`#: 5603
`
`Joint Appendix
`Exhibit 36
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 2 of 38 PageID
`#: 5604
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`_____________________________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`_____________________________
`
`WIZ, INC.,
`Petitioner,
`
`v.
`
`ORCA SECURITY LTD,
`Patent Owner.
`
`_____________________________
`
`Case IPR2024-00864
`Patent No. 11,663,032
`_____________________________
`
`DECLARATION OF DR. ANGELOS STAVROU
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 3 of 38 PageID
`#: 5605
`
`149.
`
`It is therefore my opinion that the combination of Veselov and
`
`Hufsmith teaches all aspects of elements 1.4, 18.4, and 22.4.
`
`7.
`
`Elements 1.5, 18.5, and 22.5
`
`150. These elements recite “[determining/determine] whether the matching
`
`installed applications are used by the protected virtual cloud asset.” Hufsmith
`
`teaches these elements in several ways. As I previously explained, Hufsmith
`
`describes multiple stages of the analysis during which weights (i.e., priorities/threat
`
`levels) for each detected security issue are assigned or adjusted. Supra, §XI.B
`
`(Hufsmith summary). After the set of security issues are identified, several
`
`evaluators (e.g., a malware evaluator and a CVE evaluator) assign initial weights to
`
`the identified items. EX1078 (Hufsmith), ¶¶109, 113-28. A context evaluator
`
`further considers other factors that are used to adjust these weights. Id., ¶¶135-43.
`
`Below, I first explain how Hufsmith teaches elements 1.5/18.5/22.5 with respect to
`
`the context evaluator. I then explain how Hufsmith further teaches these elements
`
`as part of the calculation of the initial weights.
`
`151. First, Hufsmith teaches adjusting previously assigned threat levels
`
`based on whether the associated file/code is dormant or active (i.e., “whether the
`
`matching installed applications are used by the protected virtual cloud asset,” as
`
`recited). Hufsmith uses a “context evaluator” to adjust weights for vulnerabilities
`
`identified by other evaluators based on various context properties. Supra, §XI.B
`
`133
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 4 of 38 PageID
`#: 5606
`
`(Hufsmith summary); EX1078 (Hufsmith), ¶¶135-43, 145. For example, the
`
`“environments” context property includes information about the execution
`
`environment of the virtual image and is used “for adjusting weights associated with
`
`one or more scanner properties.” Id., ¶140. The execution environment records
`
`“indicate which portions of a container image are dormant and which are active in a
`
`given use case,” such as whether a “a subroutine is never invoked” or whether a
`
`“command module is never called” (i.e., which portions are used and which are not
`
`used). Id. In some cases, “[v]ulnerabilities associated with dormant code may be
`
`down-weighted.” Id.
`
`152. A POSA would have understood this dormancy analysis of “code” (or
`
`“portions” of the image) to include, or at least suggest, assessing the
`
`dormancy/activity of installed applications. Supra, §XII.B.5. For example,
`
`Hufsmith notes that the distributed applications represented by the images can
`
`include multiple applications, which it refers to as “bodies of code.” EX1078
`
`(Hufsmith), Abstract, ¶¶3-4. Such applications/code are also “portions” of the
`
`image. A POSA would have recognized that identified malware/CVEs include such
`
`applications/code, meaning that the use determinations I discussed above are being
`
`applied to the “matching installed applications” detected by that analysis. E.g.,
`
`EX1078 (Hufsmith), ¶135 (“The weights determined by the context evaluator 125
`
`134
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 5 of 38 PageID
`#: 5607
`
`may be applied to one or more of the weights determined by the scanner property
`
`evaluators e.g., 105, 110, 115, and 120 to modify them responsive to contextual
`
`information.”).
`
`153. Second, Hufsmith teaches or suggests making the recited use
`
`determination as part of the initial calculation of individual weights. Discussing the
`
`calculation of
`
`individual weights for malware—which
`
`includes
`
`installed
`
`applications that are infected with malware—the malware evaluator considers
`
`“whether the file or code is executed” (i.e., whether the matching installed
`
`application is used by the protected virtual cloud asset). Hufsmith explains that the
`
`malware evaluator evaluates the malware properties of a virtual image, and the
`
`malware properties include a list of known malicious files detected within the virtual
`
`image. Id., ¶121. The list of detected malicious files further includes information
`
`about the file such as “whether the file or code is executed.” Id. The malware
`
`evaluator evaluates this information—including whether the file/code has been
`
`executed—to determine one or more malware weights. Id., ¶124; see also id., ¶¶125-
`
`28 (describing how to determine malware weights, metrics, and values).
`
`154. Additionally, a POSA would have recognized
`
`that
`
`the use
`
`determinations taught by Hufsmith are equally applicable and useful in the context
`
`of the Veselov’s assessment, since that assessment similarly evaluates software
`
`135
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 6 of 38 PageID
`#: 5608
`
`applications. EX1007, (Veselov), 1:34-40 (VMs may include containers, “and each
`
`container has its own process and network space in which the container may, for
`
`example, execute software programs”), 3:32-40 (snapshot may include “a database
`
`of software packages installed in a virtual machine”), 7:16-22 (assessment may
`
`include “software application-level CVEs assessment”), 9:59-67 (snapshot
`
`properties may include “software packages installed” and “software patches
`
`applied”), 13:58-63 (assessment VM includes “software applications 364A”), 17:23-
`
`29 (scanning service reads snapshot data, including “software application 564A”),
`
`18:34-38 (scanning service identifies “software packages”). Determining whether
`
`matching installed applications are used by a security assessment target, as taught
`
`by the combination of the Veselov and Hufsmith, was also consistent with common
`
`practice in the field, since it was well known that whether a vulnerable application
`
`is executed is relevant to level of risk presented by that vulnerability. EX1063
`
`(Hibbert), ¶80 (“[A] potential exclusion or mitigation for the vulnerability report is
`
`based on the context of the users executing Internet Explorer within your
`
`environment. But what if no one uses Internet Explorer, and you have standardized
`
`on another browser like FireFox or Chrome? Yes, the system is technically
`
`vulnerable but the offending application is not used and therefore a lower risk….”);
`
`see also id., ¶¶113-16 (security assessment considers which vulnerable applications
`
`136
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 7 of 38 PageID
`#: 5609
`
`“are being run” and other usage factors). In other words, it was well known that if
`
`a system contains a vulnerability associated with an installed application, that
`
`vulnerability tends to present a relatively greater risk if the application is used and a
`
`relatively lower risk if it is not used. Id., ¶¶80, 113-16.
`
`155.
`
`It is therefore my opinion that the combination of Veselov and
`
`Hufsmith teaches all aspects of elements 1.5, 18.5, and 22.5.
`
`8.
`
`Elements 1.6, 18.6, and 22.6
`
`156. These elements recite “[prioritizing/prioritize] the potential cyber
`
`vulnerabilities based on the use determinations.” As I explained above, Hufsmith
`
`teaches use determinations in two ways: (1) a dormancy/activity determination by
`
`the context evaluator and (2) a determination by the malware evaluator (and
`
`suggested for the CVE evaluator) of whether a file/code is executed. Supra,
`
`§XII.B.6. As I explain below, Hufsmith further teaches prioritizing the
`
`vulnerabilities based on these use determinations. See also supra, §XII.A (reasons
`
`to combine).
`
`157. First, with respect to dormancy determination by the context evaluator,
`
`Hufsmith teaches deprioritizing vulnerabilities associated with applications that are
`
`not used. As Hufsmith explains, “[t]he weights determined by the context evaluator
`
`125 may be applied to one or more of the weights determined by the scanner property
`
`evaluators,” for example, the malware and CVE evaluators, “to modify them
`
`137
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 8 of 38 PageID
`#: 5610
`
`cross-examination in this case and that cross-examination will take place within the
`
`United States. If cross-examination is required of me, I will appear for cross-
`
`examination within the United States during the time allotted for cross-examination.
`
`255.
`
`I declare that all statements made herein of my knowledge are true, and
`
`that all statements made on information and belief are believed to be true, and that
`
`these statements were made with the knowledge that willful false statements and the
`
`like so made are punishable by fine or imprisonment, or both, under Section 1001 of
`
`Title 18 of the United States Code.
`
`Dated: 5/24/2024
`
`By:
`
` /Angelos Stavrou/
`Angelos Stavrou, Ph.D.
`
`198
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 9 of 38 PageID
`#: 5611
`
`Joint Appendix
`Exhibit 37
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 10 of 38 PageID
`#: 5612
`
`
`
`
`
`
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`_____________________________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`_____________________________
`
`WIZ, INC.,
`Petitioner,
`
`v.
`
`ORCA SECURITY LTD,
`Patent Owner.
`
`_____________________________
`
`Case IPR2024-01191
`Patent No. 11,775,326
`_____________________________
`
`
`DECLARATION OF DR. ANGELOS STAVROU
`
`
`
`
`
`
`
`
`
`WIZ, Inc. EXHIBIT - 1002
`WIZ, Inc. v. Orca Security LTD.
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 11 of 38 PageID
`#: 5613
`
`offending application is not used,” it presents “a lower risk”); EX1078 (Hufsmith),
`
`¶140 (“Vulnerabilities associated with dormant code may be down-weighted in
`
`some cases.”).
`
`211. It is therefore my opinion that claim 7 was obvious.
`
`7.
`
`Claim 8
`
`212. Claim 8 depends from claim 7 and further recites, “wherein determining
`
`whether the matching installed applications are used by the respective protected
`
`virtual cloud asset includes checking configuration files of the matching installed
`
`applications to determine whether at least one of the matching installed applications
`
`is not in use, and wherein prioritizing reduces priority of the at least one matching
`
`installed application not in use.”
`
`213. At the outset, I note that “the matching installed applications” limitation
`
`has not been previously introduced in any of claims 7, 3, or 1 from which claim 8
`
`depends. This language is only used in claims 4 and 5. Nevertheless, the
`
`combination of Veselov and Basavapatna teaches matching installed applications
`
`against lists of known vulnerable applications for the reasons I discussed above in
`
`§§XII.C.3-4. As I explain below, the combination also teaches (1) determining
`
`whether these matching installed applications are used by the respective virtual
`
`cloud asset, (2) this determination includes “checking configuration files of the
`
`matching installed applications to determine whether at least one of the matching
`
`
`
`175
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 12 of 38 PageID
`#: 5614
`
`installed applications is not in use, and (3) prioritizing that reduces priority of the at
`
`least one matching installed application not in use.”
`
`214. First, the combination teaches determining whether the matching
`
`installed applications are used by the respective protected virtual cloud asset. The
`
`previously discussed application-usage determinations include use determinations
`
`for “matching installed applications.” As I explained above, Veselov and
`
`Basavapatna each teach identifying vulnerabilities on an asset by matching the
`
`asset’s installed applications against a list of applications known to be vulnerable.
`
`Supra, §XII.C.3. A POSA would have understood that Basavapatna’s application-
`
`usage determinations apply to these matching installed applications since the
`
`application-usage determinations are performed as part of risk assessments for the
`
`detected vulnerabilities, and the detected vulnerabilities include the matching
`
`installed applications. See, e.g., EX1008 (Basavapatna), ¶105 (determining whether
`
`“any test whose outcome is included in the vulnerability detection data” indicates
`
`that the asset possesses “the corresponding vulnerability or configurations
`
`evidencing the vulnerability”). At a minimum this would have been obvious since
`
`performing the risk determinations (including the application-usage determinations)
`
`for the matching installed applications would have been a straightforward way to
`
`avoid unnecessary risk determinations for thousands of vulnerable applications that
`
`
`
`176
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 13 of 38 PageID
`#: 5615
`
`are not even installed on the asset. See id., ¶72 (“In some systems, the number of
`
`identified, verified, catalogued, or otherwise known system vulnerabilities can be in
`
`excess of 30,000 or more….”).
`
`215. Second, the combination teaches that determining whether matching
`
`installed applications are used by the respective protected virtual cloud asset
`
`includes “checking configuration files of the matching installed applications to
`
`determine whether at least one of the matching installed applications is not in use.”
`
`As an initial matter, Veselov teaches that the snapshot includes configuration files
`
`of installed applications and that the security assessment considers these
`
`configuration files. EX1007 (Veselov), 2:35-41 (analysis may include “host
`
`configuration assessments”), 7:16-22 (exemplary software application-level CVE
`
`assessment based on information that includes list of “all software packages installed
`
`on the virtual machine and their configurations”), 9:55-67 (snapshot includes
`
`“software packages installed, and configurations thereof”). Basavapatna further
`
`teaches that the application-usage determination—which includes determining
`
`whether a given application is “not in use,” as I explained in §XII.C.6—involves
`
`checking application configuration files. The vulnerability-centric and threat-centric
`
`risk metrics are each determined by comparing vulnerability definition data with
`
`“asset configuration data” to determine whether or not an application is in use.
`
`
`
`177
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 14 of 38 PageID
`#: 5616
`
`Supra, §XII.C.6; EX1008 (Basavapatna), ¶¶53 (vulnerability definition data
`
`includes applicability data such as “software application versions” and “application
`
`settings”), 60 (vulnerability detection data generated by “comparing asset
`
`configuration data 207 and vulnerability definition data 205”), 79 (determination of
`
`threat-centric risk metric includes comparing “applicability data for a given threat to
`
`configuration data for a given asset”), 84. For example, determining the
`
`vulnerability-centric risk metric involves “analyzing the vulnerability detection data
`
`206 and/or asset configuration data 207 to determine whether any test whose
`
`outcome is included in the vulnerability detection data identified the asset as
`
`possessing the corresponding vulnerability or configurations evidencing the
`
`vulnerability.” Id., 105 (emphasis added). Basavapatna further explains that this
`
`asset configuration data includes “the software running on the asset, and the
`
`configuration of the software running on the asset.” Id., ¶37 (emphasis added);
`
`see also id., ¶53 (applicability data can specify “software application versions,
`
`application settings, plug-ins, etc.”).
`
`216. At a minimum, given Basavapatna’s teachings that asset configuration
`
`data includes application configurations and that application usage is determined
`
`based on the asset configuration data, it would have been obvious to determine
`
`application usage by checking application configuration files. For example, it was
`
`
`
`178
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 15 of 38 PageID
`#: 5617
`
`well known that such configuration files included information indicating whether or
`
`not applications were in use. See, e.g., EX1079 (Carroll), 31 (Windows registry
`
`tracks “system and application configuration information, as well as user activity”),
`
`32 (a “System Resource Usage Monitor” or “SRUM” collects information allowing
`
`determination of “which applications were running”), 33-34 (“With every Windows
`
`application, developers have the ability to create their own set of registry keys to
`
`track specific configuration and user activity for their application…. Each
`
`application in the Office suite has its own set of ‘FileMRU’ (most recently used files)
`
`that tracks most recent files used and when they were opened.”), 35-36 (further
`
`discussing SRUM).
`
`217. Third, Basavapatna teaches “wherein prioritizing reduces priority of the
`
`at least one matching installed application not in use.” As I discussed above with
`
`respect to claim 7, Basavapatna teaches determining that an application is not in use
`
`as part of determining the vulnerability-centric and threat-centric risk metrics.
`
`Supra, §XII.C. 6. Basavapatna further teaches that the determined risk is lowered if
`
`the application is not running. As I previously explained, both risk metrics may be
`
`based on the same vulnerability detection score, which may be set to a first value
`
`(“e.g., one”) if the application is running or a lower value if the application is not
`
`running (e.g., “zero” when the asset is not vulnerable or “some other value between
`
`
`
`179
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 16 of 38 PageID
`#: 5618
`
`zero and one” when the asset has a lower probability of being vulnerable). EX1008
`
`(Basavapatna), ¶104; see also id., ¶¶78, 83-84 (discussing similar scoring for
`
`applicability/vulnerability scores used in threat-centric risk metric). Applying a
`
`lower coefficient if the application is not running “reduces priority of the at least one
`
`matching installed application,” as recited, since this lowers the ultimate risk metric,
`
`and Basavapatna teachings prioritizing vulnerabilities based on those metrics. For
`
`example, Basavapatna teaches that the system “can list all assets, vulnerabilities, or
`
`threats, sorted by aggregate risk metric, or can list a top number, e.g., top ten, of the
`
`assets, vulnerabilities, or threats.” Id., ¶138; see also id., ¶¶120-28 (aggregate risk
`
`metric for particular vulnerability/threat based on risk metrics for
`
`that
`
`vulnerability/threat across multiple assets). As Basavapatna explains:
`
`Ranking assets, vulnerabilities, and/or threats according to aggregate
`risk scores allows a user to quickly identify which assets are most at
`risk, or which vulnerabilities and threats are most dangerous for a
`system. The user can then remediate the most at-risk assets before
`remediating other less-at-risk assets, or can apply remediations across
`the system for the riskiest vulnerabilities and threats before applying
`remediations for other, less-risky ones.
`
`Id., ¶138; see also id., ¶140 (system may filter out lower-risk vulnerabilities); supra,
`
`§XI.B (Basavapatna summary). At a minimum, reducing the priority of matching
`
`installed applications that are not in use would have been obvious in view of
`
`
`
`180
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 17 of 38 PageID
`#: 5619
`
`Basavapatna’s teachings since it was well known that vulnerabilities associated with
`
`applications that are not in use tend to present lower risk and are therefore a lower
`
`priority. EX1063 (Hibbert), ¶80 (if “the offending application is not used,” it
`
`presents “a lower risk”); EX1078 (Hufsmith), ¶140 (“Vulnerabilities associated with
`
`dormant code may be down-weighted in some cases.”).
`
`218. It is therefore my opinion that claim 8 was obvious.
`
`8.
`
`Claim 9
`
`219. Claim 9 depends from independent claim 1, and its preamble recites
`
`“wherein analyzing the snapshot of the at least one virtual disk of the respective
`
`protected virtual cloud asset further includes….” The claim then recites additional
`
`limitations that are taught by Veselov, as I explain below.
`
`a.
`
`Elements 9.1
`
`220. This element recites “parsing the snapshot of the at least one virtual
`
`disk.” As I explain below, Veselov teaches this element.
`
`221. Veselov teaches such parsing in several ways. For example, Veselov
`
`explains that “[p]rocessing the snapshot data 146 may require applying one or more
`
`rules, instructions, and/or transformations to the snapshot data 146.” EX1007
`
`(Veselov), 8:21-23 (emphasis added). Veselov also describes formatting the
`
`snapshot data with tags identifying the target resource. Id., 9:32-37 (“Additionally
`
`or alternatively, the scanning system may specify parameters related to the format
`
`
`
`181
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 18 of 38 PageID
`#: 5620
`
`includes determining the particular asset’s criticality “based on other assets in the
`
`cloud computing environment that are accessible from the particular protected
`
`virtual cloud asset” (supra, §XIV.E) “and based on the contents stored from the
`
`particular protected virtual cloud asset” (supra, §XIV.D).
`
`326. It is therefore my opinion that claim 26 was obvious.
`
`XV. CONCLUDING STATEMENTS
`
`327. In signing this declaration, I understand that the declaration will be filed
`
`as evidence in a contested case before the Patent Trial and Appeal Board of the
`
`United States Patent and Trademark Office. I acknowledge that I may be subject to
`
`cross-examination in this case and that cross-examination will take place within the
`
`United States. If cross-examination is required of me, I will appear for cross-
`
`examination within the United States during the time allotted for cross-examination.
`
`328. I declare that all statements made herein of my knowledge are true, and
`
`that all statements made on information and belief are believed to be true, and that
`
`these statements were made with the knowledge that willful false statements and the
`
`like so made are punishable by fine or imprisonment, or both, under Section 1001 of
`
`Title 18 of the United States Code.
`
`Dated: 08/07/2024
`
`
`
`
`
`
`By:
`
`
` /Angelos Stavrou/
`Angelos Stavrou, Ph.D.
`
`
`
`253
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 19 of 38 PageID
`#: 5621
`
`Joint Appendix
`Exhibit 38
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 20 of 38 PageID
`#: 5622
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 21 of 38 PageID
`#: 5623
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 22 of 38 PageID
`#: 5624
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 23 of 38 PageID
`#: 5625
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 24 of 38 PageID
`#: 5626
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 25 of 38 PageID
`#: 5627
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 26 of 38 PageID
`#: 5628
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 27 of 38 PageID
`#: 5629
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 28 of 38 PageID
`#: 5630
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 29 of 38 PageID
`#: 5631
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 30 of 38 PageID
`#: 5632
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 31 of 38 PageID
`#: 5633
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 32 of 38 PageID
`#: 5634
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 33 of 38 PageID
`#: 5635
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 34 of 38 PageID
`#: 5636
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 35 of 38 PageID
`#: 5637
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 36 of 38 PageID
`#: 5638
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 37 of 38 PageID
`#: 5639
`
`
`
`Case 1:23-cv-00758-JLH-SRF Document 203-2 Filed 11/22/24 Page 38 of 38 PageID
`#: 5640
`
`