dissenting.
Because the Board of Patent Appeals and Interferences (“Board”), now the Patent Trial and Appeal Board, erred in relying on a facially incomplete reference and was not supported by substantial evidence in finding that the same reference was publicly available as of the critical date, I would reverse its decision. I cannot endorse allowing the Board to strip Enhanced Security Research, LLC (“ESR”) of its right to a validly issued patent on such a suspect record. I, thus, respectfully dissent.
The Board relied on an incomplete reference — Haystack Labs, Inc., NetStalkerO, Installation and User’s Guide, Version 1.0.2 (1996) (“NetStalker”) — a reference which was missing all the even pages in one of the two chapters to which the Board cited to support its finding of obviousness, was missing entire sections of other chapters, and bore indicia of being a draft document. The reference was obtained from an interested party — a paid expert for a party opposing ESR in litigation, the same party who initiated the reexam of United States Patent No. 6,119,236 (“'236 Patent”). That paid expert, Stephen Sma-ha, was the only person who apparently had access to the reference, could explain whether a complete reference existed, could explain why, if so, the reference was submitted in incomplete form, and could explain what was in the missing portions of the reference. While Smaha submitted a declaration in support of the reference, he neither claimed that a more complete reference existed — at any time — explained why the reference was submitted in its incomplete state, or explained what the missing portions discussed. The government asserts no positive theory allowing it to rely on such a reference, arguing simply that it was ESR’s burden to prove a negative — i.e., that the pages of the manual to which it has been denied access teach away from or undercut the teachings in the pages Smaha and the requestor selectively chose to provide to the Patent and Trademark Office (“PTO”). The government is mistaken, as is the majority. The PTO should have refused to rely on the NetStalker manual as a reference, and should have refused to instigate or maintain a reexamination on such grounds.
The Board compounded its error by finding that the NetStalker Manual was publicly accessible by the critical date, despite the omission of important details in the supporting declaration. The Smaha declaration was telling more for what it failed to state than for what little it actually did say with regard to accessibility. Given his undisputed bias, the Board and majority should demand precision with re*1360spect to such important facts, and not rely on what appeared to be half-truths. If the manual really was publicly accessible as of the critical date, it would not have been difficult for Smaha to actually say so, and to support his statements with verifiable facts.
I. Incompleteness
The Board and the examiner relied on the facially incomplete NetStalker reference to find the claims of the '236 Patent obvious. Specifically, the Board relied on pages from chapters 5 and 6 of NetStalker. Chapter 6, entitled “Configuring Misuse Detector,” appears to be complete. But chapter 5, entitled “Running NetStalker,” is woefully incomplete. The only pages present are 5-1, 5-3, 5-5, and 5-7. The Board relies specifically on page 5-5, despite the fact that the preceding and following pages are both missing. The very instructions on which the Board relied in chapter 5 are incomplete. While not cited by the Board, chapter 7, entitled “Managing and Analyzing Log Files,” contains only the first three pages of the chapter; despite that fact, the examiner found it meaningful to his analysis. The Board concluded that the portions of NetStalker that are present “serve to disclose that portion of NetStalker that presumably is relevant to the patentability of the '236 patent” and that ESR did “not indicate[ ] that the portions of NetStalker provided are in any way irrelevant to the patentability of the '975 [sic] patent.” Ex Parte Enhanced Sec. Research, L.L.C., No. 2012-008692, Reexamination No. 901010,849, 2012 WL 3801778, at *3 (B.P.A.I. Aug. 30, 2012) (“Enhanced Sec.”) (emphasis added).
Though the Board did not find or even consider whether the missing pages were unlikely to teach away from or further clarify the provided pages, the majority proceeds to make factual findings on those issues. The majority finds it implausible that the missing pages of NetStalker would contain evidence contrary to an obviousness finding and says it sees nothing in the submitted portions to clearly indicate that the missing pages would have been meaningful to the Board’s analysis. This is speculation on the part of the majority, however. Speculation cannot substitute for actual evidence that the missing pages are meaningless. Without those pages, neither this court nor the Board can determine whether the missing pages of NetStalker teach away from the claimed invention. Nor can we clarify whether the missing pages would reveal that NetStalker is actually less similar to the claimed invention than it might appear. And, we are unable to determine whether the Net-Stalker Manual was only an incomplete draft and, thus, not likely to be publicly accessible. Where a reference is proffered by an interested party with control over all information relating to that reference, it is not too much to ask that the proffer be complete in all material respects.
Though the majority disagrees, I believe the Board’s analysis was legally insufficient because it was based only on a consideration of the evidence supporting a finding of obviousness, and did not consider the possibility of evidence contrary to such a finding. In an obviousness analysis, a reference must be considered “in its entirety, i.e., as a whole, including portions that would lead away from the invention in suit.” Panduit Corp. v. Dennison Mfg. Co., 810 F.2d 1561, 1568 (Fed.Cir.1987). While the majority distinguishes Panduit on the ground that the contrary evidence ignored was clearly in the record in Pan-duit, I think that is a distinction without a meaningful difference. There is no doubt that what was missing from the reference here related to the operation of the Net-Stalker product — or at least that version of it — -and related specifically to the disclosures in chapter 5 which the Board found central to its obviousness analysis. *1361Whether the Board discounts evidence before it (what the majority says are the Panduit circumstances) or turns a blind eye to the existence of such evidence should not make a difference; in either instance, the Board’s analysis is flawed.
This does not mean that the PTO can never rely on a reference that is incomplete. Considering an incomplete reference may be consistent with the obligation to consider a reference “as a whole” when only an incomplete reference is currently available and reliable evidence about what is missing from the reference is provided. The incomplete reference, in those circumstances, may be deemed the entire existing or relevant reference. This case does not fall into that category; the supporting declaration provides no explanation for Net-Stalker’s incompleteness, and never even addresses that incompleteness. In fact, the Smaha declaration indicates that the version of NetStalker provided “is a true and correct copy,” (J.A. 9705), which raises a question as to whether the manual ever existed in final form. Similarly, considering an incomplete reference may be consistent with the requirement to consider a reference “as a whole” when omitted portions of a voluminous reference clearly are not relevant because they are not directed to the field of the invention. In such a case, the reference is “whole” at least in relevant part. Again, this case is not that one; NetStalker’s missing pages are in the very sections relevant to the field of the claimed invention. Thus, NetStalker cannot be considered by the Board “as a whole,” Panduit, 810 F.2d at 1568, because it is facially incomplete in the relevant portions and there is no explanation for that incompleteness.
The government cites In re NTP, Inc., 654 F.3d 1279, 1296 (Fed.Cir.2011) for the proposition that a patentee “ha[s] the burden to prove that [a] document [i]s not authentic.” Even assuming that a paten-tee by analogy has the burden to show the relevance of missing portions of a reference, NTP does not address relevance, and even if it did, ESR met that burden because the missing pages are from a chapter relevant to the field of the invention and contain portions of the very instructions on which the Board relied. Again, the pages on either side of the main page cited by the Board are not there. When a patentee does not have access to the missing pages, it can show little else, and this showing should be sufficient to render those pages relevant to the content of the prior art and to evidence potentially contrary to a conclusion of obviousness.
While I agree that section 2218 of the Manual of Patent Examining Procedure (8th ed. Rev.9, Aug. 2012) is not disposi-tive, I believe it is consistent with a prohibition against the Board and the examiner relying on a reference that is incomplete in relevant part with no explanation of why that is so. Section 2218 requires that “a copy of each patent or printed publication relied on or referred to in the request, be filed with the request.” Id. It does not indicate that a portion of a reference can be filed with a request. Section 2218 also provides that “[i]f any of the documents are not in the English language, an English language translation of all necessary and pertinent parts is also required.” Id. Thus, the entire non-English document must be provided, which means that a patentee can gain access to the entire document by having it translated, and even the required partial translation must be sufficiently complete to include the relevant parts. No language in section 2218 suggests that a reference that is incomplete in relevant part can be submitted.
Given the potential impact of a Board decision on reexamination, due process concerns arise when, as here, a complete version of a reference is unavailable to a *1362patentee, but the PTO relies on it with no explanation from the provider as to why it is incomplete. “[A] patent is a property right protected by the Due Process Clause.... ” Abbott Labs. v. Cordis Corp., 710 F.3d 1318, 1327 (Fed.Cir.2013). While due process considerations frequently focus on notice and hearing, cf. id. at 1328, this court has acknowledged that additional procedures may be mandated by due process when the PTO acts. See id. at 1327. Beyond notice and an opportunity to be heard, “what additional procedures are guaranteed by due process requires balancing the various interests at stake.” Id. at 1328 (citing Mathews v. Eldridge, 424 U.S. 319, 334-35, 96 S.Ct. 893, 47 L.Ed.2d 18 (1976)). While “excluding compulsory production of testimony in inter partes reexamination proceedings [did not] raise[] a serious constitutional problem” on the facts of Abbott, id. (internal quotation marks omitted), the same cannot be said in this case.
In determining what due process requires, the court is to consider three factors: “[fjirst, the private interest that will be affected by the official action; second, the risk of an erroneous deprivation of such interest through the procedures used, and the probable value, if any, of additional or substitute procedural safeguards; and finally, the Government’s interest, including the function involved and the fiscal and administrative burdens that the additional or substitute procedural requirement would entail.” Mathews, 424 U.S. at 335, 96 S.Ct. 893. The second factor examines “the fairness and reliability of the existing ... procedures.” Id. at 343, 96 S.Ct. 893.
First, the patentee has a significant interest in the retention of its rights in a validly issued patent. This is unlike the situation in Patlex Corp. v. Mossinghoff, 771 F.2d 480 (Fed.Cir.1985), in which this court concluded that a regulation barring a patentee “from communicating with the PTO during the three-month statutory period during which the PTO is required to decide whether any substantial new question of patentability is raised by a reexamination request” was not inconsistent with due process. Id. at 483, 486. In that case, the property interest was only “the temporary deprivation of full enjoyment of patent rights, for the period needed to correct an erroneous determination to reexamine [the] patents.” Id. at 485. Here, the pat-entee is permanently, not temporarily, deprived of its enjoyment of patent rights.
Second, there is risk of an erroneous deprivation of those rights when the provider of an incomplete document is the one asking that a reexamination be instituted and is involved in active litigation with the patent holder. This is especially so where the only one with access to both the reference and information about the reference is a paid representative of that party. In such circumstances, minimal additional safeguards clearly are warranted. Allowing the PTO to rely on a reference that is unavailable and incomplete without explanation threatens the reliability and fairness of the proceedings.
The facts of this case illustrate the risk. Smaha, whose declaration purports to support NetStalker’s use as a reference, was the Chief Executive Officer and chairman of the board of NetStalker’s authoring organization. Smaha stated that he was “engaged as an expert by Juniper Networks, Inc.... in connection with the litigation against [ESR],” although he further stated that he had “no interest, personal or otherwise, in the outcome of Juniper’s disputes with ESR.” J.A. 9704. A paid expert for a party adverse to the patentee is not un-qualifiedly disinterested; saying he lacks an interest does not change that fact. The provider of the NetStalker manual was in the best position to provide it in its entire*1363ty or explain the absence of the missing parts. ESR, on the other hand, had no formal procedural mechanisms to obtain the document or any further explanation from Smaha, and ESR’s efforts to obtain that information informally were rebuffed. When PTO procedures do not require the provider of a reference to provide a complete reference or at least provide — under penalty for falsification — a statement that the document is complete in all relevant parts, that there is an explanation for any missing portions of the document, or that a document is otherwise available to the pat-entee, the incentive to mislead with partial submissions is great. Basic fairness to patentees should demand more.
Again, this case is different from Patlex in which the “risk of examiner error due to lack of information” is related “only to the question ‘whether a substantial new question of patentability ... is raised’, 35 U.S.C. § 303, not the answer to the question.” 771 F.2d at 485. Here, the risk of error does relate to the “answer to the question” of patent validity. While in Pat-lex the PTO’s “expertise [wa]s a factor to be given weight in considering the risk of error at this stage," id. (emphasis added), the PTO’s expertise can only be applied to the information provided to the examiner. When that information is fundamentally incomplete at the resolution of the validity inquiry, the benefit afforded by a “disinterested expert!]” cannot cure the problem.
Third, the PTO need not adopt any new, complex evidentiary procedures to cure this problem. The most straightforward corrective action is for the examiner or the Board to refuse to rely on a reference like the one proffered here. There may be some inconvenience and additional cost to the requester who would need to resubmit the reexamination request, but there would be none to the PTO. Requiring a complete document, an explanation for incompleteness, or an indication of public accessibility will strongly incentivize providers of references to meet at least one of these requirements in the first instance. The government’s burden from this additional procedure would be minimal.
The relative dearth of other protective procedures in a reexamination reinforces the need to allow the patentee to challenge the examiner’s and the Board’s reliance on a reference that is unavailable and incomplete without explanation. A patentee may not seek discovery or resort to subpoenas to seek information in a reexamination. See Abbott, 710 F.3d at 1328. The absence of vehicles for discovery makes the ability of the patentee to challenge a reference on completeness grounds all the more critical; otherwise the patentee would be defenseless against one who chooses to provide only those portions of a reference which undercut the validity of a patent. An interested provider of a reference should not be able to use a reference as a sword, while failing to provide the portions of it that may shield the patentee.
The due process balance also requires inquiry into the extent to which “judicial-type procedures must be imposed upon administrative action to assure fairness.” Mathews, 424 U.S. at 348, 96 S.Ct. 893. Prohibiting the PTO from relying on non-probative evidence is hardly an elaborate judicial procedure and is certainly one that is necessary to assure the fairness of the proceedings. The Supreme Court has explained that “procedural due process rules are shaped by the risk of error inherent in the truthfinding process as applied to the generality of cases, not the rare exceptions.” Mathews, 424 U.S. at 344, 96 S.Ct. 893. This does not bar finding a due process violation in an individual case, however, because “[a] fundamentally fair adjudication ... is constitutionally required in all cases, and not just in the *1364large majority.” Cushman v. Shinseki, 576 F.3d 1290, 1299-1300 (Fed.Cir.2009).
Due process requires “a fair hearing on the merits” of a claim. Cushman, 576 F.3d at 1299. In Cushman, the initial determination of the veteran’s claim “was tainted by the presence of an improperly altered document,” and “[t]he source of the fundamental unfairness that tainted the initial evaluation of Mr. Cushman’s claim was never removed from any prior proceedings.” Id. “The presentation of improperly altered material evidence has been found to constitute a due process violation in analogous cases.” Id. at 1300. When the procedures applied by the PTO do not provide either this court or the patentee some means to determine whether, or some assurance that, the evidence on which the PTO relied was not improperly altered, those procedures cannot be consistent with due process.
The majority and the government cite nothing that allows the Board or the examiner to rely on an incomplete reference. I believe the majority errs in concluding that nothing prohibits the Board from relying on an incomplete reference; due process and concepts of fundamental fairness do. Accordingly, I would reverse the Board’s obviousness finding because it is based on an unreliable reference.1
II. Public Accessibility
The majority also errs in concluding that Smaha’s declaration was sufficient to establish that the NetStalker manual was accessible to the public before the critical date.
Whether a reference is publicly accessible is a question of fact that we review for substantial evidence. A reference is publicly available if it was disseminated or otherwise made available to the extent that persons interested and ordinarily skilled in the subject matter or art exercising reasonable diligence, can locate it.
NTP, 654 F.3d at 1296 (citation omitted) (internal quotation marks omitted). “The proponent of the publication bar must show that prior to the critical date the reference was sufficiently accessible, at least to the public interested in the art....” In re Hall, 781 F.2d 897, 899 (Fed.Cir.1986).
Smaha said, “This version of the Net-Stalker manual was available in May 1996. Members of the public showing an interest in buying or licensing the NetStalker product could have obtained a copy of the manual by contacting Haystack ... and requesting one.” J.A. 9705. Smaha also said, “NetStalker was advertised no later than 1995.” Id. The Board and the majority both conclude that members of the public would have known of NetStalker based on the advertisements as early as 1995 and would have received the manual upon request in May 1996. The first weakness in that conclusion is the supposed connection between the product advertised in 1995 and the May 1996 Manual. Smaha says “[tjhis version,” of the manual was “available” in May 1996. The front page of the manual indicates that it is “version 1.0.2.” Nowhere does Smaha say that the 1.0.2 version of NetStalker is what was advertised in 1995. Nor does Smaha ever say that the version described in the manual was ever advertised to the public. Pointedly, while Smaha says members of the public interested in the NetStalker product could have gotten the relevant manual upon request, there is no indication that the public had any information available to it which would have prompted any*1365one to make such a request for that particular manual. And, there is no evidence that version 1.0.2 of NetStalker was ever manufactured or offered for sale.
Thus, not only is there no evidence that the version of NetStalker discussed in the manual was ever advertised, but there is no evidence — from the Smaha declaration or otherwise — that any sales ever occurred prior to the critical date, that the sales that did occur were of the version of the product described in the reference, or that any of those sales were accompanied by the relevant manual. Smaha’s declaration was submitted by counsel. Had the gaps in his testimony been tillable — rather than conveniently omitted — it is likely those gaps would have been filled. The majority reads facts into the declaration that are simply not there; the absence of those facts is not harmless as the majority seems to believe, they are telling.
Perhaps most troubling is Smaha’s statement that the reference submitted is a “true and correct copy” of the manual which he said would have been made available to members of the public in May 1996, if requested. It is undisputed, however, that the submitted manual was incomplete. And it is undisputed that the submitted manual bore several indicia of a draft document: a cryptic date legend on its cover, question marks in the index, and the absence of the last ten pages of the final chapter. If an incomplete and unfinished manual is the “true and correct” version of the reference in existence in May 1996, any claim that the public would have been given access to it, or even would have known to request it, is even more suspect.
It is worth noting, moreover, that Sma-ha filed his own patent application — after the critical date of the '236 patent — to similar technology, but did not list his own manual as prior art. If we assume Smaha was not purposely misleading the PTO with that filing, the failure to cite the manual indicates that it was either an unfinished draft document or never available to the public.
For all these reasons, I believe the Board’s finding that the reference was publicly accessible before the critical date is not supported by substantial evidence. Smaha’s statements are sufficiently ambiguous to encompass both scenarios in which the NetStalker manual would have been publicly accessible and those in which it would not have been so. The Board cannot conclude that a reference was publicly accessible when no evidence provides sufficient specificity to support that conclusion. The majority should not endorse its having done so.
III. Obviousness
Because I believe the Board should never have reached the question of obviousness on this record, I do not analyze the majority’s obviousness analysis in detail. I take issue, however, with the fact that the majority bases its judgment on grounds that differ from those upon which the Board relied. This Court may not stray from the Board’s reasoning for purposes of supporting its judgment. See SEC v. Chenery Corp., 332 U.S. 194, 196, 67 S.Ct. 1575, 91 L.Ed. 1995 (1947) (“[A] reviewing court, in dealing with a determination or judgment which an administrative agency alone is authorized to make, must judge the propriety of such action solely by the grounds invoked by the agency.”); see also, In re Applied Materials, Inc., 692 F.3d 1289, 1294 (Fed.Cir.2012) (“The Board’s judgment must be reviewed on the grounds upon which the Board actually relied.”). Notably, the Board did not, as the majority states, find that, “in combination, these two pieces of prior art [Net-Stalker Manual and Liepins] disclosed all of the elements of the '236 patent.” Maj. at 1352.
*1366Instead, the Board found, and the government admits that, NetStalker and Lie-pins do not “specifically teach[ ] using the assigned strength or severity level as a basis for blocking communications.” Ap-pellee’s Br. 1. The Board filled this gap by relying on what it characterized as “ordinary creativity”:
NetStalker discloses not only a user defined severity level of a security breach but also triggering an alarm when a certain number of (security) events are recognized and blocking communications when the alarm is triggered. Based on the NetStalker reference, one of ordinary skill in the art would have known to characterize a security breach based on level of severity (i.e., user defined severity) and block communications based on when a condition has been achieved (for example, when a threshold number of security events have been encountered).
Given NetStalker’s disclosure of blocking communications when a threshold criteria is met indicating a security breach and given that one of ordinary skill in the art is a person of ordinary creativity, not an automaton, one of skill in the art would have understood the practice of blocking communications when a security breach is detected, the security breach being of sufficient severity as to exceed a threshold. NetStalker further discloses that a severity level is assigned to security breach events thus further indicating that blocking communications when a severity level of security breach is identified would have been obvious (as assigning severity levels to security breaches were known to those of ordinary skill in the art), or at least obvious to try, as a matter of ordinary creativity and common sense.
Enhanced Sec., at *8 (citations omitted) (internal quotation marks omitted).
As ESR argues, however, obviousness is not shown by the mere fact that each of the elements “was, independently, known in the prior art.” KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 418, 127 S.Ct. 1727, 167 L.Ed.2d 705 (2007). “Although common sense directs one to look with care at a patent application that claims as innovation the combination of two known devices according to their established functions, it can be important to identify a reason that would have prompted a person of ordinary skill in the relevant field to combine the elements in the way the claimed new invention does.” Id. (emphasis added). The Board identified no such reason. Specifically, the Board identifies no “design need or market pressure to solve a problem,” id. at 421, 127 S.Ct. 1727, such that the new combination using severity as the basis for blocking communication would have been “obvious to try,” id. And, the Board identifies no “problem” that one of ordinary skill was trying to solve at the time of the invention.
Apparently recognizing these gaps in the Board’s analysis, the majority accepts the government’s suggestion .that it fill them with an alternative analysis. The government argues that the undisclosed use of severity assessments for blocking purposes is actually disclosed in the Net-Stalker reference because “NetStalker itself includes the idea of tailoring the response to the severity of a threat, i.e., by only initiating action after a threshold number of events has occurred.” Appel-lee’s Br. 20. The Board expressly found this teaching missing in NetStalker, however. I agree.
While the majority relies on the statement in NetStalker that “you may want NetStalker to take an automatic action (a ‘response’),” a “[rjesponse” is simply another type of alarm that is listed in a “description of supplied alarms.” Net-*1367Stalker at 4-2, 4^4. It is described as “[p]rovid[ing] a general purpose response to activities taking place on the network [reserved for future use].” Id. at 4-4 (final brackets in original). The pages referenced by the Board also indicate that “[r]esponse” is “Reserved for future use.” Id. at 5-5, 6-16. While the “User Defined” alarms are also referred to as “responses,” the response “can be as simple as sending a beep to the system console or more complex such as logging the event in syslog.” Id. at 4-5, 5-5, 6-16. None of this suggests that the response is necessarily related to a shun alarm or that the response would be triggered by the severity-
Thus, not only does the majority violate the principles described in Chenery governing review of administrative agency determinations, but its independent obviousness analysis seems inconsistent with the very reference upon which the majority’s alternative analysis relies.
IV. Conclusion
For these reasons, I cannot join the majority in its conclusions on the reliability of the NetStalker reference, on public accessibility, or on the obviousness of the relevant claims of the '236 patent. I cannot join in depriving a patentee of its patent rights on these grounds. I respectfully dissent.
. While the majority emphasizes the Board's reliance on the Liepins article, the parties agree that Liepins provides only limited support for the Board’s obviousness determination. Without the NetStalker Manual, there would have been no obviousness finding.