CAFC in SOPHOS LIMITED v. IANCU. PTAB's understanding of process found to be backwards.

The outcome of SOPHOS LIMITED v. IANCU:


In this inter partes review, Finjan Holdings, Inc. challenged
various claims of Sophos Limited’s U.S. Patent
No. 8,776,218, which describes and claims computer
programs that perform runtime behavior-based detection
of malicious software. The Patent Trial and Appeal Board
of the Patent and Trademark Office (PTO) determined
that certain claims are unpatentable for obviousness.
Sophos appeals from the Board’s decision. With Finjan no
longer participating, the PTO’s Director has intervened.
Because the Board’s claim construction, as the Board
understood its scope, is incorrect, we vacate the Board’s
decision and remand for further proceedings.




This began as an inter partes review:


Finjan petitioned for an inter partes review of claims
1 through 20 under 35 U.S.C. §§ 311–19. The Board,
acting as the delegate of the PTO’s Director under 37
C.F.R. § 42.4(a), instituted a review of claims 1, 11, and
12 on multiple grounds, all under 35 U.S.C. § 103. Finjan
Holdings, Inc. v. Sophos Ltd., IPR2015-01405, Paper No.
9, at 27 (P.T.A.B. Dec. 15, 2015) (Institution Decision).

(...)

In its Final Written Decision, the Board construed the
claim phrase requiring that phenotypes “are ranked.”
Finjan Holdings, Inc. v. Sophos Ltd., IPR2015-01405,
2016 WL 7987957, at *4–6 (P.T.A.B. Nov. 30, 2016) (Final
Written Decision). The Board first construed the term
“ranked” to mean simply “ordered,” id. at *4, meaning
that the phenotypes must be ordered vis-à-vis each other.
The Board then concluded that, in the phrase “are
ranked,” ranking “may occur as part of the comparing
step such that phenotypes are not ‘already ranked’” before
the comparing step begins. Id. at *6 (emphasis added).
The “may” in that conclusion, however, ultimately
does not accurately reflect the Board’s understanding of
“are ranked.” As the key substantive basis for its claim
construction, the Board explained its understanding of
the claimed ranking in a way that actually precludes precomparison
ranking. Specifically, the Board said that
“[a]s a result of the comparison, the phenotypes ‘are
ranked’”; “[i]t is only in the context of the comparison that
the ranking occurs”; and “[t]he reason for the comparison
between the ‘malicious behavior indications’ and the
‘phenotypes’ is to identify the most problematic phenotype,
which is ‘ranked’ according to how similar it is to the
‘executing computer program.’” Id. at *5 (emphases added).
In short, the Board’s understanding is that the
invention ranks phenotypes based on their similarity to a
particular runtime object being examined.

(...)

But neither the Board’s explanation nor the Director’s
defense of that explanation in this court indicates how the
claim can reasonably be understood to call for that ranking
to be based on how similar phenotypes in a database
are to a particular runtime object being scrutinized. The
“broadest reasonable interpretation . . . is an interpretation
that corresponds with what and how the inventor
describes his invention in the specification.” In re Smith
Int’l, Inc., 871 F.3d 1375, 1382–83 (Fed. Cir. 2017); see
Microsoft Corp. v. Proxyconn, Inc., 789 F.3d 1292, 1298
(Fed. Cir. 2015) (prosecution history is relevant to this
inquiry as well), overruled in another respect by Aqua
Prods., Inc. v. Matal, 872 F.3d 1290 (Fed. Cir. 2017); In re
NTP, Inc., 654 F.3d 1279, 1288 (Fed. Cir. 2011) (“While
the Board must give the terms their broadest reasonable
construction, the construction cannot be divorced from the
specification and the record evidence.”); In re Suitco
Surface, Inc., 603 F.3d 1255, 1260 (Fed. Cir. 2010) (“The
broadest-construction rubric . . . does not give the PTO an
unfettered license to interpret claims to embrace anything
remotely related to the claimed invention. Rather, claims
should always be read in light of the specification and
teachings in the underlying patent.”). Here, the Board’s
central understanding is unreasonable in light of the
specification and, indeed, of the claim’s own statement of
its objective.
We have been pointed to nothing in the ’218 patent
that discloses ranking the predetermined phenotypes
according to their degree of similarity to a particular set
of malicious behaviors observed in a particular runtime
object. To the contrary, the patent speaks consistently of
seeking a “match” of a phenotype for the runtime object’s
set of malicious-behavior indications, never of examining
degrees of similarity. ’218 patent, Abstract; col. 17, lines
41–48, 64. In addition, the Board’s notion would call for
re-ranking the group of phenotypes with every new
runtime object that is evaluated. The patent says nothing
to that effect.
More fundamentally, the Board’s notion of ranking
the phenotypes by how similar each is to a particular
runtime object is detached from the essential function of
the invention. As claim 1 itself makes clear, the function
of the comparison of a runtime object to phenotypes is to
develop information about the threat presented by a
runtime object, whose risk of being any type of malware is
not known before the comparison.



As to PTAB's understanding:


The Board fundamentally misread the patent when it
said, in its crucial paragraph, that “[t]he reason for the
comparison . . . is to identify the most problematic phenotype.”
Final Written Decision, 2016 WL 7987957, at *5
(emphasis added). That statement gets the direction of
inference in the patented process backwards. The point of
the comparison is not to start with a known danger presented
by the runtime object and infer how problematic a
phenotype is by the degree of similarity to the runtime
object. The danger of the runtime object is not yet known.
The point of the comparison is to infer something about
precisely that danger, by using independent information
(through ranking) about how problematic a particular
matched phenotype is. Ranking of phenotypes independently
of similarity to the runtime object is an essential
component of that process.
In short, nothing in the claims, the specification, or
the prosecution history supports the Board’s understanding
that the predetermined phenotypes are ranked based
on their similarity to the observed malicious behavior
indications. We conclude that the Board’s understanding
is “divorced from the specification” and is “legally incorrect.”
Smith, 871 F.3d at 1382 (quoting Microsoft, 789
F.3d at 1298).