IBM's improvement arguments overcome Alice rejections, win random number generation claims at the PTAB

Takeaway: The PTAB reversed Alice rejections of pure-software random digital generation claims, crediting the appellants' improvement argument in the face of examiner criticism of that argument and even in the absence of a reply brief addressing that criticism.

Details:

Ex parte Sherwood
Appeal No. 2017-006552; Application No. 13/906,056; Tech. Center 2100
Decided: Sep. 28, 2017

The application on appeal, titled "Balancing Consumption of Random Data", described improved operation of an apparatus for generating random data in a computer system, and in particular, to the improved operation of obtaining entropy data for seeding the random data generation.  A representative claim on appeal read:

21.  A digital data apparatus for producing random digital data, comprising:
       at least one physical processor;
       a physical system memory;
       a plurality of entropy sources each generating respective random source digital data having a corresponding level of entropy, including a first entropy source generating random source digital data having a first level of entropy, and a second entropy source generating random source digital data having a second level of entropy lower than said first level of entropy, said second entropy source being independent of said first entropy source;
       a random number generator embodied as computer program code storable in said physical system memory and executable on said at least one physical processor, said random number generator receiving input from each said plurality of entropy sources, said random number generator generating a random digital data output by a deterministic algorithm using input from a selective one of said plurality of entropy sources as a seed for said deterministic algorithm;
       an entropy manager embodied as computer program code storable in said physical system memory and executable on said at least one physical processor, wherein said entropy manager automatically selects one entropy source among said plurality of entropy sources as input for the seed for said deterministic algorithm used by said random number generator, said entropy manager automatically selecting one entropy source among said plurality of entropy sources by determining a minimum level of entropy required by a consuming entity from among multiple possible minimum levels of entropy required, wherein the consuming entity consumes random digital data output by said random number generator to perform at least one data processing function, wherein the minimum level of entropy required by the consuming entity is a minimum level required as input for the seed for said random number generator to produce the random digital data output consumed by the consuming entity, said entropy manager further automatically selecting an entropy source from among said plurality of entropy sources having the lowest corresponding level of entropy which meets said minimum level of entropy required by the consuming entity.

(Emphasis added.)

The final rejection held that the claims were directed to the abstract idea of "performing mathematical steps drawn to generating random data, receiving inputs, selecting entropy sources, determining levels of entropy, consuming random data, monitoring/analyzing events, accessing rules, determining levels of entropy, preventing flipping of entropy selection states, and performing cryptographic operations."  The rejection stated that

[t]hese steps merely employ basic concepts drawn to manipulating information using mathematical and logical concepts which is similar to the basic concept of manipulating information using mathematical relationships found to be an abstract idea by the courts. The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the claims do not add a meaningful limitation to the abstract steps as they would be routinely used by those having ordinary skill in the art on a general purpose computer.

With regard to the second step of the Alice framework, the examiner found that the recitation of processor and memory in the claims did not amount to significantly more, and likewise,

the recitation of elements such as "entropy sources", "random number generator", "entropy manager", and "cryptographic apparatus" are also generic because there is no evidence in the specification that these elements must be specific hardware elements. For example, the claimed random number generator and entropy manager are embodied as code and the claimed cryptographic functionality is defined in the specification as software. A general purpose computer can be programmed to execute code and software.

The examiner also addressed the applicant's argument that the claims were directed to an improvement in computer technology:

[T]he invention as claimed is not drawn to an improved encryption engine for a computer system that improves efficiency. . . . [T]he applicant has failed to point to evidence that using "relatively lower entropy data for certain less essential purposes" improves efficiency of the claimed invention.  The applicant's arguments cannot take the place of evidence.

(Emphasis in original.)

The appeal brief is worth quoting at length:

[The] invention relates to "improved operation of an apparatus for generating random data in a computer system, and in particular, to the improved operation of obtaining entropy data for seeding the random data generation" . . . Although improved performance is the primary motivation, Appellant's technique . . . may, by using highest quality random data only when necessary, permit a higher quality of randomness to be used in those circumstances which require it.
A computer-implemented cryptographic engine or random data generator necessarily receives input in the form of data, i.e., strings of '1's and '0's, and performs a series of mathematical steps and takes branches based on determinations made using data, to produce a final result, which is again data, i.e., strings of '1's and '0's.  . . . [A]ny computer-implemented process can ultimately be reduced to receiving data as input, performing a series of mathematical steps and branches based on determinations made using the data, and producing a result in the form of data.  That is the very essence of a deterministic sequential state machine.  No computer, no matter how fast, sophisticated or advanced over other computers, can do anything beyond these basic steps.
If any machine or process which is limited to receiving data input, performing mathematical manipulations of the data, making decisions based on results, and outputting data, is deemed an "abstract idea", the Patent Office might as well give up patenting computer implemented inventions, for all such inventions, no matter how complex or sophisticated, ultimately can be broken down into simple mathematical steps and branches.  This is clearly not what the law is and not what was intended by the recent Supreme Court decision in Alice Corp. v. CLS Bank. . . . [T]t is true that in a very general sense Applicant's invention does [manipulate information using mathematical relationships].  As does each and every computer-implemented invention that has been patented by the Patent Office. . . . The Examiner's reasoning appears to be exactly the type of overly broad application of the "abstract idea" doctrine which was disapproved by the Federal Circuit in Enfish. . . .  Appellant's invention does not come close to pre-empting the field of "manipulating information using mathematical relationships".

(Emphasis in original.)

The examiner answered that:

[P]reemption is not the test for judging subject matter eligibility under the Alice analysis.  Rather, the test consists of (in summary) determining whether the claimed invention is drawn to a judicially recognized exception.  If so, then the claimed invention is further analyzed to determine whether there is additional subject matter recited that amounts to significantly more than the judicial exception.
The claims are drawn to apparatuses, methods, and computer program products which perform mathematical steps. These steps manipulate information using mathematical and logical concepts. Ideas such as this have been found by the courts to be abstract. See Parker v. Flook, 437 U.S. 584 (1978); Gottschalk v. Benson, 409 U.S. 63 (1972).  In Flook, the claimed invention was drawn to the abstract idea of gathering numerical information and manipulating it mathematically.  In the instant case, entropy information is gathered in the form of binary numbers and is manipulated mathematically to generate random data, which is also represented in binary.

The examiner called the appellants' improvement argument "flawed":

Appellant's only explanation that the claimed invention is an improvement is that computer system efficiency is improved "by using relatively lower entropy data for certain less essential purposes." . . . However, Appellant never defines what these "purposes" are.  More importantly, the entropy data is used for random number generation, this being the only "purpose" that can be inferred from Appellant's argument.  How this improves computer system efficiency is not explained in Appellant's argument. Appellant merely asserts that these purposes exist without citing any evidence as to what they are.

The appellants did not file a reply brief to address this rebuttal.  Nevertheless, the Board "agree[d] with Appellants that the Examiner has overgeneralized the claimed invention by summarizing it as the mere performance of mathematical steps, or as information gathered in the form of binary numbers that is manipulated mathematically to generate random data", and took it upon themselves to scour the specification for the inventive concept:

[The] invention is directed to generating random digital data for use by a consuming entity.  For certain purposes, e.g., strong encryption, random data having higher entropy is required; for other purposes, lower entropy (and thus quicker to gather) random data will suffice. . . . In the invention, an entropy manager determines the minimum level of entropy required by a consuming entity, and selects the random data (i.e., entropy) source that supplies random data having a requisite level of entropy for the consuming entity’s purposes.
We do agree generally with the Examiner’s conclusion that Appellants’ claimed invention is drawn to an abstract idea. The claims under appeal are drawn to method and apparatus for producing random digital data, including a random number generator for generating said random digital data by following a deterministic algorithm using input from one of a plurality of entropy sources as a seed.
Appellants persuade us, however, that the claims are, nonetheless, statutory under the second prong of the Alice analysis.  Representative claim 21 recites, inter alia, an entropy manager that “automatically selects one entropy source” from among a plurality, as the seed for the deterministic algorithm. The entropy manager makes this selection “by determining a minimum level of entropy required by a consuming entity . . . the minimum level of entropy required by the consuming entity is a minimum level required as input for the seed for said random number generator . . . said entropy manager further automatically selecting an entropy source from among said plurality of entropy sources having the lowest corresponding level of entropy which meets said minimum level of entropy required by the consuming entity.”
In support, Appellants’ Specification discloses that “[h]igh-entropy data is difficult for a computer to generate,” and even if one resorts to techniques such as monitoring network traffic, “it takes time in order to gather such random data.” . . . Appellants’ invention, thus, sets forth a system that automatically balances the tradeoff between (a) high entropy, hard to predict (and thus hard to attack) data that is scarce, time-consuming to produce, or both, and (b) lower entropy, less difficult to predict (but less secure) data that is easier to gather, less computationally intensive to produce, or both. See Spec. 9-10.
We conclude that the function of the entropy manager in the claimed invention results in claims drawn to significantly more than an abstract idea.  See Alice, 134 S. Ct. at 2355.  Like the animation method in McRO Inc. v. Bandai Namco Games America Inc., 837 F.3d 1299 (Fed. Cir. 2016), we conclude that the claims under appeal are limited to rules with specific characteristics.  McRO, 837 F.3d at 1314—15.  Like the self-referential logical table in Enfish, LLC v. Microsoft Corp., 822 F.3d 1327 (Fed. Cir. 2016), we conclude that the appealed claims focus on a specific means or method that improves the relevant technology.  See Enfish, 822 F.3d at 1335.  Here, Appellants’ invention focuses on a specific method of providing an entropy manager to make a decision concerning the appropriate entropy level to be employed, in order to supply random digital data of sufficient entropy while conserving computer processing power and/or time.

(Emphasis added.)  Thus, the Board reversed the subject-matter eligibility rejections, and the appellants were issued their patent (No. 9,934,000).


My two cents:

This post in the fifth in a series examining improvement arguments and their rapidly mounting significance in winning reversals of Alice rejections at the Board.  The first post in the series provided a set of practice tips.  This case is an example of an improvement argument winning at step two of the Alice framework, and illustrates the examiner demanding evidence in support of the improvement argument (see practice tip #7) but the appellants not needing it to win.

The abstract idea to which the examiner alleged the claims were directed ("performing mathematical steps drawn to generating random data, receiving inputs, selecting entropy sources, determining levels of entropy, consuming random data, monitoring/analyzing events, accessing rules, determining levels of entropy, preventing flipping of entropy selection states, and performing cryptographic operations") is a mouthful, but the Board nonetheless agreed that the claims were drawn to an abstract idea, albeit one of a different description: "producing random digital data, including a random number generator for generating said random digital data by following a deterministic algorithm using input from one of a plurality of entropy sources as a seed."

The Board gave no credit to the examiner's argument that additional evidence was required to show that using "relatively lower entropy data for certain less essential purposes" resulted in an efficiency improvement.  Instead, the Board found the description in the specification to be self-evident of the cause and effect reasoning necessary to support the improvement argument.

Only about six months elapsed between the docketing of the appeal and the handing down of the Board's decision.  This seems unusual, as most practitioners are more accustomed to appeal pendency on the order of years rather than months.  According to contemporary PTAB statistics, average appeal pendency for cases originating from technology center 2100 at the time of this decision was 13.2 months, down a spectacular 11.1 months from the same time the previous year.  Although appeal pendency has been dealt with very effectively and continues to fall (it's down to 13.0 months in 2100 at last count), this appeal was still delivered in less than half the average time.  One can only speculate whether this case was hastened by a PTO directive to build Alice guidance by expediting appeals deciding § 101 controversies, or because of the particular Board section to which the appeal was assigned and the workload thereof, or because of the identity of the real party in interest (IBM) and the volume of applications filed by that particular assignee.  Whatever the reason for the favoritism (if any), the appellants got a good deal with this appeal, only waiting about twice as long as they would have for another Office action.

We're not done yet with looking at improvement argument cases, so stay tuned for more.