computer "hacking" in general have become new avenues for entertainment, achievement and criminal activity. To date, very few, if any, absolutely secure information encyrption algorithms exist.

Genetic algorithms are generally parallel processes that develop populations of individual data objects, usually developing binary character strings into new populations of the same data type, using methods that mimic biological genetics, such as recombination, or crossover, and a proportional reproductive schema based on the notion of Darwin's survival of the fittest. Such algorithms start with some initial population of data objects, usually created in some pseudo random fashion. These data objects are then evaluated iteratively for fitness as it pertains to the problem at hand and genetic like operations are performed on various data objects within the population to evolve a new population.

John Holland of the University of Michigan developed the initial concept of genetic algorithms for fixed-length binary character strings in Adaptation in Artificial and Natural Systems, by Professor John H. Holland, 1975. Subsequent and significant works in genetic algorithms and genetic classifier systems may be referenced in Grefenstette, Proceedings of the Second International Conference on Genetic Algorithms, 1987; M. Srinivas, et al., "Genetic Algorithms: A Survey", Computer, vol. 27, No. 6, pages 17-26, June 1994; Goldberg, Genetic Algorithms, pages 10-20, 80-139, Addison Wesley 1989; W. B. Dress, "Darwinian Optimization of Synthetic Neural Systems", IEEE First International Conference on Neural Networks, San Diego, June 1987, vol. No. 3, pages769-775; Schaffer et al., An Adaptive Crossover Distribution Mechanism for Genetic Algorithms, Proceedings of the 2^nd International Conference on Genetic Algorithms, July 28- 31 , 1987, pages 36-40; and Melanie Mitchell, "An Introduction to Genetic Algorithms", pages 87-113, MIT Press 1996.

Several improvements have been made to Holland's basic premise over the ensuing years, but none has addressed the lack of parallelism between these genetic algorithms and their natural genetic paradigm, namely that in conventional genetic algorithms, sexual-like recombination or crossover, regardless of its geometry or relative sophistication, occurs on a population member or data object only in its final form, that is its phenotypical form. This is the biological equivalent, for example, of grafting the legs of a very fast runner onto the body of a person having great upper body strength in order to achieve some environmental fitness objective.

Indeed, all efforts toward improvements in genetic algorithms have been made on the same plane. That is to say, the prior art procedures of crossover and fitness selection in genetic algorithms, regardless of their variations, are performed on the same level of member development, without regard for the complexities surrounding the integrated behavior of the genomic regulatory systems underlying ontogeny, or in other words, the unfolding of events involved in an organism changing gradually from a simple to a more complex level.

In reality, biological sexual recombination occurs at the genotypical level, where the genotype is a group of organisms sharing a specific genetic constitution. The phenotype, the constitution of an organism as determined by the interaction of its genetic constitution and the environment, is the phase in the overall scheme of things where natural selection occurs. In natural genetics and ontology, ". . . the chromosomal and point mutations are continuously 'scrambling' the 'wiring diagram' and the 'logic' of the regulatory system". Stuart A. Kauffman, "The Origins of Order", pages 411-439, Oxford University Press, Inc. 1993. Furthermore, the developmental process is highly recursive and in fact far more parallel than the von Neumann architecture that slants the view of the way in which the "genetic code" is perceived to be executed.

In fact, the two procedures, biological crossover or sexual recombination and fitness selection, occur not only at very different levels but on an incredibly different scale.

"It has been estimated that the sperm that fertilized all the eggs from which the present human population of the world developed could be packed into a container smaller than an eraser on a pencil . . . the biologically inherited qualities of human beings - the similarities as well as the differences that distinguish one human from another and from all other living things - have their basis in a minute mass of sperm . . . ". Adrian M. Srb, "General Genetics", pages 2-26, 265-384, W. H. Freeman and Company 1965.

Volume rendering is a computer graphics technique whereby the object or phenomenon of interest is sampled or subdivided into many cubic building blocks, called voxels, or volume elements. A voxel is the three dimensional (hereafter 3-D) counterpart of the two dimensional (hereafter 2-D) pixel and is a measure of unit volume. Each voxel carries one or more values for some measured or calculated property of the volume and is typically represented by a unit cube. The 3-D voxel sets are assembled from multiple 2-D images, and are displayed by projecting these images into 2-D pixel space where they are stored in a frame buffer. Volumes rendered in this manner have been likened to a translucent suspension of particles in 3-D space.

In surface rendering, the volumetric data must first be converted into geometric primitives, by a process such as isosurfacing, isocontouring, surface extraction or border following. These primitives, such as polygon meshes or contours, are then rendered for display using conventional geometric rendering techniques. Both techniques have advantages and pitfalls. A major advantage of the volume rendering technique is that the 3-D volume can be displayed without any knowledge of the geometry of the data set and hence without intermediate conversion to a surface representation. This conversion step in surface rendering can sometimes be quite complex, especially if surfaces are not well defined, e.g. noisy 2-D images, and can require a lot of user intervention, such as manual contour tracing.

On the other hand, because the 3-D data set is reduced to a set of geometric primitives in surface rendering, this can result in a significant reduction in the amount of data to be stored, and can provide fast display and manipulation of the 3-D reconstructions produced by this method.

By contrast, since all of the image stack data is used for volume rendering, computers with lots of memory and processing power are required to handle volumes rendered in this manner. Because the entire data set is preserved in volume rendering, any part, including internal structures and details, which may be lost when reducing to geometric structures with surface rendering, may be viewed.

Since most applications of interest wish to preserve the internal structures and details resulting from volume rendering, but would desire to perform more efficiently, particularly with massive amounts of data, as in the case of surface rendering, it would be highly desirable to provide a technique that achieved the advantages of both methods, that is, the speed of surface rendering with the detail of volume rendering.

The prior art of data encryption is at least two thousand years old and the entire field of cryptology has evolved to a very high degree of specialization. The most commonly used methods employ a "key" or system of keys. A key may be a binary equivalent of a phrase, e.g.- "the quick brown fox". This binary number is in turn used to encrypt an information data set. As a general rule, the longer the binary number, the more difficult it is to break into an encrypted message. Thus, so-called "strong" encryption algorithms might use a 128 bit method in contrast to a more standard 40 bit method. One form of early key-encryption algorithms was the "symmetric" key approach. Symmetric key encryption methods require that both the sender and receiver of encrypted messages have access to the same key and that same key is used to both encrypt and decrypt, thus, "symmetric". Such systems assume that those communicating have an alternative means of secure communication, otherwise there could be no means of agreement on what key(s) to use.

In 1976, Whitfield Diffie and Martin Hellman at Stanford University, proposed the "public key" encryption system. The system was soon named "RSA". A user's RSA software generates a pair of keys. Each is a large integer, sometimes more than 500 digits. The two keys are mathematically related. Either key can be input into certain software to encrypt a message, and the other key can be input later to decrypt the same message. That is, encrypt with one key and decrypt with another. In practice, however, RSA encryption is used more as a security "envelope". Thus, what is transmitted is a message encrypted with a symmetric key and a copy of the key used, wrapped in the RSA envelope. At the receiving end, the decryption software first unwraps the RSA envelope, obtains the symmetric key and uses that key to decode the message.

In any event the same encryption will result in each instance that the same key is used on the same information data set. That is, the process can be replicated.

Regardless of the type of key system used, symmetric, public etc., the replication is ultimately the vulnerable aspect of the prior art of information encryption.

Therefor, it would be desirable to provide an improved method for absolutely secure information encryption that can not be replicated.

SUMMARY OF THE INVENTION

The invention uses biological genetics as a metaphor to achieve a greater parallel with the natural genetic paradigm by incorporating synthetic genes into the genetic algorithmic process. Employing synthetic genes within the structure of computer readable memory provides increased flexibility and power to genetic algorithms, allows for efficient and compact information encoding, provides secure information encryption, and provides a problem solving tool which can be applied to a large number of very diverse problems.

Another embodiment of the invention provides a method and system for nonlinear encoding of ordered information data sets employing synthetic genes. Yet another embodiment of the invention provides a method and system for highly nonlinear information encryption using synthetic genes which is completely non-replicative.

Additional advantages and novel features of the invention will be set forth in part in the description that follows, 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.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is an overview of the distinction between the genotypical and phenotypical planes on which the disclosed invention operates in executing genetic algorithms;

FIG. 2 is a flow diagram of a generalized method for implementing the invention in genetic algorithms;

FIG. 3 is a pseudo code listing of a recursive subroutine used in an example problem employing synthetic genes in genetic algorithms;

FIG. 4 is a flow diagram of synthetic chromosome expansion and fitness evaluation of the example "Magic Square" problem to be solved by one embodiment of the invention;

FIG. 5 is a flow diagram of the generalized procedure used in encoding volumetric information using synthetic genes;

FIG. 6 is a flow diagram of the example "Cubes within a Cube" encoding problem using synthetic genes according to one embodiment of the invention;

FIG. 7a is a table of detailed macro parameters used to produce the outer cube in the example encoding problem;

FIG. 7b is a table of detailed macro instructions used to display the four inner cubes in the example encoding problem;

FIG. 8 is a pseudo code listing for a recursive subroutine used in the example "Cubes within a Cube" problem according to one embodiment of the invention;

FIG. 9 is a table and description of encoded synthetic genes for the "Cubes within a Cube" problem; FIG. 10 is an overview of information encryption using synthetic genes according to one embodiment of the invention;

FIG. 11 is a flow diagram of an example of a message encryption using synthetic genes according to one embodiment of the invention;

FIG. 12 is a pseudo listing of a recursive subroutine used in the message encryption example; and

FIG. 13 is a table showing two instances of encryption of the same message resulting in different sets of synthetic genes.

DETAILED DESCRIPTION OF THE INVENTION

Referring now to the figures, the use and implementation of synthetic genes and computer-readable memory structures will be more fully described. Additionally, procedures for the creation, storing and manipulation of synthetic chromosomes and their synthetic genes will be more fully described. Finally, the construction of phenotypical structures containing the encoded parameters or other aspects of physical problem sets as derived from the base information contained within the synthetic chromosomes and component synthetic genes will be more fully described.

While it is true that biological chromosomes and the genes which they comprise have evolved in a relatively 'unsophisticated' gene to gene relationship over billions of years, particularly when viewed from a linear mathematical perspective, it is nonetheless impressive, to say the least, that so complex an organ as the human eye, for example, has come into existence at all. Indeed, the sheer power of biological evolution stems from "genetic systems containing thousands of genes whose products turn one another on and off. It is the integrated dynamical behavior of this regulatory system which coordinates the expression of different genes in each cell type of the organism and underlies the orderly unfolding of ontogeny." Stuart A. Kauffman, "The Origins of Order", pages 441-442, Oxford University Press, Inc. 1993. Thus, from a simplified view, the force of biological evolution appears to be derived from the constant cycling between fitness evaluation at the organism's developed level and reproduction, or crossover, at its genetic level. FIG. 1 shows the two main levels on which one embodiment of the invention for genetic algorithms operates. The fitness selection process in the upper phenotypical level dictates which synthetic genes ultimately take place in the recombination, or crossover, and possible mutation processes in the lower genotypical level.

An overview of using synthetic genes and chromosomes within the genetic algorithm is shown schematically in the flow diagram of FIG. 2. In the first step, global assignments are made that will be used by the expansion subroutine, EXPAND, as required by a given problem's parameters. These may include divisor maxima for the modulo operators, the number of genes in each chromosome, the size of the final phenotypical entity, and the individual population member whose final characteristics will be evaluated in the objective function.

The next step consists of generating the entire population of synthetic chromosomes, made up of individual synthetic genes. This step may be, and usually is, done in a pseudo random fashion.

The succeeding steps are done iteratively until some objective is reached. They consist of: proportionately producing offspring, e.g. synthetic chromosomal crossover, according to rank until the entire population is replaced; introducing a small chanced mutation; expanding each member's set of genes into a fully formed phenotypical entity, or problem solution candidate; and ranking the results of the entire phenotypical population according to their match against some target or objective function.

In the example presented below, the objective function is a solution to the so-called "Magic Square" problem, in this case, a four by four cellular matrix having one of the integers 1 through 16 in each cell and arranged in such a way that the sums of the individual cells in each row, column and diagonal are equal.

FIG. 3 shows the pseudo code of a simple subroutine that can be used to expand a synthetic chromosome into a fully developed population member having sufficient structure to potentially solve the example problem. Since it is a recursive call, global arrays, "dx[]", a temporary vector to be assigned the current population member's synthetic chromosome's genes and "exp_rslt[]", a temporary array of expanded results, i. e. fully developed phenotypical population member, as well as the modulo operator's divisor, "md", are defined elsewhere.

In pseudo code form, the prototype of the call is EXPAND(t, x, d, n), where: t = the number of genes to expand; x = the last calculated datum (initially zero); d = gene array element counter (i. e. - chromosome locus) (initially zero); and n = size of the resulting population member (i.e. in this the linear length of the coded, fully developed population member).

Because in practice, most fully developed, or expanded, population members are significantly larger in length than their corresponding synthetic chromosomes, the gene array counter, i.e. the chromosome locus, is first checked for being out of range of the chromosome loci. For instance, if d < 0, then t is added to the existing value of d, i.e. d = d + t. If d > t, then t is subtracted from the existing value of d, i.e. d = d - 1.

Having established that the chromosome's locus is in the appropriate range, the next step is to add the value of the chromosome's current locus, i.e. gene array element "d" to the last calculated, i.e. "passed", value "x", resulting in "m", a temporarily defined local variable within the subroutine. The modulo operator then determines the remainder of "m" less multiples of its divisor "md". Since, in this particular example, gene values were pseudo randomly assigned values between 0 and 999, a fair amount of variability is assured. A 1 must be added to the result because the base number is 0 and our coding requirement of possible values is the range 1 - 16.

Finally, a check is made for termination of the recursive call if the 0^tn element has been reached, otherwise, another call is made (to itself) until such condition is met.

FIG. 4 shows a detail flow diagram of the chromosomal expansion and fitness evaluation for the resulting phenotype in the "Magic Square" problem.

Perhaps one of the more important aspects of synthetic genes is the ability to perform highly compressed and ordered encoding of an information data set. As discussed above in the Background section relating to volume rendered data, one of the advantages of surface rendered data is the speed of rendering because the cartesian coordinates are known, whereas volume rendered data are not geometrically specific and can store, using voxels, detailed information about the inside structure of an object, albeit often at great costs in terms of storage and computational overhead. The relative speed and storage advantage of the surface rendering approach is achieved because there is no information that is stored regarding the internal structure of an object in 3D space.

One embodiment of the invention offers a method and system to achieve the advantages of volume rendered, voxel like information stored in the spirit and efficiency of the surface rendering approach.

FIG. 5 is an overview of the synthetic gene encoding process as it may concern a simple volumetric shape, complete with internal details.

FIG. 6 shows a flow diagram of a process for encoding a simple set of data used to generate a 3D cube having four smaller cubes positioned equidistantly within it, of different colors, shades and face information.

FIG. 7a and FIG. 7b are tables showing the different levels of actual data used as macro specifications in the example below and which was implemented using "Geomview", a 3D visualization software environment developed by the University of Minnesota. FIG. 7a and FIG. 7b are an amplification of the first step in FIG. 6, "Define Target Information".

Each data set is considered a level for transformation within the overall process. It begins with the coordinate parameters for the outside cube and all subsequent levels of transformation are subordinate to those coordinates.

The second step in FIG. 6 is the generation of a pseudo random population of synthetic chromosomes. In this particular example, a population of 1 ,500,000 synthetic chromosomes were thus created, each having four synthetic genes with numeric values ranging between 1 - 999.

FIG. 8 is a pseudo code listing showing the third step in FIG. 6 and the first part of the iterated loop where the actual transformation takes place. It describes the expansion of an individual synthetic chromosome into a state model such that, when it is "stepped through" or executed, a trial matrix of encoded parameters will be generated within the range of values as passed to the recursive subroutine "GEOMGROW". Each such constructed, i.e. expanded, trial matrix is compared to the target data defined initially in the procedure and ranked in order of ascending differences.

The next part of the iteration performs the synthetic chromosomal crossover in a quantity proportionately scaled to the ranking of the respective, expanded, state models, wherein a slight chance of mutation on a synthetic gene is made possible within the code.

The iteration process continues until a perfect match is achieved for the current level of target data. The entire process is repeated until all levels of target data have so transformed.

Thus, in this example, each level of complexity of the overall volumetric quantity will have its own set of four synthetic genes, beginning with the coordinate genes for the outer cube, the definitional characteristics of the outer cube and then proceeding to the inner cubes, their definitional characteristics and so on.

FIG. 9 shows the actual set of synthetic genes generated by this method. It is interesting to compare FIG. 7a and FIG. 7b with FIG. 9 and note the difference in scale while still allowing a faithful transcription of the synthetic genes therein to the final volumetric form.

FIG. 10 is an overview of a process of encrypting data through the use of synthetic genes. An advantage of the invention as it relates to data encryption is the fact that it is extremely unlikely that any two sets of synthetic genes representing the exact same data set will be the same. That is, the process is non-replicative. In other words, the same data set encrypted in separate instances by the invention will almost never result in identical sets of synthetic genes. In fact, the chance of such an occurrence is roughly the repeatability rate of most present day pseudo random number generators, usually around three times 10 raised to the 38^tn power.

Another feature of the disclosed method allows a variety of avenues for utilizing encryption "keys", or other information extraneous to the information data set, in such a way as to be imbedded into the encoded synthetic gene structure. Such encryption keys may follow more standard usage of symmetric and/or public keys, as discussed in the Background section relating to data encryption above, or may be employed in more novel ways. Because of its highly non-linear encoding scheme, this method of encryption leaves open the possibility of more and varied forms of "key-like" information, including indexed keys, key-per- character and compound keys, than currently exist in the prior art.

FIG. 11 is a flow diagram of an example encryption process using synthetic genes. The example encryption further illustrates a method of embedding an encryption key within the synthetic genes thus generated from the information data set. In a general sense, the procedure is a fitness selection process wherein possible exact matches to a particular data subset are generated by a series of recursive calls to the subroutine CYPHER.

The phrase "Fall not from grace" is to be encrypted using a single encryption key 'secret'. In this example, the phrase is decomposed into data subsets of three characters, ("Fa), (II ), (not), and so on. The encryption key 'secret' is converted into a numerical quantity and is an argument in the subroutine CYPHER, described below, for each initial call to a particular data subset.

FIG. 12 is a pseudo code listing describing the subroutine CYPHER used in the example encryption. The global gene array "dxQ" is defined ahead of the procedure CYPHER and contains the trial genes used in the transformation process. "ptype[]" is also defined globally and will contain the resulting transformed data having values of 0 through 255. "cnt" is a globally defined counter for "ptype[]" and terminating quantity for CYPHER. The purpose of CYPHER is to transform a series of numeric synthetic genes and any encryption key(s) into a series of numeric quantities falling within the ASCII range of 0 through 255, which represents the range of printable characters on standard U.S. keyboards. Different ranges can be employed by simply changing the modulo operator divisor.

Any encryption key information is initially passed to CYPHER as an argument in its call and becomes enmeshed in the resulting transformation as its original value becomes recursively modified each time a call to CYPHER is made to itself. That is, synthetic gene and encryption key values are operated on as a sum by modulo 255 to produce a potential solution to the numeric ASCII character equivalent of the target data. In this particular example, the encryption key, "secret", is converted into a numerical quantity representing the sum of the numeric equivalents of each of its ASCII characters. This numeric key value is initially passed to CYPHER for each data subset of the decomposed phrase representing the information data set. In subsequent calls of CYPHER to itself, previously modified values of "x", initially 0, as represented by "m", as well as modified values of "k", initially the encryption key, as represented by "nk", are both passed back in their modified form to be further modified, i.e. a modification of a previous modification, and so on.

FIG. 13 shows the encoded synthetic genes from two separate instances of encrypting the phrase "Fall not from grace" with the encryption key 'secret'. It shows that the invention produced two totally different sets of synthetic genes, each representing exactly the same information data set and encryption key. Each set of synthetic genes can be decrypted back into the exact same phrase, "Fall not from grace", provided that the encryption key, "secret", is used.

While there is shown and described certain embodiments of the invention, it is to be distinctly understood that this invention is not limited thereto but may be variously embodied to practice within the scope of the following claims.

I claim: