APPLE INC.

Patent Trials and Appeals Board

Jul 1, 2020

2019000919 (P.T.A.B. Jul. 1, 2020)

UNITED STATES PATENT AND TRADEMARK OFFICE
UNITED STATES DEPARTMENT OF COMMERCE
United States Patent and Trademark Office
Address: COMMISSIONER FOR PATENTS
P.O. Box 1450
Alexandria, Virginia 22313-1450
www.uspto.gov
APPLICATION NO. FILING DATE FIRST NAMED INVENTOR ATTORNEY DOCKET NO. CONFIRMATION NO.
13/743,989 01/17/2013 Gregory T. Lydon 18962-0695001/
P14669US1
9683
147271 7590 07/01/2020
APPLE/JWMH
7501 Village Square Drive
Unit 206
Castle Pines, CO 80108
EXAMINER
BEJCEK II, ROBERT H
ART UNIT PAPER NUMBER
2123
NOTIFICATION DATE DELIVERY MODE
07/01/2020 ELECTRONIC
Please find below and/or attached an Office communication concerning this application or proceeding.
The time period for reply, if any, is set in the attached communication.
Notice of the Office communication was sent electronically on above-indicated "Notification Date" to the
following e-mail address(es):
Apple@jwmhlaw.com
eofficeaction@appcoll.com
howard.hamilton@jwmhlaw.com
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
BEFORE THE PATENT TRIAL AND APPEAL BOARD
Ex parte GREGORY T. LYDON and
SYLVAIN RENÉ YVES LOUBOUTIN
Appeal 2019-000919
Application 13/743,989
Technology Center 2100
Before ALLEN R. MacDONALD, DEBRA K. STEPHENS, and
JAMES B. ARPIN, Administrative Patent Judges.
STEPHENS, Administrative Patent Judge.
DECISION ON APPEAL
STATEMENT OF THE CASE
Pursuant to 35 U.S.C. § 134(a), Appellant1 appeals from the
Examiner’s decision to reject claims 1–11, 21–27 and 31–37 (see Final Act.
2–12). We have jurisdiction under 35 U.S.C. § 6(b).
We AFFIRM.

1 We use the word Appellant to refer to “applicant” as defined in 37 C.F.R.
§ 1.42. Appellant identifies the real party in interest as Apple Inc. (Appeal
Br. 1).
Appeal 2019-000919
Application 13/743,989

2
CLAIMED SUBJECT MATTER
The claims are directed to generating notifications upon detecting
unusual user behavior (Spec. ¶ 1). Claim 1, reproduced below, is illustrative
of the claimed subject matter:
1. A method for determining behavior associated with a user
device, comprising:
receiving first sensor data, the first sensor data
including readings from one or more sensors configured to
measure motion of the user device and to measure an
environment in which the motion is performed;
creating behavior clusters from the first sensor data,
each behavior cluster being associated with a respective
behavior type, a respective set of one or more
representative readings, and a respective magnitude
threshold, creating the behavior clusters including:
determining that a first set of sensor readings
of the first sensor data correspond to a first behavior
type at least in part by determining that differences
between the first set of sensor readings and sensor
readings of other behavior types exceed a quality
threshold;
responsive to determining that the first set of
sensor readings of the first sensor data correspond
to the first behavior type, creating a first behavior
cluster from the first set of sensor readings;
determining a first set of one or more
representative readings based on magnitudes of the
first set of sensor readings;
determining a first magnitude threshold
based on variances in the first set of sensor readings;
and
associating the first set of one or more
representative readings and the first magnitude
threshold with the created first behavior cluster;
comparing second sensor data with the behavior
clusters, the second sensor data including readings from
Appeal 2019-000919
Application 13/743,989

3
the one or more sensors received after the first sensor data
is received;
determining that a behavior type corresponding to
the second sensor data is different from each behavior type
of the behavior clusters upon determining, based on results
of the comparison, that the second sensor data is out of
range of each magnitude threshold corresponding to each
respective cluster from each respective set of one or more
representative readings; and
causing the user device to perform a security action
in response to determining that the behavior type
corresponding to the second sensor data is different.
REFERENCES
The Examiner relies on the following references:
Name Reference Date
Young US 2003/0154072 A1 Aug. 14, 2003
Kashi US 2009/0049544 A1 Feb. 19, 2009
Oppenheimer US 2014/0089243 A1 Mar. 27, 2014
Akella US 9,092,802 B1 July 28, 2015
REJECTIONS
Claims 1–11, 21–27, 31–35, and 37 are rejected under pre–AIA 35
U.S.C. § 103(a) as unpatentable over the combined teachings of
Oppenheimer, Kashi, and Akella (Final Act. 2–10).
Claim 33 is rejected under pre–AIA 35 U.S.C. § 103(a) as
unpatentable over the combined teachings of Oppenheimer, Kashi, Akella,
and Hsu (id. at 10–11).
Claim 36 is rejected under pre–AIA 35 U.S.C. § 103(a) as
unpatentable over the combined teachings of Oppenheimer, Kashi, Akella,
and Young (id. at 11–12).

Appeal 2019-000919
Application 13/743,989

4
We have only considered those arguments that Appellant raised in the
Briefs. Arguments Appellant could have made but chose not to make in the
Briefs have not been considered and are deemed waived (see 37 C.F.R.
§ 41.37(c)(1)(iv)).
OPINION
35 U.S.C. § 103(a): Claims 1–11, 21–27, 31–35, and 37
Claims 1, 8, and 21
Appellant contends the method as recited in claim 1, is not obvious
over the combined teachings of Oppenheimer, Kashi, and Akella (Appeal
Br. 11, 22). In particular, Appellant contends neither Oppenheimer nor
Akella discloses “creating behavior clusters” in the manner recited in
claim 1. Appellant argues, “Oppenheimer generally discloses the use of
‘clusters’”; however, Oppenheimer fails to disclose “how these ‘clusters’ are
created” (id. at 11). Rather, according to Appellant, “as Oppenheimer’s
‘clusters’ appear to be simply pre-defined by the programmer of the system,
Oppenheimer fails to teach that behavior clusters are dynamically created for
a specific user device based on motion and environment readings collected
from that same user device” (id. (emphasis in original); Reply Br. 2).
Appellant further argues, “Akella does not cure” Oppenheimer’s defects;
rather “Akella simply discloses the use of ‘clustering’ in a highly
generalized manner” (Appeal Br. 12).
We do not find Appellant’s argument persuasive. The Examiner finds
Oppenheimer (Final Act. 2–3 (citing Oppenheimer ¶¶ 401, 1375–1378,
1392, 1412–1413, 1505, Fig. 5B, Fig. 6B,)) in combination with Akella (id.
at 5 (citing 7:9, 7:59, 8:50–54)) teaches the disputed limitations.
Specifically, Akella discloses “the compressed relevance model may be
Appeal 2019-000919
Application 13/743,989

5
generated as a sequence of clustering of usage time series, possibly a usage-
based predictor, both together based on categorizing users, types, and
devices” (Akella, 7:57–61). Akella further discloses:
The User Behavior Clustering module may receive input from
and provide output to the Big Data Store and the Macro/Micro
Profiles (Graph) module. The User Behavior Clustering module
may analyze groupings of data or potentially group data
pertaining to user behavior as described herein.
(id. at 8:50–54). Therefore, Akella teaches creating behavior clusters.
Appellant argues Akella does not disclose the recited steps for
creating behavior clusters (Appeal Br. 13). However, the Examiner relies on
Oppenheimer to teach the recited steps of “determining that a first set of
sensor readings. . .”; “determining a first set of one or more representative
readings”; and “determining a first magnitude threshold” (Final Act. 3) and
Akella to teach “responsive to determining that the first set of sensor
readings . . .” and “associating the first set of one or more representative
readings” (id. at 5). As noted by the Examiner,
[O]ne cannot show nonobviousness by attacking references
individually where the rejections are based on combinations of
references. … it is the combination of references which render
the claim obvious to a person of ordinary skill in the art.
Specifically, Oppenheimer discloses the use of clusters to
determine normal and abnormal use of the device. While
Oppenheimer uses pre-created clusters for the states of the
device, it is Akella which discloses the creation of new clusters.
(Ans. 3–4). Here, Appellant is arguing the references individually while the
Examiner is relying on the combination of Oppenheimer and Akella to teach
the disputed limitations.
Appellant does not provide specific arguments against the Examiner’s
findings but generally states “Oppenheimer cannot possibly teach ‘creating
Appeal 2019-000919
Application 13/743,989

6
behavior clusters’ by performing each of the specific steps recited in claim 1
(e.g., creating behavior clusters using a specific sequence of steps, upon
satisfaction of certain specific conditions” (Appeal Br. 11). The Examiner,
however, sets forth with specificity where each limitation is taught (Final
Act. 2–6; see also Ans. 3–4 (further explaining why an ordinarily skilled
artisan would have found the limitations obvious over the combined
teachings of Oppenheimer and Akella)).
Appellant additionally contends “Oppenheimer fails to teach that
behavior clusters are dynamically created for a specific user device based on
motion and environment readings collected from that same user device”
(Appeal Br. 11). Again, Appellant is arguing the references individually.
The Examiner relies on the combination of references to teach “creating the
behavior clusters” while Appellant is arguing Oppenheimer and Akella
individually. Accordingly, Appellant does not persuade us the combination
of Oppenheimer and Akella fails to teach or suggest “creating the behavior
clusters,” as recited in claim 1.
Appellant challenges the rejection of claims 8 and 21 on the same
basis as claim 1 (Appeal Br. 14–16); therefore, for the reasons set forth with
respect to claim 1, Appellant does not persuade us the combination of
Oppenheimer and Akella fails to teach or suggest “creating the behavior
clusters,” as recited in claims 8 and 21. Accordingly, we sustain the
rejection of claims 1, 8, and 21 under 35 U.S.C. § 103(a) for obviousness
over the combined teachings of Oppenheimer, Kashi, and Akella.
We further note, although not relied upon in affirming this rejection,
Oppenheimer also discloses creating behavior clusters. For example,
Appeal 2019-000919
Application 13/743,989

7
Oppenheimer describes a BIRD [(Portable Item Reporting Device2)]
Training (Self Configuration) operation, in which “Realtime or Time-
Proximate Environmental Conditions/Usage Conditions/Usage Data” is
combined with a determination of “typical item usage based on sensor
readings during BIRD training” (see Oppenheimer, Fig. 1D, item 700, item
1000). The system of Oppenheimer can determine behavior clusters either
by a BIRD configuration pre-defined by a user or system administrator, or
alternatively self-configured (id. ¶¶ 388–389). Specifically, Oppenheimer
states:
In an alternative embodiment, the ExD criteria (170) may be
determined in whole or part by the BIRD (200) itself during one
or more training sessions or configuration time periods. During
the training periods, an authorized user (AU) (not shown) uses
the item (for example, her keys (100.K) in her purse) in ways
designed to train the BIRD (200) to distinguish normal item
usage from anomalous item usage (503. 2). In this case, on-board
BIRD Navigation (1000)–possibly augmented at points by BIRD
logic (500) on a configuration computer (335) ––uses the sensor
data (700) collected during the training period to determine
normal (503.3) vs. anomalous (503.2) item usage, that is, ExD
criteria (170).
(id. ¶ 389 (emphases added)). Thus, rather than simply being “pre-defined
by the programmer of the system,” as Appellant has argued, Oppenheimer
discloses the system may itself, through a user, create behavior clusters
based on sensor data.

2 The acronym was purposefully changed to BIRD from PIRD because
“‘BIRD’ sounds beautiful and the letter ‘B’ looks much like the letter ‘P.’
Further, in both Danish and Norwegian, ‘portable’ is ‘brerbare,’ as well as
being ‘biirbara’ in Swedish, so we find the letter ‘B’ for portable after all.
Also, birds are generally pretty smart when it comes to finding their way
home.” (Oppenheimer ¶¶ 38–39).
Appeal 2019-000919
Application 13/743,989

8

Claim 34
Appellant contends that the method as recited in claim 34, is not
obvious over the combined teachings of Oppenheimer, Kashi, and Akella
and, in particular, that the combination fails to teach or suggest “wherein the
first magnitude threshold is a standard deviation for the first set of sensors
readings,” as recited in claim 34 (Appeal Br. 16). More specifically,
Appellant argues, “‘a standard deviation for the first set of sensor readings’
(i.e., ‘the first magnitude threshold’) is used to determine when to perform a
security action on a user device” (id. (emphasis in original); Reply Br. 4).
However, according to Appellant, “Oppenheimer teaches that a ‘standard
deviation’ is used to determine when to discard data entirely, such that the
data is not used at all” (Appeal Br. 16).
We are not persuaded. As explained by the Examiner, Oppenheimer
“discloses the concept of using standard deviation to determine the value of
the data (i.e., [outlier] or reliable)” (Ans. 5; see Final Act. 9 (citing
Oppenheimer ¶¶ 2301, 2466)). Specifically, Oppenheimer discloses “[d]ata
values previously recorded for the selected sensor, spanning the BIRD
training period, are retrieved from the historical environmental data log via
the data storage and management module” (Oppenheimer ¶ 2299 (citations
omitted)). Oppenheimer further discloses “[o]utlier values may be
determined based on a number of criteria including …: the number of
standard deviations of variance from a mean or normal value for the data”
(id. ¶ 2301). Thus, Oppenheimer discloses “wherein the first magnitude
threshold is a standard deviation of the first set of sensor readings,” as
claimed.
Appeal 2019-000919
Application 13/743,989

9
Appellant also argues “[a]lthough Oppenheimer generally discloses
that ‘standard deviation’ can be used to determine ‘outliers,’ Oppenheimer
specifically teaches that outliers are ‘suppressed’ or ‘ignored’” (Appeal Br.
16; Reply Br. 4). We do not find Appellant’s argument persuasive.
Oppenheimer states, “[a] determination as to whether to keep or ignore
outlier values … may be made based on a number of parameters and
criteria” –– not “suppress” or “ignore” (Oppenheimer ¶ 2301 (emphasis
added)).
Lastly, Appellant contends Oppenheimer fails to disclose “using ‘a
standard deviation for [a] first set of sensors readings’ to determine when to
perform a security action on a user device, a substantially different purpose”
(Appeal Br. 16–17). The Examiner sets forth with specificity where
Oppenheimer teaches the step of “perform[ing] a security action,” as recited
in claim 1 (Final Act. 3 (citing Oppenheimer ¶¶ 572–574)). Appellant does
not address the Examiner’s findings instead, Appellant only challenges the
Examiner’s findings regarding determination of the first magnitude
threshold. Accordingly, Appellant does not persuade us Oppenheimer fails
to disclose “wherein the first magnitude threshold is a standard deviation for
the first set of sensors readings,” as recited in claim 34. Therefore, we
sustain the rejection of claim 34 under 35 U.S.C. § 103(a) for obviousness
over the combined teachings of Oppenheimer, Kashi, and Akella.
Claim 35
Appellant contends the method as recited in claim 35, is not obvious
over the combined teachings of Oppenheimer, Kashi, and Akella (Appeal
Br. 17). The issue presented by the arguments is whether the combination of
Oppenheimer, Kashi, and Akella teaches, suggests, or otherwise renders
Appeal 2019-000919
Application 13/743,989

10
obvious “determining that differences between the first set of sensor
readings and sensor readings of other behavior types exceed a quality
threshold comprising determining that the differences between the first set of
sensor readings and sensor readings of other behavior types are greater than
a threshold distance,” as recited in claim 35. More specifically, Appellant
argues, “Oppenheimer does not contemplate anything analogous to
‘threshold distances’ at all, much less ‘determining that the differences
between the first set of sensor readings and sensor readings of other behavior
types are greater than a threshold distance’” (Appeal Br. 18; Reply Br. 5).
We are not persuaded. Oppenheimer discloses, “determining that
differences between the first set of sensor readings and sensor readings of
other behavior types exceed a quality threshold,” as recited in claim 1 (Final
Act. 3 (citing Oppenheimer ¶ 1375, Figs. 5B, 6B)). In addition,
Oppenheimer describes that BIRD logic “provides criteria and/or methods to
compare usage data . . . against usage expectations” (Oppenheimer ¶ 1375;
Final Act. 9; Ans. 5; see also Oppenheimer ¶¶ 1381–1382, Fig. 5B
(describing a plurality of clusters which contain criteria to determine which
state the BIRD logic is currently in)). Moreover, Oppenheimer discloses
determining “just how ‘acceptably close’ the usage data must be to the usage
expectations” (id. ¶ 401; Final Act. 3).
In light of this disclosure, we agree with the Examiner that
Oppenheimer teaches or suggests comparing one set of sensor readings
against another to determine a state of the item or if it is “otherwise in an
anomalous state” (Ans. 5). We determine an ordinarily skilled artisan would
have found it obvious to set the exceeding of a quality threshold to be
“greater than a threshold distance,” as recited in claim 35, as this is
Appeal 2019-000919
Application 13/743,989

11
comparing two sensor readings. Indeed, the determination of how close the
usage data must be to the usage expectation, to be a value within a particular
behavior cluster, teaches or suggests the claimed “threshold distance.”
Appellant’s argument that “Oppenheimer does not teach how to
create new clusters” (Appeal Br. 18) is not persuasive as the Examiner relied
on Akella to teach this limitation (Final Act. 5; Ans. 5). As previously
discussed with respect to claim 1, the Examiner is relying on the
combination of Oppenheimer, Kashi, and Akella to teach limitations of
claim 35 while Appellant is arguing the references individually. Therefore,
we are not persuaded the combination of Oppenheimer, Kashi, and Akella
fails to teach or suggest the limitation as recited in claim 35. Accordingly,
we sustain the rejection of claim 35 under 35 U.S.C. § 103(a) for
obviousness over the combined teachings of Oppenheimer, Kashi, and
Akella.
Remaining Dependent Claims
Dependent claims 2–11, 21–27, 31–33, and 37 are not separately
argued (see Appeal Br.); therefore, these claims fall with their respective
independent claims. Accordingly, we sustain the rejection of claims 1–11,
21–27, 31–35, and 37 under 35 U.S.C. § 103(a) for obviousness over
Oppenheimer, Kashi, and Akella.
35 U.S.C. § 103(a): Claim 36
Appellant contends the method as recited in claim 36, is not obvious
over the combined teachings of Oppenheimer, Kashi, Akella, and Young
(Appeal Br. 18) and, in particular contends the combination of
Oppenheimer, Kashi, Akella, and Young fails to teach or suggest:
Appeal 2019-000919
Application 13/743,989

12
“determining the first set of one or more representative readings
for the first behavior cluster comprises” [(i)] determining . . . a
maximum distance between the sensor reading and another
sensor reading of the first behavior cluster; [and (ii)] identifying
a particular sensor reading of the first behavior cluster having a
lowest maximum distance,”
as recited in claim 36 (id. at 19). Appellant argues “Young simply refers to
a ‘K-means clustering algorithm’” but does not disclose “a maximum
distance between the sensor reading and another sensor reading of the first
behavior cluster," or "having a lowest maximum distance" (id.). Nor,
according to Appellant, “does Young disclose or suggest performing such
steps specifically to determine a ‘representative reading’ for a behavior
cluster (id.; Reply Br. 7). Appellant further argues, “the Examiner did not
address these particular steps at all, much less identify any particular
teaching in the cited references that allegedly render obvious the
performance of such steps” (Appeal Br. 20).
We are not persuaded by Appellant’s contention that Young’s
teaching of the K-means clustering algorithm, which the Examiner
explained, fails to teach the disputed limitation. Appellant contends,
“Young does not refer to distances at all, aside from generally assigning
points to the ‘nearest cluster center’ without regard for a ‘maximum
distance’ or a ‘lowest maximum distance’” (Reply Br. 7). However, the
Examiner finds Young teaches the argued limitations (Final Act. 12 (citing
Young ¶ 52)) and, more specifically, explains:
[b]oth of these common clustering elements are found in Young
in paragraph 52 which recites calculating the center of the cluster
(a sensor reading with the lowest maximum distance would be
zero, i.e. the center of the cluster) as well as finding the distance
a data point can be from the center and still be part of that cluster
(i.e. maximum distance between two sensor readings).
Appeal 2019-000919
Application 13/743,989

13
(Ans. 6). Thus, Young teaches calculating the center of a cluster by finding
the sensor with the lowest maximum distance. And by determining the
distance a data point can be from the center, yet still be part of the cluster,
Young teaches “determining, for each sensor reading of the first behavior
cluster, a maximum distance between the sensor reading and another sensor
reading of the first behavior cluster,” as recited in claim 36. Moreover,
Young teaches determining average distance of cluster points to their cluster
centroids and therefore, teaches determining the distance of each cluster
point from its cluster centroid (Young ¶ 52). Appellant provides an
“illustrative example” of various sensor readings (Reply Br. 7), but,
Appellant has not identified how Young’s K-means clustering algorithm,
which calculates distances from a centroid, fails to teach or suggest the
disputed limitation when taken in combination with the other relied-upon
references. Additionally, as set forth by the Examiner, the claim does not
specify which sensor is the “another sensor” (Ans. 6).
Appellant additionally argues Young does not teach “determin[ing] a
‘representative reading’ for a behavior cluster” (Reply Br. 7); however, the
Examiner relies on Oppenheimer, not Young, to teach the limitation of
“determining the first set of one or more representative readings for the first
behavior cluster” (Final Act. 3) and Young to teach specific steps (id. at 12).
Therefore, we are not persuaded the combination of Young and
Oppenheimer fails to teach determining “determining the particular sensor
reading as a representative reading for the first behavior cluster,” as recited
in claim 36.
Accordingly, Appellant does not persuade us the combination of
Oppenheimer, Kashi, Akella, and Young fails to teach or suggest the
Appeal 2019-000919
Application 13/743,989

14
limitations as recited in claim 36. Therefore, we sustain the rejection of
claim 36 under 35 U.S.C. § 103(a) for obviousness over the combined
teachings of Oppenheimer, Kashi, Akella, and Young.
CONCLUSION
The Examiner’s rejections are affirmed.
More specifically,
The rejection of claims 1–11, 21–27, 31–35, and 37 under pre–AIA 35
U.S.C. § 103(a) as unpatentable over Oppenheimer, Kashi, and Akella is
affirmed;
The rejection of claim 33 under pre–AIA 35 U.S.C. § 103(a) as
unpatentable over Oppenheimer, Kashi, Akella, and Hsu is affirmed; and
The rejection of claim 36 under pre–AIA 35 U.S.C. § 103(a) as
unpatentable over Oppenheimer, Kashi, Akella, and Young is affirmed.
DECISION SUMMARY
In summary:
Claims
Rejected
35 U.S.C. § Reference(s)/Basis Affirmed Reversed
1–11, 21–
27, 31–35,
37
103(a) Oppenheimer, Kashi,
Akella
1–11, 21–
27, 31–35,
37

33 103(a) Oppenheimer, Kashi,
Akella, Hsu
33
36 103(a) Oppenheimer, Kashi,
Akella, Young
36
Overall
Outcome:
1–11, 21–
27, 31–37

Appeal 2019-000919
Application 13/743,989

15
TIME PERIOD FOR RESPONSE
No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a) (see 37 C.F.R.
§ 1.136(a)(1)(iv)).
AFFIRMED