IN THE SUPERIOR COURT OF FULTON COUNTY

STATE OF GEORGIA

STATE OF GEORGIA, )
)
vs. ) Case No. 23SC188947
)
HARRISON FLOYD, et al )
)
DEFENDANT. )

DEFENDANT HARRISON FLOYD’S CONSOLIDATED

OPPOSITION TO MOTIONS TO QUASH SUBPOENAS

Harrison Floyd files this consolidated opposition to the Motions to Quash Mr. Floyd’s

Subpoena Duces Tecum (“Subpoena”) filed by the Fulton County Clerk of Superior Court

(“Clerk”), Fulton County Board of Registration and Elections (“Board”), and the Secretary of

State (“Secretary”). 1 The information that Mr. Floyd seeks is relevant, vital, and crucial to his

defense.

Introduction

Mr. Floyd intends to present as his defense, evidence that President Trump did not lose

the November 3, 2020 election in the State of Georgia. He intends to produce, for example,

evidence that more than 41,079 presidential votes should have been excluded from the final

Fulton County vote tally because they were not and cannot be reconciled or properly validated as

legitimate votes. If the 41,079 votes were not properly accounted for or verified, they should not

have been included in the final tally and President Trump would have won Georgia's election by

a net margin of at least 3,886 votes. See Exhibit A, Sample #11 Summary.

Mr. Floyd intends to show that these votes should not have been included in Fulton

County’s final result because of pervasive and systemic failures to comply with the processes

Mr. Floyd has consolidated this response for convenience because the objections allege identical grounds. The
1

Clerk, Board, and Secretary may be collectively referred to as the “Non-Parties” in this opposition.
and rules promulgated by the State Election Board. 2 For example, there is evidence that hundreds

of thousands of ballot images including the 41,079 sample above are missing, that ballots were

counted two and three times, there are a lack of Scanner Recap sheets, missing tabulator sheets,

missing tabulator system logs, the use of 10 tabulators that were not part of Fulton County’s

inventory and for which serial numbers do not exist, and the deletion of approximately 132,280

“.sha” files which are automatically created and produce a unique file identifier for each ballot

image in order to authenticate ballots. See Exhibit B, Sample #11 Support.

While this small 41,079 sample may have other explanations, only compliance with the

subpoena can assist Mr. Floyd’s evaluation of this and other evidence that he intends to present

to negate criminal intent, and to rebut the indictment’s “knowing and willful” allegations and

false statements. It is vital that Mr. Floyd be allowed to determine truth or falsity of these things.

So it seems strange that the Non-Parties are all moving to quash Mr. Floyd’s subpoenas on

relevance grounds when the political indictment before this Court alleges:

• 17 times that President Trump lost the 2020 election including in Georgia and
Fulton County;

• 49 times that the defendants engaged “knowingly” and “willfully” in various

activities to overturn the 2020 election when they allegedly knew their statements
about the election results were false;

• 113 times that there were attempts to “unlawfully” change the outcome of the
election, including in Georgia;

• 145 times that the defendants sought to put the November 3, 2020 election results
at issue with the implication that the results cannot be questioned; and

• 182 times that the RICO defendants engaged in “false statements” concerning
“fraud” in and about the November 2020 elections;

2
See, e.g., Ga. Comp. R. & Regs. § 183-1-14-.02.

2
The Non-Party objections are strange because this is a prime opportunity for them to show with

complete transparency that the Fulton County election results in the 2020 election are inherently

trustworthy, reliable, and true. It is an opportunity to discredit Mr. Floyd and all the defendants,

including President Trump, who claim the election was stolen and to once and for all put the

concerns of so-called “election deniers” to bed.

However, it is more likely that the Non-Parties objections constitute an effort to avoid

putting on display, the legion of irregularities in Georgia’s November 2020 election that clearly

affected the result, especially with respect to Fulton County. It is Fulton County’s untrustworthy

process and gross irregularities that made the election results unreliable and fomented the lack of

voter confidence. Governor Kemp wrote to the State Election Board about 36 errors in Fulton

County’s data and pointed to flaws in the County’s audit, asserting that its 2020 audit report was

“sloppy, inconsistent, and presents questions about what processes were used by Fulton County

to arrive at the result.” See Exhibit C, Gov. Kemp Letter to SEB dated November 17, 2021

(emphasis added). From the evidence in Mr. Floyd’s possession, 35 of the 36 errors appear to

favor President Biden.

As another example, the declaration of an expert employed by a Democrat plaintiff in the

ongoing federal case, Curling v. Raffensperger, 3 shows why the information sought is relevant to

Mr. Floyd’s defense. Curling’s expert, Professor Phillip Stark, is an internationally renowned

statistician and Associate Dean of Mathematical and Physical Sciences at the University of

California University, Berkely. He is literally the inventor of the risk limiting audit used by many

governments and which Georgia claimed to employ. His conclusions in his 32-page 2019

declaration in the Curling litigation (reaffirmed in 2022) are startling. He opines that “[t]here is

3
Curling, et al. v. Brad Raffensperger, et al., No. 1:17-cv-2989-AT, (N.D. Ga. 2020.).

3
no trustworthy inventory of ballots to check the results against, because of Georgia’s lax

canvass.” Exhibit D, Prof. Philip B. Stark Report, ¶ 49 at 15. Professor Stark went on to state that

inventory was critical and what that meant:

The ballot manifest must be based on physical inventories of the ballots, keeping track of
where the ballots are and how they are organized. Absent that, it is impossible to account
for votes reliably, and impossible to limit the risk that an incorrect electoral outcome will
be certified: applying risk-limiting audit procedures to an untrustworthy collection of
ballots is “security theater.”

Ex. D, ¶ 87(g) at 31 (emphasis added). Even more compelling are Prof. Stark’s conclusions about

Fulton’s complete lack of organization and record keeping of election materials:

Based on the observations in paragraphs 58 through 78, supra, it seems that Fulton County
did not keep track of which ballots and BMD cards had been scanned and which had not,
in both the original count and in the machine recount. . . . This is a surprising lack of
tracking and protecting election materials: the most basic election safeguard is to check
whether the number of voters who participated is equal to the number of ballots and BMD
printout cards that were cast and to the number that were tabulated. Moreover, I would
expect all electronic election materials to be backed up onsite and offsite, at least for the
federally mandated retention period of twenty-two months, so the loss of hundreds of
thousands of image files from the first machine count and of nearly 18,000 images from
the second machine count is hard to fathom.

Ex. D, ¶ 80 at 25-26 (emphasis added). On behalf of a Democrat organization and Democrat

plaintiff, Prof. Stark plainly states that Fulton County’s election results are not to be trusted:

Fulton County’s chaotic, unaccountable curation and processing of cast ballots, cast
BMD printout, and electronic records make a true risk-limiting audit impossible. It is
unreasonable for voters to trust that their votes were counted at all, much less counted
correctly. Voters have good reason to believe that some votes counted more than others:
some votes were included twice or thrice in the totals. There is no way to know how many
votes were omitted from the tabulation, absent access to the physical ballots and BMD
printout and evidence that the chain of custody is intact. From the records produced so
far, it is impossible to determine whether malware, bugs, misconfiguration, or
malfeasance disenfranchised voters or altered the election results.

Ex. D, ¶ 82 at 26 (emphasis added).

And finally, from civil discovery obtained in the Curling matter, the County has made

glaring admissions which will impact the evidence in this case. For example, the County admits

4
that it has not preserved “the majority of ballot images from in-person voting for the November

3, 2020 election”; that it had not received from the Secretary, “instructions, guidelines, or

recommended procedures for reconciling the discrepancies” in the recount; and, that it has “not

been required by the Secretary to provide an explanation of audit tallies and machine tallies in

the November 3, 2020 election.” See Exhibit E, Defendant Fulton County Responses to Requests

for Admissions, ¶1 at 1 and ¶¶23-24 at 8.

The Motions to Quash are, quite frankly, based on misstatements of law and fact. The

requested documents and materials go to the heart of Mr. Floyd’s defense—a defense

necessitated by the indictment’s claim that Mr. Floyd “willfully and knowingly” engaged in

alleged crimes to overturn a valid and lawful election and engaged in false statements concerning

the November 3, 2020 election. If Mr. Floyd’s evidence, such as the County’s inability to

reconcile 41,079 suspicious votes, are true, then the State’s indictment about many things

including false statements evaporates—instantly. If Mr. Floyd’s analysis of the produced

evidence shows that the Fulton County election results are inconclusive, the State’s indictment

including false statements evaporates—instantly. If Mr. Floyd’s evidence turns out to in fact be

false, Mr. Floyd intends to assert an affirmative defense of mistake of fact pursuant to O.C.G.A.

§ 16-3-5 (2022). In any case scenario, the information is highly relevant and completely

necessary.

MEMORANDUM OF LAW

I. Overview of Criminal Discovery in Georgia

A subpoena for documents or materials is allowed in criminal cases “insofar as consistent

with the Constitution.” Coleman v. State, 301 Ga. 720, 726 (2017); see also, O.C.G.A. § 24-13-

20 (2022). Documents and materials requested must be relevant and “specific enough to the

questions at issue” for which the items are procured. Reese v. State, 252 Ga. App. 650, 653
5
(2001) (emphasis added). Relevant evidence is defined as that “evidence having any tendency to

make the existence of any fact that is of consequence to the determination of the action more

probable or less probable than it would be without the evidence.” O.C.G.A. § 24-4-401 (2022)

(emphasis added). Mr. Floyd has the burden of showing relevance.

When a motion to quash a subpoena is filed, the party serving the subpoena has the initial
burden of showing the documents sought are relevant. Where the evidence sought in a
subpoena duces tecum is demonstrably relevant and material to the defense, it is error for
a trial court to quash the subpoena.
Harris v. State, 341 Ga. App 831 (2017). Additionally, the State, as the prosecuting agency, is

required to “turn over evidence in its possession that is material to guilt or punishment and is

favorable to the accused.” Bello v. State, 300 Ga. 682, 683 n.3 (2017), citing Brady v. Maryland,

373 U.S. 83 (1963). Mr. Floyd asserts that the evidence in possession of the Non-Parties is

material to issues related to his criminal culpability and to any punishment he might receive if

convicted.

II. The Amended Subpoena is not overly broad

The Secretary claims that the subpoena is

overly broad, generally lacking in a reasonable scope with respect to subject matter or
time and seeks responses to multiple types of overbroad and unbounded requests that
Georgia courts have repeatedly quashed, including requests for ‘any and all documents’
in varying mediums without any restriction of subject-matter, time, or any meaningful
measure.

Sec’y of State Motion to Quash or Modify the Subpoena Served by Harrison Floyd (“Secretary’s

Motion”), at 3. Despite the fact that Mr. Floyd’s amended subpoena limits the discovery to the

November 3, 2020 election, the Secretary claims that this limitation is merely “a distinction

without meaning, as explained below,” yet fails to adequately explain why this oft-used phrase is

somehow confusing or too broad in scope. Id. at 6.

6
 The requested materials in the amended subpoena track the same language used by the

prosecutor in the indictment. One has to wonder how this same language can be legally sufficient

for the indictment, yet fail to be sufficient for a subpoena duces tecum used to defend against the

indictment. The amended subpoena is also not general discovery, because the requests within the

amended subpoena are specifically tailored to events related to the indictment. Neither is the

amended subpoena a fishing expedition, for the requests are based on the Secretary’s own

response in that they state that they are in possession of a fair amount of material related to

investigations of complaints of voter fraud (approximately 296 complaints) about the November

3, 2020, presidential election. Thus, the amended subpoena is neither overly broad nor does it

lack in scope.

III. The requested materials are relevant and material to Mr. Floyd’s defense

The indictment has swung the door wide open when it made whether former President

Donald J. Trump won the state of Georgia in the 2020 presidential election a pivotal and material

issue. If the material that Mr. Floyd seeks clearly supports Mr. Floyd’s lack of criminal intent,

then evidence of the election results are material and probative. As discussed above in the

introduction, Mr. Floyd does not even have to prove that Trump won. It would be enough to

show that the results are inconclusive and should not have been certified. If Mr. Trump is denied

the opportunity to review this evidence, then the State would likewise be barred from introducing

any evidence concerning the election—including whether Trump lost. Yet, the State has made

the election result a material issue throughout the indictment in the following allegations:

a) acted “unlawfully” (Count 1 at 13);

b) “knowingly and willfully joined a conspiracy to unlawfully change the outcome

of the election in favor of Trump” (Count 1 at 14);

7
c) “At these public hearings, members of the enterprise made false statements
concerning fraud in the November, 3 2020, presidential election (Count 1, ¶1 at 16);

d) that the purpose of these statements was to reject “lawful” electoral votes cast
(Count 1, ¶1 at 16);

e) “…to unlawfully appoint their own presidential electors…” (Count 1, ¶1 at 16);

f) “by unlawfully changing the outcome of the November 3, 2020, presidential

election in Georgia…” (Count 1, ¶2 at 16-17);

g) “…to unlawfully change the outcome of the November 3, 2020, presidential

election…” (Count 1, ¶3 at 17);

h) “…to unlawfully change the outcome of the November 3, 2020, presidential

election…” (Count 1, ¶4 at 17);

i) Solicited the Vice President of the United States to violate the United States
Constitution and federal law by unlawfully rejecting Electoral College votes cast in
Fulton County, Georgia by the duly elected and qualified presidential electors from
Georgia” (Count 1, ¶6 at 18);

j) “falsely claimed voter fraud” (Count 1, Act 1 at 20);

k) made “statements concerning fraud in the November 3, 2020, election in Fulton

County, Georgia” (Count 1, Act 2 at 20);

l) made “false statements concerning fraud in the November 3, 2020, presidential

election in Georgia and elsewhere” (Count 1, Act 2 [sic] at 20);

m) “knowingly, willfully, and unlawfully, making . . . false statements to the Georgia

Senate” about mail-in ballots counted in the November 3, 2020, presidential election in
Georgia (Count 1, Act 24 at 25);

n) “knowingly, willfully, and unlawfully” making “false statements to the Georgia

Senate” about felons illegally voting, underaged voters voting, unregistered voters voting,
illegally registered voters voting, and dead voters voting, in the November 3, 2020,
presidential election in Georgia (Count 1, Act 25 at 26);

o) making “false statements concerning fraud in the November 3, 2020, presidential

election in Georgia” (Count 1, Act 29 at 27);

p) placing “a telephone call to Georgia Attorney General Chris Carr for the purpose
of making false statements concerning fraud in the November 3, 2020, presidential
election in Georgia and elsewhere” (Count I, Act 43 at 30);
8
 q) acting to appoint, instruct, or cast electoral votes of “Trump presidential elector
nominees in Georgia . . . despite the fact that DONALD JOHN TRUMP lost the
November 3, 2020, presidential election . . .” (Count I, Act 46 at 31 along with Acts 48
and 49 at 31);

r) “knowingly, willfully, and unlawfully making . . . false statements to the Georgia

House of Representatives about the November 3, 2020, presidential election in Georgia”.
(Count 1, Act 56 at 34);

s) creating documents for presidential elector nominees in Georgia “despite the fact
that DONALD JOHN TRUMP lost the November 3, 2020, presidential election in
Georgia.” (Count 1, Acts 61 at 35 and 71 at 38);

t) The Indictment states specifically that an alleged co-conspirator stated that:

“During the phone call DONALD JOHN TRUMP falsely stated that he won the
November 3, 2020, presidential election in Georgia . . .” (Count 1, Act 93 at 44);

u) “knowingly, willfully, and unlawfully” making false statements pertaining to the

election results of the November 3, 2020, presidential election. (Count 1, Act 113 at 51);

v) making “false statements concerning fraud in the November 3, 2020, presidential

election . . .” (Count 1, Act 137 at 62);

w) tweeting “States want to correct their votes, which they know were based on
irregularities and fraud . . .” (Count 1, Act 139 at 63);

x) knowingly, willfully, and unlawfully making the following false statement: “As
stated to you previously, the number of false and/or irregular votes is far greater than
needed to change the Georgia election result” (Count 1, Act 157 at 68) and;

y) “knowingly and willfully making a false statement and representation concerning

events at State Farm Arena in the November 3, 2020, presidential election in Georgia…”
(Count 30 at 89).

Therefore, who actually won or lost in the 2020 election is material and relevant because

it goes to the defendants’ criminal intent. Mr. Floyd stands charged with Conspiracy to Solicit

False Statements and Writings, among other offenses, in connection with the alleged election

fraud surrounding the 2020 presidential election. What Mr. Floyd knew or believed regarding the

alleged falsity of statements and/or writings is an element of the purported crime for which he

9
stands accused. The requested materials are exculpatory, because the materials will show that on

January 4, 2021, Mr. Floyd’s beliefs were not unreasonable based on the evidence in the Non-

Parties’ possession, custody, and control.

These materials will show that this Defendant did not knowingly make false statements

or writings because the statements attributed to him in Count I were true. Alternatively, if the

statements turn out to be in fact false, these materials can show that Mr. Floyd did not knowingly

and willfully make false statements or writings—a mistake of fact defense. See O.C.G.A. § 16-3-

5 (2022).

The amended subpoena clearly narrows the time frame concerning the requested

materials to the relevant time period in which the Secretary of State received complaints

regarding election fraud after the election, and it cannot be narrowed further. The time frame is

relevant because it goes to reasonable belief, which negates criminal intent. The amended

subpoena is certainly specific enough for the Non-Parties to discern what materials are being

requested. The Secretary freely admits that his office opened investigations into approximately

296 complaints surrounding the 2020 primaries and elections, and continues to receive

complaints regarding the November 3, 2020 election. See SOS Motion to Quash at 8-9. Mr.

Floyd has narrowed this to include only the November 3, 2020 election and not primaries or

other elections. For the Non-Parties to aver that the requested materials are not relevant, or that

the requests were neither specific enough in their description (after identifying documents), or

that the requested material would be too burdensome to produce, or any of the other myriad of

excuses that the Non-Parties choose to use to justify their non-compliance are merely attempts to

stonewall the defense.

10
IV. Production of the Subpoenaed Materials is neither unreasonable nor oppressive

Production of the requested materials should not be nearly as burdensome as claimed by

the Secretary, and as such, the requests are neither unreasonable nor oppressive. If the provided

materials of Requests 1 and 2 are any indication, the materials in Requests 3-7 of the amended

subpoena are scanned and filed on a computer. Locating and copying such files to an external

drive would not take the months of time or effort claimed by the Secretary by even the least tech-

savvy employee of his office.

The Secretary attempts to inflate the time required to produce the requested materials by

claiming that the files would need to be redacted prior to release. Such redaction is unnecessary.

The Court can simply place a protective order on the documents and items produced and instruct

Mr. Floyd not to divulge the contents of the documents to outside third parties. If Mr. Floyd

wishes to introduce any of these documents in a hearing or at trial, the Parties can stipulate to an

in-camera inspection of any materials prior to such a hearing or trial, to allow the Secretary to

instruct what must redacted for said hearing. Further, the Court could order the information

unsealed and provide it in accordance state law. The Georgia Bureau of Investigation, which has

been unsurprisingly much more forthcoming in producing requested materials to a subpoena

duces tecum, has agreed to such a stipulation. Therefore, it is a stretch of the imagination to think

that the Secretary cannot do the same.

Instead of requesting a reasonable cost to produce the documents and materials, the

Secretary attempts to outright deny this Defendant of the requested records. Georgia law allows

for the Secretary to charge a reasonable cost to produce the requested materials. The Court may

“[c]ondition denial of the motion upon the advancement by the person in whose behalf the

subpoena is issued of the reasonable cost of producing the evidence.” O.C.G.A. § 24-13-

11
23(b)(2). Although the Secretary has not proposed payment of a reasonable cost to produce the

requested materials, this Defendant should be given the opportunity to pay a reasonable cost for

the production of the requested documents before the Court rules on the Non-Parties’ Motion to

Quash.

V. Conclusion

This case is about the liberty interests of a Defendant charged with alleged activities

surrounding an attempt to expose the fraud that he believed occurred in the election. The Non-

Parties’ contention that these subpoenas are not relevant is facially not true because the election

result is relevant and probative to the material issues in the indictment. Mr. Floyd has shown

that the requested materials are relevant to his charges and he did not create the Non-Parties’

concerns.

It is the State which put the outcome of the election front and center in the indictment,

and what the Defendant knew or believed to be true concerning the outcome. This contention

goes directly to Mr. Floyd’s guilt or innocence. If the outcome was that Trump actually won the

state of Georgia, then the Defendant did NOT knowingly make false statements and writings. If

the outcome was that Trump did not win the state of Georgia, then that it supports a mistake of

fact defense.

The requested documents and materials are not overly broad, unreasonable or oppressive.

The requests are specifically tailored to events related to the indictment, and tracks the same

language used in the indictment. If the language in the indictment is sufficient as written, then it

stands to reason that the language used in the amended subpoena is also sufficient as written. The

Defendant has shown that he is amenable to paying a reasonable cost to produce the materials

and is willing to stipulate to a protective order of the Court. Therefore, there remains no valid

12
reason why the Non-Parties should not be ordered to produce the requested materials, much like

the Georgia Bureau of Investigation has already produced.

Respectfully submitted this the 31st day of October 2023.

HARDING LAW FIRM, LLC

___________________________
Christopher I. Kachouroff, Esq.*
MCSWEENEY, CYNKAR & KACHOUROFF, PLLC
13649 Office Place, Suite 101
Woodbridge, Virginia 22192
(703) 365-9900

Todd A. Harding, For the Firm

Ga. Bar No.: 101562
HARDING LAW FIRM, LLC
Attorney at Law
113 E. Solomon Street
Griffin, Georgia 30223
(770) 229-4578
(770) 228-9111 facsimile

* Admitted Pro Hac Vice

Attorneys for Harrison Floyd

13
 IN THE SUPERIOR COURT OF FULTON COUNTY
STATE OF GEORGIA

STATE OF GEORGIA, )
)
vs. ) Case No. 23SC188947
)
HARRISON FLOYD, et al )
)
DEFENDANT. )

CERTIFICATE OF SERVICE
I hereby certify that on this day I have served counsel of record with the foregoing
DEFENDANT HARRISON FLOYD’S CONSOLIDATED RESPONSE TO NON-
PARTIES’ MOTION TO QUASH SUBPOENA, filed by electronic transmission addressed to
the following:
Fani T. Willis, DA
136 Pryor Street, SW
3rd Floor
Atlanta, Georgia 30303

Respectfully submitted this the 31st day of October 2023.

HARDING LAW FIRM, LLC

_______________________________
Christopher I. Kachouroff, Esq.*
MCSWEENEY, CYNKAR & KACHOUROFF,
PLLC
13649 Office Place, Suite 101
Woodbridge, Virginia 22192
(703) 365-9900

Todd A. Harding, For the Firm

Ga. Bar No.: 101562
HARDING LAW FIRM, LLC

* Admitted Pro Hac Vice

Attorneys for Harrison Floyd

14
 EXHIBIT A

Summary of Vote Irregularity #11

Biden Votes Trump Votes Source

Official 2020 Georgia Results 2,473,633 2,461,854 Biden +11,779

Missing Ballot Images (13,390) (4,319) Note 1 below

Duplicate Ballots (1,708) (1,404) Note 2 below

No Tabulators Tapes (13,274) (6,984) Note 3 below

Total Irreconcilable Votes (28,372) (12,707) (41,079) (See Notes 1, 2, and 3)

Corrected Results 2,445,261 2,449,147

Corrected Margin of Victory Trump +3,886

Notes:

1. Fulton County has confirmed 17,852 ballot images from the officially reported results
(i.e., Machine Count 2) are “missing.” In the presidential race, 17,709 missing ballot images
represent 13,390 Biden Votes and 4,319 Trump Votes.

2. Through analysis of the Official Results (Machine Count 2), Cast Voter Records, and
Ballot Images, there are an addition 3,125 ballots that have been scanned and counted multiple
times). In the presidential race, 3,183 duplicate ballots represented 1,708 Biden Votes and 1,404
Trump Votes.

3. Upon examining the Batches Loaded Report, Cast Vote Records, and in light of Fulton
County's admission that ten Tabulator Tapes are non-existent, it is evident that 20,713 votes lack
the necessary tabulator tapes detailing their counting process and timing. Within the presidential
contest, of the 20,532 votes for which tapes are missing, 13,724 were cast for Biden and 6,984
for Trump. The required Tabulator Tapes, which should be dated and signed by the responsible
tabulation personnel, are absent, calling into question the legitimacy of these votes, as evidenced
by the Cast Vote Records.

In light of the multiple tallying attempts, each marked by procedural discrepancies,

validation issues, and inadequate documentation, it becomes apparent that the results may not
endure scrutiny from an independent third-party auditor. While varied explanations were offered
for these discrepancies, there were numerous instances where no justification was presented for
the observed errors or lapses.
 1

Exhibit for Case No. 23SC188947

DEFENDANT HARRISON FLOYD’S CONSOLIDATED

OPPOSITION TO MOTIONS TO QUASH SUBPOENAS

EXHIBIT B

SUPPORT FOR SAMPLE 11

Summary:

This document outlines sources of the included records and methods of the calculations
described in the Opposition to Motions to Quash Subpoenas.

The data suggests that 41,079 votes from the Fulton County 2020 General Election (Presidential
contest) weren't accurately verified or accounted for. Consequently, these votes should not have
been counted in the official results.

These votes should have been omitted from the final count due to their unverified or unvalidated
status.

Fulton County 2020 Presidential Election Vote Totals

Biden Votes Trump Votes Vote Totals

A. Missing Ballot Images (13,390) (4,319) (17,709)

B. Duplicate Ballots (1,708) (1,404) (3,112)

C. No Tabulators Tapes (13,274) (6,984) (20,258)

Not Properly Accounted (28,372) (12,707) (41,079)

For Votes

The data represented in this analysis is from the Fulton County Original November 3rd, 2020
count (“Machine Count 1”) - And from the Fulton County Official results for the Presidential
Recount (“Machine Count 2”).
 2

A. Missing Ballot Images

17,852 ballot images are missing from the officially reported results (i.e., Machine Count 2)
In the presidential race, 17,709 missing ballot images represent 13,390 Biden Votes and
4,319 Trump Votes.

● A total of 17,852 ballot images from the Official Results (Machine Count 2) are
“missing.”
● Fulton County has confirmed in recent filings in a pending Federal lawsuit that it does
not have these 17,852 ballot images.
● These Votes were counted and included in the official recount results.

Missing Ballot Images Candidate

13,390 Biden

4,319 Trump

17,709 Total Duplicate Ballots

Analysis:

The total number of votes as reported by Fulton County for Machine Count 2 exceeds the total
number of ballot images by 17,852.

Upon examining the filenames of the ballot images supplied by Fulton County and cross-
referencing them with the filenames mentioned in the Cast Voter Records (CVR) from the
Machine Count 2 Audit, a list of absent filenames from the provided ballot images was
compiled. Subsequently, the votes from the CVR for each candidate were tallied.

Supporting Reference:

In Prof. Stark’s Declaration in the Curling vs. Raffensperger:

63. Fulton County did not produce the image file corresponding to every cast vote
record. For the first machine count, production included images of ballots or BMD
printout cards for only 168,726 of the 528,776 cast vote records: 376,863 image files are
missing. For the second machine count, Fulton County’s production included images of
ballots or BMD printout cards for 510,073 of the 527,925 cast vote records: 17,852
image files are missing
 3

64. Entire batches of images are missing from Fulton County’s production, for example,
images from Scanner 801 batch 117 and Scanner 801 batch 118 are referred to in the
cast vote records for the second machine count but the images were not among the
electronic records. Without additional discovery it is impossible to determine whether the
missing images are missing because of human error, programming errors (bugs), or
malware in Fulton County’s election management system (EMS). Of course, those
possibilities are not mutually exclusive.

Source: Curling-Phllip-Stark-Declaration-2022-03-09.pdf
Curling v. Raffensperger
Civil Action File No: 1:17-cv-2989-AT

Data Source:
2020 Fulton County Machine Count 2 - Ballot Images
Curling v. Raffensperger
Civil Action File No: 1:17-cv-2989-AT

2020 Fulton County Machine Count 2 - Cast Vote Records (CVRs)

B. Duplicate Ballots

Through analysis of the Official Results (Machine Count 2), Cast Voter Records, and Ballot
Images, there are 3,125 ballots that have been scanned and counted multiple times). In the
presidential race, 3,112 duplicate ballots represented 1,708 Biden Votes and 1,404 Trump Votes.

Duplicates Votes Candidate

1,708 Biden

1,404 Trump

3,112 Total Duplicate Ballots

Analysis:
The analysis involved detecting duplicate ballot IDs in the same sequential order for
corresponding batches of ballots using the Cast Vote Record. Upon identifying anomalies, both
machine-based and visual examinations of the ballot images were conducted to ascertain if the
ballots displayed characteristics indicating they were identical.
 4

Supporting Reference:

In Prof. Stark’s Declaration in the Curling vs. Raffensperger:

75. For Fulton County as a whole, Coalition plaintiffs gave me a list that identified
images of 2,871 ballots and BMD printout cards that they believe were counted two or
three times in the second machine count. Some were identified by visual inspection of the
images; others were inferred to be duplicates because a sequence of cast vote records
was identical (or reversed) for long portions of two scan batches. As mentioned in
paragraph 72, supra, I confirmed that 214 of the purported duplicate scans of BMD
cards were indeed duplicates. I understand that this list of 2,871 are a sample from a
larger list of images of ballots and BMD printout cards that Coalition Plaintiffs assert
were included in the tabulation twice or more. I confirmed that all 6,118 images in
question were referenced in cast vote records in the second machine count, so all
presumably contributed to the tabulation.

Source: Curling-Phllip-Stark-Declaration-2022-03-09.pdf
Curling v. Raffensperger
Civil Action File No: 1:17-cv-2989-AT

*While Prof. Stark’s analysis was limited by the time constraints of the trial, further examination
revealed more duplicated ballot images. In Fulton County alone, at least 3,183 duplicate ballots
were identified. 1 Subsequent reviews of other counties in Georgia also uncovered similar
discrepancies.

In Duncan Buell’s Declaration in the Curling vs. Raffensperger:

24. The JSON filename, tabulator, batch, and sequence numbers will (unless there is an
extraordinary coincidence) be different for a ballot recorded in the original CVR when
compared to a ballot in the recount CVR. However, one can view the rest of each line for
a ballot as a "signature": in that precinct portion, with that ballot style, cast in that
fashion, what the actual choices are recorded on the ballot image.

I use the term "signature" of the ballot image to mean a index created in our analysis of
the ballot images where we combine key attributes of the ballot (like precinct and style)
with letter codes indicating the ballot choice in each contest. The combined codes create
a pattern that gives us considerable information in one alpha-numeric string about each
ballot. If there were say 35 ballots in a given precinct with exactly the same candidate

1
Analysis of the election records and identification of duplicate ballots continues. At the time of this
writing 3,921 duplicate ballots have been identified.
 5

choices made we can verify that there should be 35 votes for each of the candidates
chosen, coming from that particular pattern of choices made, and for the purpose of
counting votes for candidates, and of matching ballots in the original with ballots in the
recount.

This would be 35 instances of the "signature" for each such ballot. It is only necessary to
match the count for each pattern (signature); if there are 35 instances of that pattern in
that precinct, style, and counting group, in the original, but only 34 in the recount, then
one of those ballot records is missing from the JSON files I am analyzing, and each
candidate chosen in the original ballots has had their vote total diminished by one. The
individual ballot content indices, the "signatures," are a fundamental analytical tool in
detecting duplicate and ballots that cannot be matched from the original to the recount
data.

28. By creating "signatures" for each ballot image available, Coalition Plaintiffs'
analysts identified examples of ballot images that appeared to be duplicate and triplicate
images of exactly the same ballot and presented them to me for review. While it is
infeasible to visually review all ballot images, I reviewed a significant number of images
which appear to me to be of duplicates or triplicates of the same ballot. I can confirm
from the cast vote records that these identical ballot images were actually counted in the
tabulation multiple times.

29. This is not a normal expected typical election administration error. It is completely
unacceptable for a system to operate in a manner where widespread double­ and triple-
counting of ballots can occur undetected. Certainly this represents a failure of both the
post election audit and the certification and canvassing process, although we do not
know the root cause of the multiple counts of the same ballots.

33. The possibility of such malware or programming or configuration errors cannot be

ruled out until the paper ballots and related electronic records are reviewed. In my
opinion, it is imperative for Plaintiffs and State officials promptly to determine factual
reasons for these mismatches and mitigate the unacceptable flaw in the voting system and
processes. Whether the cause of the mismatches is physically different ballots being
scanned in the two counts or conflicting electronic interpretation of voters, neither is an
acceptable condition for a public election.

Source: Curling-Duncan-Buell-Declaration-2022-01-22.pdf
Curling v. Raffensperger
Civil Action File No: 1:17-cv-2989-AT
 6

Data Source:
2020 Fulton County Machine Count 2 - Ballot Images
Curling v. Raffensperger
Civil Action File No: 1:17-cv-2989-AT

2020 Fulton County Machine Count 2 - Cast Vote Records (CVRs)

C. No Tabulators Tapes

Upon examining the Batches Loaded Report, Cast Vote Records, and in light of Fulton County's
admission that ten Tabulator Tapes are non-existent, it is evident that 20,713 votes lack the
necessary tabulator tapes detailing their counting process and timing. Within the presidential
contest, of the 20,258 votes for which tapes are missing, 13,274 were cast for Biden and 6,984
for Trump. The required Tabulator Tapes, which should be dated and signed by the responsible
tabulation personnel, are absent, calling into question the legitimacy of these votes, as evidenced
by the Cast Vote Records as required by Georgia State Law.

Batches Loaded Report From Machine Count 1

of Tabulators that Do Not Have Poll Tapes

Ballots Tabulator Tabulator Loaded

Name ID Date Time

133 AV-State Farm Arena ICP 3 303 11/9/2020 9:11:20 AM

198 AV-State Farm Arena ICP 10 311 11/9/2020 9:11:51 AM

558 AV-State Farm Arena ICP 11 312 11/9/2020 9:11:51 AM

3,377 AV-So Fulton Srvc Center ICP3 712 11/3/2020 10:11:22 PM

2,252 AV-Wolf Creek Library ICP4 714 11/4/2020 1:56:39 AM

4,216 AV-Park Place at Newtown ICP3 724 11/3/2020 7:59:33 PM

2,511 AV-Northeast Library ICP3 727 11/4/2020 1:57:07 AM

1,830 AV-Ponce De Leon Library ICP3 754 11/3/2020 8:31:23 PM

1,396 East Point Library ICP3 763 11/3/2020 8:39:58 PM

4,242 AV-Johns Creek ENV Campus ICP 2 2535 11/4/2020 1:53:39 AM

= 20,713 No Tabulator Tapes (C)

 7

Analysis:
Fulton County Open Records Request for Machine Count 1 provided 138 Tabulator Poll tapes,
yet 148 Tabulators were reflected in the official results via the Batches Loaded Report and Cast
Vote Records. The results from 10 unaccounted tabulators were discerned through an analysis of
the Cast Vote Records.
Machine Count 1 - Candidate Totals for President
From Tabulators without Poll Tapes from Cast Vote Records.

Tabulator ID Biden Trump

303 125 8

311 183 15

312 496 51

712 3,109 216

714 2,099 109

724 1,676 2,427

727 1,072 1,373

754 1,407 369

763 1,234 122

2535 1,873 2,294

Total 13,274 6,984

Grand Total 20,258

Supporting Reference:
In Moncla Rossi SEB Complaint; Case No. SEB-2023-025:

In addition to the problems detailed above, the original Election Night vote count
includes results for ten (10) Advance Voting tabulators for which Fulton County has no
records. That is, the tabulators do not exist – there are no poll open tapes, no daily status
tapes and no poll closing tapes.. We submitted Open Records Requests to Fulton County
 8

specifically seeking the 10 tabulator tapes, but Fulton County responded by saying that
they had “No such records”.

To follow up, we sent two emails to the Fulton County Records Department and the
Fulton County Custodian of Records, Steve Rosenberg, seeking clarification to determine
if the records were missing or if they exist; the Records Department replied, “The
records do not exist.” See attached emails and official certification of records.

These ten (10) tabulator tapes total 20,713 votes, all of which were included in Election
Night results:

Source: Moncla Rossi SEB Complaint; Case No. SEB-2023-025 - Page 10

Two emails to the Fulton County Records Department and the Fulton County Custodian
of Records, Steve Rosenberg, seeking clarification to determine if the records were
missing or if they exist; the Records Department replied, “The records do not exist.” See
attached emails and official certification of records from Feb 4, 2022.

Source: Moncla Rossi SEB Complaint; Case No. SEB-2023-025 - Page 19

Data Source:

2020 Fulton County Machine Count 1 - Batches Loaded Report

2020 Fulton County Machine Count 1 - Cast Vote Records

List of Poll Tapes that were provided to determine which Poll Tapes were missing.
ORR #: R008635-1201121 - Received 01/02/2022
Request: All Tabulator Poll Tapes for the 11/3/2020 General Election, Advanced Voting
File: 671461587-Fulton-2020-Av-Poll-Tapes-Certified.pdf

D. Compromised Ballot Authentication: Missing and Deleted SHA Files

Approximately 132,842 ".sha" files from Fulton County's 2020 Machine Count 1 were deleted.
These files are automatically generated, producing a unique identifier for each ballot image to
authenticate the ballots. A SHA file, a variation of MD5, is used for data hashing to ensure the
ballot image remains unaltered.
 9

2020 Fulton County Machine Count 1 - Fulton Absentee by Mail Ballot Files Analysis

File Type Count

Ballot Images (.tif) 148,876

Secure Hashing Algorithm (.sha) 16,034

Ballot Images with missing SHA Files 132,842

In the 2020 Fulton County Machine Count 2, Ducan Buell highlighted that approximately 17,800
ballot images were missing their corresponding ".sha" files.

Supporting Reference:

In Duncan Buell’s Declaration in the Curling vs. Raffensperger:

16. Coalition Plaintiff's attorneys' correspondence indicates the estimated

number of electronic records missing. In Fulton County ballot images and their attached
"AuditMark" tabulation interpretations are missing for approximately 360,000 BMD
ballots and 6.4 million BMD votes they contain for in-person voting for the official
certified count of all down ballot races. Approximately 17,800 images and the same
number of .sha files are missing from the presidential recount. Approximately 11,000
.dvd files are missing from the presidential recount for Fulton's November 2020 election.
Such electronic files are essential when testing the election records for consistency and
accuracy. Significant numbers of missing files inhibits a complete authoritative analysis
of causes or effects of potential impacts of malware, software bugs, equipment
malfunction, human error or malfeasance by insiders.

20. The significance of the missing and unverified files cannot be overstated in
the limitations they place on expert analysis. For example, while I worked extensively
with Fulton's cast vote records, roughly 34% of the ballot images and often the .dvd and
.sha related files were simply missing, preventing me from consulting those files for
confirmation of the fidelity of the data with the image and its related files. It is my
understanding that while Coalition Plaintiffs desire that I analyze Cobb County's cast
vote records and other electronic records, a full 50% of the required records have not
been preserved by Cobb County.
 10

Source: Curling-Duncan-Buell-Declaration-2022-01-22.pdf
Curling v. Raffensperger
Civil Action File No: 1:17-cv-2989-AT

Data Source:
2020 Fulton County Machine Count 1 - Fulton Absentee by Mail Ballot Files
File: Fulton-absentee-by-mail.7z
Curling v. Raffensperger
Civil Action File No: 1:17-cv-2989-AT
 IN THE UNITED STATES DISTRICT COURT FOR
THE NORTHERN DISTRICT OF GEORGIA
ATLANTA DIVISION

)
DONNA CURLING, et al. )
)
Plaintiff, )
) CIVIL ACTION FILE NO.:
vs. ) 1:17-cv-2989-AT
)
BRAD RAFFENSPERGER, et al. )
)
Defendant. )
)
)

DECLARATION OF PHILIP B. STARK

PHILIP B. STARK hereby declares as follows:

1. This statement supplements my declarations of September 9, 2018; September 30,

2018; October 22, 2019; December 16, 2019; August 23, 2020; August 31, 2020;

September 13, 2020; and August 2, 2021. I stand by everything in the previous

declarations and incorporate them by reference. This declaration includes and

augments a declaration submitted on 11 January 2022. Aside from adding the

italicized material in this paragraph and correcting minor typographical errors, the

differences between the previous version and this version are confined to (new)

paragraphs 58 through 84. Paragraphs 85 through 89 were in the earlier version but

have been renumbered. Appendix 1 has been updated to the current version of my CV.

Appendices 2, 3, and 4 are unchanged. Appendices 5 through 9 are new.

Qualifications and Background
2. I am Professor of Statistics at the University of California, Berkeley, where I am also a

faculty member in the Graduate Program in Computational Data Science and

Engineering; a co-investigator at the Berkeley Institute for Data Science; principal

investigator of the Consortium for Data Analytics in Risk; director of Berkeley Open

Source Food; and affiliated faculty of the Simons Institute for the Theory of

Computing, the Theoretical Astrophysics Center, and the Berkeley Food Institute.

Previously, I was Associate Dean of Mathematical and Physical Sciences, Interim

Regional Associate for the College of Chemistry and the Division of Mathematical

and Physical Sciences, Chair of the Department of Statistics, and Director of the

Statistical Computing Facility.

3. I have published more than two hundred articles and books. I have served on the

editorial boards of archival journals in physical science, Applied Mathematics,

Computer Science, and Statistics. I currently serve on three editorial boards. I have

lectured at universities, professional societies, and government agencies in thirty

countries. I was a Presidential Young Investigator and a Miller Research Professor. I

received the U.C. Berkeley Chancellor’s Award for Research in the Public Interest, the

Leamer-Rosenthal Prize for Open Social Science, and a Velux/Villum Foundation

Professorship. I am a member of the Institute for Mathematical Statistics and the

Bernoulli Society. I am a Fellow of the American Statistical Association, the Institute

of Physics, and the Royal Astronomical Society. I am professionally accredited as a

statistician by the American Statistical Association and as a physicist by the Institute

of Physics.

2
 4. I have consulted for many government agencies, including the U.S. Department of

Justice, the U.S. Department of Agriculture, the U.S. Department of Commerce, the

U.S. Department of Housing and Urban Development, the U.S. Department of

Veterans Affairs, the Federal Trade Commission, the California Secretary of State, the

California Attorney General, the California Highway Patrol, the Colorado Secretary of

State, the Georgia Department of Law, the Illinois State Attorney, the New Hampshire

Attorney General, and the New Hampshire Secretary of State. I currently serve on the

Board of Advisors of the U.S. Election Assistance Commission and its Cybersecurity

Subcommittee. (The opinions expressed herein are, however, my own: I am not

writing as a representative of any entity.)

5. I have testified before the U.S. House of Representatives Subcommittee on the

Census; the State of California Senate Committee on Elections, Reapportionment and

Constitutional Amendments; the State of California Assembly Committee on Elections

and Redistricting; the State of California Senate Committee on Natural Resources; and

the State of California Little Hoover Commission.

6. I have been an expert witness or non-testifying expert in a variety of state and federal

cases, for plaintiffs and for defendants, in criminal matters and a range of civil matters,

including, inter alia: truth in advertising, antitrust, construction defects, consumer

class actions, credit risk, disaster relief, elections, employment discrimination,

environmental protection, equal protection, fairness in lending, federal legislation,

First Amendment, import restrictions, insurance, intellectual property, jury selection,

mortgage-backed securities, natural resources, product liability class actions, qui tam,

3
 risk assessment, toxic tort class actions, trade secrets, utilities, and wage and hour class

actions.

7. I have been qualified as an expert on statistics in federal courts, including the Central

District of California, the Northern District of Georgia, the District of Maryland, the

Southern District of New York, and the Eastern District of Pennsylvania.

8. I have also been qualified as an expert on statistics in state courts.

9. I have used statistics to address a wide range of questions in many fields.1

10. I served on former California Secretary of State Debra Bowen’s Post-Election Audit

Standards Working Group in 2007.

11. In 2007, I invented a statistical approach to auditing elections (“risk-limiting audits,”

referred to below as “RLAs”) that has been incorporated into statutes in California

(AB 2023, SB 360, AB 44), Colorado (C.R.S. 1-7-515), Rhode Island (RI Gen L §17-

19-37.4 (2017)), Virginia (Code of Virginia 24.2-671.1), and Washington (RCW

29A.60.185), and which are in pending federal legislation (the PAVE Act of 2018 and

S.1 of 2021). My election auditing methods have been used in roughly 20 U.S. States

and in Denmark. (The State of Georgia has piloted some RLA procedures, but has not

conducted an actual RLA, as I explain below.)

12. RLAs are widely viewed as the best way to check whether the reported winner(s) of an

election really won. They have been endorsed by the Presidential Commission on

1
For example, I have used statistics to analyze the Big Bang, the interior structure of the Earth
and Sun, earthquake risk, the reliability of clinical trials, the accuracy of election results, the
accuracy of the U.S. Census, the risk of consumer credit default, food safety, the causes of
geriatric hearing loss, the effectiveness of water treatment, sequestration of carbon in agricultural
soils, the fragility of ecological food webs, risks to protected species, the effectiveness of
Internet content filters, high-energy particle physics data, and the reliability of models of climate,
among other things.

4
 Election Administration; the U.S. National Academies of Sciences, Engineering, and

Medicine; the American Statistical Association; the League of Women Voters;

Verified Voting Foundation; Citizens for Election Integrity Minnesota; and other

groups concerned with election integrity.

13. I have worked closely with state and local election officials in California and Colorado

to pilot and deploy RLAs. The software Colorado uses to conduct RLAs is based on

software I wrote. All of the genuinely risk-limiting methods in VotingWorks “Arlo”

software used by the State of Georgia were invented by me.2

14. I worked with Travis County, Texas, on the design of STAR-Vote, an end-to-end

cryptographically verifiable voting system.

15. I testified as an expert witness in the general area of election integrity, including the

reliability of voting equipment, in 2016 presidential candidate Jill Stein’s recount suit

in Wisconsin, and filed a report in her suit in Michigan.

16. I have testified as an expert in election auditing and the accuracy of election results in

two election-related lawsuits in California.

17. I have testified to both houses of the California legislature regarding election integrity

and election audits. I have testified to the California Little Hoover Commission about

election integrity, voting equipment, and election audits.

18. I have advised the election commissions of Denmark, Mongolia, and Nigeria on issues

related to election integrity, security, and audits.

2
Arlo also implements a method that is not risk-limiting in practice.

5
 19. I was a member of the three-person team that conducted a statutory forensic audit of

the State Representative contest in Windham, NH, in 2021.3

20. Since 1988, I have taught statistics at the University of California, Berkeley, one of the

top two statistics departments in the world (see, e.g., QS World University Rankings,

2014) and the nation (US News and World Reports, 2018). I teach statistics regularly

at the undergraduate and graduate levels. I have created five new statistics courses at

Berkeley. I developed and taught U.C. Berkeley’s first for-credit online course in any

subject, and among the first approved for credit throughout the ten campuses of the

University of California system. I also developed and co-taught online statistics

courses to over 52,000 students, using an online textbook and other pedagogical

materials I wrote and programmed.

21. Appendix 1 is my current curriculum vitae, which includes my publications for the last

ten years and all cases in the last four years in which I gave deposition or trial

testimony.

Opinions

22. I have been asked to assess whether the State of Georgia’s current Dominion Ballot

Marking Device (“BMD”) voting system and the protocols for its use—including

audits—provides reasonable assurance that voters’ selections will be counted, and

counted as cast. The answer is a clear “no.”

The 2020 “Audit”

23. Georgia Secretary of State Brad Raffensperger has claimed, referring to the post-

election audit of the November 3, 2020 presidential contest, “Georgia’s historic first

3
See https://www.doj.nh.gov/sb43/index.htm, last accessed 8 January 2022.

6
 statewide audit reaffirmed that the state’s new secure paper ballot voting system

accurately counted and reported results.”4 And “[] we did a 100 percent risk-limiting

audit with a hand recount which proved the accuracy of the count and also proved that

the machines were accurately counting it, and that no votes were flipped.”5

VotingWorks Executive Director Ben Adida claimed “Georgia’s first statewide audit

successfully confirmed the winner of the chosen contest and should give voters

increased confidence in the results.”6 Per the official report of the audit, “The audit

confirmed the original result of the election, namely that Joe Biden won the

Presidential Contest in the State of Georgia. The audit [] provides sufficient evidence

that the correct winner was reported.”7 I shall explain why these claims about the audit

are false.

24. There are many things the audit did not check (including the outcome), and the thing it

was positioned to check—the tabulation of validly cast ballots—was not checked

properly, as data from the audit itself show.

25. I shall start by listing some things the audit did not check. My statements are true and

correct to the best of my knowledge, and they are consistent with the audit

documentation available at the Secretary of State’s website at the URL

https://sos.ga.gov/index.php/elections/2020_general_election_risk-limiting_audit (last

accessed 9 January 2022).

4
https://sos.ga.gov/index.php/elections/historic_first_statewide_audit_of_paper_ballots_upholds_
result of presidential race, last accessed 9 January 2022
5
https://www.effinghamherald.net/local/raffensperger-spread-election-misinformation-
bipartisan-endeavor/ last accessed 9 January 2022.
6
Ibid.
7
https://sos.ga.gov/admin/uploads/11.19_.20_Risk_Limiting_Audit_Report_Memo_1.pdf. last
accessed 9 January 2022

7
 26. The audit did not check whether BMDs correctly printed voters’ selections. No audit

can check that, as I have previously declared. (As a consequence, Secretary

Raffensperger has no basis to assert that no votes were flipped.) The declarations and

testimony of Prof. J. Alex Halderman establish that BMDs can be hacked,

misprogrammed, or misconfigured to print votes that differ from voters’ selections as

confirmed onscreen or through audio. As Prof. Andrew Appel has testified and as

elaborated in my declarations, only voters are in a position to check—but few do, and

those who do check generally check poorly. To the best of my knowledge, the State of

Georgia has no procedures in place to log, investigate, or report complaints from

voters that BMDs altered votes, so it is not clear whether any voters did notice

problems. My previous declarations also explain why logic and accuracy testing can

never be adequate to establish that BMDs behave correctly in practice.8

27. The audit did not check whether every validly cast ballot was scanned exactly once.

The audit could not check whether every validly cast ballot was scanned, because

Georgia’s rules for ballot accounting, pollbook and voter participation reconciliation,

physical chain of custody, etc., are not adequate to ensure that every cast ballot is

accounted for.

28. The audit did not check whether every memory card used in the election was

accounted for, nor whether every memory card containing votes was uploaded to a

8
See, e.g., Stark, P.B. and R. Xie, 2019. Testing Cannot Tell Whether Ballot-Marking Devices
Alter Election Outcomes, ArXiV, https://arxiv.org/abs/1908.08144, last accessed 9 January
2022.

8
 tabulator. The audit found that some had not been,9 but to my knowledge, there has

been no check to confirm there are no other cards with votes outstanding.

29. The audit did not check whether any scans were duplicated, deleted, replaced or

altered.

30. The audit did not check whether QR code encoding the votes on BMD printout

matches the human-readable selections on any ballot.

31. The audit did not check whether the voting system correctly interpreted any ballot or

BMD printout. (Again, as a consequence, Secretary Raffensperger has no basis to

assert that no votes were flipped.)

32. The audit did not do a very good job of checking the tabulation, as I shall demonstrate.

I focus on Fulton County. I have not investigated other counties, but I have no reason

to believe the problems and errors are confined to Fulton County. I have been told by

Coalition Plaintiffs that similar problems occurred in other counties, but I have not

independently verified their findings.

33. I downloaded the detailed “audit spreadsheet” from the URL

https://sos.ga.gov/admin/uploads/audit-report-November-3-2020-General-Election-

2020-11-19.csv on 9 January 2022.

34. I downloaded images of the Fulton County RLA manual tabulation batch sheets

(“Audit Board Batch Sheets”, ABBSs henceforth) from

https://sos.ga.gov/admin/uploads/Fulton%20RLA%20Batches.zip on 9 January 2022.

That file contains five .pdf files, “Fulton Audit Documents 1_redacted.pdf,” through

“Fulton Audit Documents 4_redacted.pdf,” which contain images of ABBSs, and

9
See notes 13 and 14, infra.

9
 “Fulton Audit Documents 5.pdf” which contains images of “Vote Review Panel Tally

Sheets.”

35. My understanding is that ABBSs are filled in by hand by the counting teams who

counted the votes from the paper ballots (including BMD printout). Each ABBS

reflects the manual tally of votes from one physically identifiable batch of ballots. I

understand that after the ABBSs were filled out, other workers transcribed data from

the ABBSs into VotingWorks audit software “Arlo.” My understanding is that every

ballot validly cast in Fulton County in the 2020 Presidential Election should be

reflected in exactly one ABBS, and data from every ABBS should have been entered

exactly once into the database from which the audit spreadsheet was exported.

36. The four ABBS image files contain 349 pages, 636 pages, 578 pages, and 364 pages,

respectively, a total of 1,927 ABBSs. But the audit spreadsheet contains only 1,916

rows of data for Fulton County. It appears that at least eleven ABBSs are entirely

missing, not counting possible duplicate entries in the spreadsheet.10 This sort of

“sanity check” is simple to perform, but apparently was not performed by the auditors,

the County, or the Secretary of State.

37. Many ABBSs were not completely filled in. The “Batch Type,” signifying the mode of

voting (absentee, election day, advance) was often blank, and many numbers were

blank, presumably intended to denote zeros.

10
However, I did see at least one ABBS marked “Dup” (presumably meaning “duplicate”) for
instance, page 11 of “Fulton Audit Documents 2_redacted.pdf.” However, as the table after
paragraph 38, supra, shows, there are at least 11 ABBSs that are not accounted for in the audit
spreadsheet. Thus, there are presumably duplicated entries in the audit spreadsheet.

10
 38. Coalition Plaintiffs have identified a sample of at least eleven ABBSs for Fulton

County that do not appear in the audit spreadsheet, and I have verified their work. The

software I wrote for that purpose is in Appendix 2.

39. The following table lists these examples; the final column indicates which page of

which ABBS image file contains the image (for instance, “4 at 162” means page 162

of “Fulton Audit Documents 4_redacted”). The fact that the vote data in the last two

rows are identical is suspicious, but the corresponding ABBS images are clearly

different; see Appendix 3. Regardless, neither appears in the audit spreadsheet.

Scanner Batch Mode of Trump Biden Jorgensen Write-In Undervote Overvote Image
voting or blank source
1 3 48 absentee 4 93 2 0 0 0 4 at 162
2 2 52 absentee 6 92 0 0 0 0 1 at 1
3 3 12–14 ? 12 83 1 0 0 0 4 at 128
4 3 239 ? 13 87 0 0 0 0 3 at 177
5 1 80–84 ? 118 329 3 2 2 1 3 at 519
6 3 260 absentee 30 66 0 0 0 0 4 at 355
7 AP01A-1 election day 84 62 6 2 1 0 1 at 170
8 3 179–181 absentee 85 224 5 1 2 0 4 at 293
9 2 239 absentee 4 42 0 0 0 0 2 at 153
10 Chastain 12 advance 613 605 24 7 4 0 3 at 351
11 Chastain 114 advance 613 605 24 ? 4 0 3 at 270

40. I searched the audit spreadsheet for tallies that matched the numbers in these missing

ABBSs. There are no data in the audit spreadsheet matching rows 4–11 of the table.

There are data that match rows 1, 2, and 3, but with distinctively different batch

identifiers.11 It is plausible that these are genuinely different batches, and I have no

reason to believe otherwise: some identical counts in different batches are to be

11
The data that match row 1 are identified as “Scanner 3 Ballot [sic] 162” rather than batch 48.
The data that match row 2 are identified as “Absentee Scanner 2 Batch 400” rather than batch 52.
The data that match row 3 are identified as Absentee Scanner 3 Batch 253 rather than batches
12–14.

11
 expected. Indeed, in the entire audit spreadsheet, there are 16,807 rows that duplicate

other ABBS vote counts within the same county, out of a total of 41,881 rows.

41. I checked vote totals for Donald J. Trump, Joseph R. Biden, and Jo Jorgensen derived

by summing ABBS entries in the audit spreadsheet against the vote totals in the

summary audit result spreadsheet posted by the Secretary of State at the URL

https://sos.ga.gov/admin/uploads/Georgia%202020%20RLA%20Report.xlsx, which I

downloaded on 9 January 2022. (The spreadsheet does not list write-ins, undervotes,

or overvotes.) Both show Trump receiving 137,620 votes, Biden receiving 381,179,

and Jorgensen receiving 6,494. Thus, the ABBSs that are missing from the audit

spreadsheet are also missing from the audit’s reported vote totals.

42. On the assumption that the ABBSs—the original source of the manual tally data

entered into the audit spreadsheet—are correct, the omission of that sample of 11

ABBSs deprived Trump of 1,582 votes, Biden of 2,288, and Jorgensen of 65, not to

mention write-ins. This sample alone has a total of over 3,900 votes that the audit

tabulated but were not included in the audit’s reported vote totals.

43. The original tabulation in Fulton County showed 524,659 votes; the reported audit

results showed 525,293, a difference of 634 votes, about 0.12 percent.12 Accounting

for those 11 omitted ABBSs increases the apparent error of the first count from 634

votes to over 4,569 votes or 0.87 percent, far larger than the statewide margin of

12
Data from https://sos.ga.gov/admin/uploads/Georgia%202020%20RLA%20Report.xlsx, last
accessed 9 January 2022.

12
 victory. It is also larger than 0.73 percent, which Secretary of State Raffensperger

claimed was the maximum miscount in any Georgia county.13

44. However, there is no way to know whether including that sample of 11 ABBSs would

make the audit tabulation a complete count of the votes in Fulton County. That is

because Georgia’s canvass is inadequate: many ballots might still remain untabulated.

The proof that at least some of Georgia’s jurisdictions do not keep adequate track of

ballots, memory cards, and other election materials is reflected in the fact that

thousands of ballots and scans were “discovered” during the audit.14 There is no

trustworthy inventory of ballots to check the results against, because of Georgia’s lax

canvass.

45. Governor Brian P. Kemp has pointed out similar flaws in the audit, saying the audit

report was “sloppy, inconsistent, and presents questions about what processes were

used by Fulton County to arrive at the result.”15 Governor Kemp’s letter points out that

13
Per Secretary Raffensperger, “[i]n Georgia’s recount, the highest error rate in any county
recount was 0.73%.” https://sos.ga.gov/index.php/elections/2020 general election risk-
limiting audit, last accessed 9 January 2022.
14
https://www.cbs46.com/news/floyd-county-election-director-fired-after-audit-reveals-2-600-
votes-went-uncounted/article_bbd08d90-2aa2-11eb-9e4d-bf96ac56ad54.html, last accessed 10
January 2022. https://www.news4jax.com/news/georgia/2020/11/18/4th-georgia-county-finds-
uncounted-votes-as-hand-count-deadline-approaches/, last accessed 10 January 2022.
https://www.mdjonline.com/elections/cobb-elections-finds-350-uncounted-ballots-during-
audit/article_0d93e26e-22bd-11eb-8bce-17067aceee33.html, last accessed 10 January 2022.
https://www.11alive.com/article/news/politics/elections/fayette-county-election-results-ballots-
uncovered-during-audit/85-f79dd838-a15c-4407-80b2-9dfbc2466188, last accessed 10 January
2022.
15
Letter from Brian P. Kemp, Governor, to the Georgia State Election Board, dated 17
November 2021, addressing the work of Mr. Joseph Rossi; Review of Inconsistencies in the Data
Supporting the Risk Limiting Audit Report, Office of Governor Brian P. Kemp, 17 November
2021. These documents are attached hereto as Appendix 4.

13
 the audit data include duplicated entries, which I understand Coalition Plaintiffs have

verified. I have not tried to verify those findings.

First Count, Audit, and Recount Differ Substantially

46. I understand that Plaintiff Donna Curling votes in Fulton County precinct RW01. On

10 January 2022, I downloaded the official precinct-level results for the original

tabulation from

https://results.enr.clarityelections.com//GA/Fulton/105430/271723/reports/detailxls.zi

p and for the recount from

https://results.enr.clarityelections.com//GA/Fulton/107292/275183/reports/detailxls.zi

p to examine the results in that precinct.

47. The following table shows the counts of election-day votes in Fulton County precinct

RW01 for the three presidential candidates, according to the original machine count,

the machine recount, and the “audit,” and vote-by-mail and advance votes for the

original election and the recount. (The audit did not report precinct-level results for

vote-by-mail or advance voting.)

Count Election Day Advance Absentee by Mail Provisional

Trump Biden Jorgensen Trump Biden Jorgensen Trump Biden Jorgensen Trump Biden Jorgensen
Original 193 88 11 1455 1003 23 619 833 15 9 4 1
Recount 162 73 9 1487 1015 25 619 809 15 5 3 1
Audit 243 88 11

48. There are large, unexplained differences among these results.16 I do not see how

Plaintiff Donna Curling can have reasonable confidence that her vote was counted at

all, much less counted as cast.

16
There appears to be some cancellation of error, but I understand that the hand count kept
ballots cast in different ways (advance in-person, absentee by mail, and election day) separate. It

14
 49. The Secretary of State attributed all differences between the audit and the original

count to human counting error, citing a 2012 study that found hand-count error rates as

high as 2 percent.17 This is simplistic, unfounded, and disingenuous.

50. While human error almost certainly accounts for some of the difference, there is no

evidence that it accounts for most of the difference, much less the entire difference, as

Secretary of State Raffensperger claimed.

51. The original count and audit agree with each other (but not with the recount) regarding

the number of election-day votes for Biden and Jorgensen. The audit found 50 more

election-day votes for Trump than the original tally, and 81 more than the machine

recount found: a difference of almost 50 percent. These differences have not been

investigated and are unexplained. A hypothesized error rate of 2 percent in hand

counts does not suffice.

52. A fact central to this case is that the differences might result from discrepancies

between the QR-encoded votes and the human-readable votes on BMD printout and/or

from misconfiguration, bugs, or malware on the scanners or tabulators. As discussed

above, the audit checked none of these things. There is no basis whatsoever to

conclude that the differences result entirely from human error without investigating the

other possibilities.

is not clear how misclassification of the mode of voting would affect one candidate’s totals much
more than the other candidates. Regardless, these discrepancies are large and should be
investigated, including inspecting the physical ballots.
17

https://sos.ga.gov/index.php/elections/historic_first_statewide_audit_of_paper_ballots_upholds_r
esult_of_presidential_race, last accessed 10 January 2022.

15
 53. The hand count could easily be more accurate than the machine count. Indeed, it is

well known that hand counts of hand-marked paper ballots are often more accurate

than machine counts, in part because human readers can interpret light, improper, and

ambiguous marks better than machines can, even when the machines are working

properly. Similarly, experience in Georgia in 2020 shows that Dominion’s scanner

settings (low resolution, black-and-white) can cause voters’ selections not to appear at

all in images of ballots, selections that human readers looking at the actual ballots can

easily discern.18

54. Evidence that hand counts are more accurate than machine counts comes from

recounts and studies of the “residual vote,”19 the number of undervotes and overvotes.

Hand counts generally find more valid votes than machine counts.20

55. Hand-count error rates are known to depend on many factors, including ballot design,

the method for hand counting (“sort-and-stack” versus “read-and-mark”), and the size

of each counting team. They presumably also depend on whether there are additional

18
See, e.g., Judge Amy Totenberg’s Opinion and Order of 11 October 2020 in the present
matter, at 4, 30, 95, 101, 103, 114–135.
19
Ansolabehere, S., and Reeves, A., 2004. Using Recounts to Measure the Accuracy of Vote
Tabulations: Evidence from New Hampshire Elections 1946–2002, in Confirming Elections:
Creating Confidence and Integrity Through Election Auditing, Alvarez, R.M., L.R. Atkeson, and
T.E. Hall, eds., Palgrave MacMillan, NY. Alvarez, R.M., D. Beckett, D., and C. Stewart, 2013.
Voting Technology, Vote-by-Mail, and Residual Votes in California, 1990–2010. Political
Research Quarterly, 66(3), 658–670. https://doi.org/10.1177/1065912912467085. Alvarez, R.M.,
L.R. Atkeson, and T.E. Hall, 2013. Evaluating Elections: A Handbook of Methods and
Standards, Cambridge University Press, NY.
20
See, e.g., Ansolabehere, S., and C. Stewart, 2005. Residual Votes Attributable to Technology.
The Journal of Politics, 67(2), 365–389. https://doi.org/10.1111/j.1468-2508.2005.00321.x;
Carrier, M.A., 2005. Vote Counting, Technology, and Unintended Consequences, St. John’s Law
Review, 79(3), 645–687; Ansolabehere, S., B.C. Burden, K.R. Mayer, and C. Stewart III, 2018.
Learninbg from Recounts, Election Law Journal, 17(2), 100–116, DOI: 10.1089/elj.2017.0440

16
 quality control measures in place, such as checking sorted piles of ballots to ensure

that each pile really has votes for just one candidate.

56. The study21 cited by the Georgia Secretary of State is a laboratory study with 108

subjects and 120 ballots, each containing 27 contests with two candidates. It used three

kinds of “ballots”: printout from two kinds of DRE (direct-recording electronic) voting

system and an optical scan ballot. The highest error rates were for thermal printout

from DREs, which does not resemble Georgia’s BMD printout nor Georgia’s hand-

marked paper ballots. The method with the highest error was the “sort-and-stack” tally

method that Georgia chose to use in its audit. This study did not observe hand vote

tabulation in a real election, nor did it involve BMD summary printout. To my

knowledge, there is no study of the accuracy of counting votes from BMD summary

printout.

57. Differences between the original count and the machine recount are also large and

unexplained. The difference between the two machine counts of Biden’s Absentee

votes is almost 3 percent. Absent access to the physical ballots, software, and

equipment, it is impossible to know what went wrong, nor whether the differences are

primarily attributable to malware, bugs, misconfiguration, or human error.

Pursuant to the Court’s Order (Order Doc. 1322) permitting the supplementation of this

report in the context of my engagement concerning the processes, adequacy, and quality of

21
Goggin, S.N., M.D. Byrne, and J.E. Gilbert, 2012. Post-Election Auditing: Effects of
Procedure and Ballot Type on Manual Counting Accuracy, Efficiency, and Auditor Satisfaction
and Confidence, Election Law Journal: Rules, Politics, and Policy, 36–51, DOI:
10.1089/elj.2010.0098

17
Georgia’s audit procedures, Paragraphs 58–84, infra, were added in the 9 March 2022

version.

The two machine counts in Fulton County

58. I examined the internal consistency of the two machine counts (the original machine

count and the machine recount) in Fulton County. I relied on election data provided in

electronic form by Coalition Plaintiffs. I understand those data to be the election

management system data for Fulton County for the two machine counts. A declaration

from Marilyn Marks attesting to the provenance of the data is attached hereto as

Appendix 5. The data include cast vote records, scanned images of ballots and BMD

printout, and other files.

59. I analyzed the Fulton County election data using software I wrote, attached hereto as

Appendix 6. I also relied on two spreadsheets provided by the Coalition Plaintiffs.

Those spreadsheets purport to identify groups of images (among the Fulton County

election materials) that appear to be repeated images of the same pieces of paper. I do

not know in detail how those spreadsheets came to exist—but as described below, I

checked the accuracy of those spreadsheets as part of this report.

60. To confirm that I had received the correct Fulton County election data from Coalition

Plaintiffs and that I was reading it correctly, I counted the votes for Donald J. Trump,

18
 Joseph R. Biden, and Jo Jorgensen. For both machine counts, I found the same totals

officially reported for Fulton County:22

Candidate First machine count Second machine count

Donald J. Trump 137,240 137,247
Joseph R. Biden 381,144 380,212
Jo Jorgensen 6,275 6,320

61. The number of cast vote records (the voting system’s record of the votes on each ballot

or BMD printout card, from which the system tabulates results) in the two machine

counts in Fulton County were rather different: 528,776 in the first count and 527,925

in the second count, a difference of 851. To my knowledge, Fulton County has not

explained this discrepancy.

62. The number of cast vote records in the two machine counts should be equal.

Differences might occur if (i) some ballots or BMD printout cards were misplaced or

found between the two machine counts, so a different number pieces of paper was

scanned in the two machine counts; (ii) malware, bugs, misconfiguration, or a bad

actor added, deleted, or altered records in the election management system in one or

both machine counts; (iii) Fulton County did not scan every validly cast ballot or

BMD printout card exactly once in each machine count. Below, I present compelling

evidence that (ii) or (iii) is true, but all three possibilities could be true simultaneously.

In particular, without further discovery, it is impossible to rule out any of the

possibilities.

22
First machine count results:
https://results.enr.clarityelections.com/GA/Fulton/105430/web.264614/#/summary (last visited 8
March 2022) Second machine count results:
https://results.enr.clarityelections.com/GA/107231/web.264614/#/detail/5000?county=Fulton
(last visited 8 March 2022)

19
 63. Fulton County did not produce the image file corresponding to every cast vote record.

For the first machine count, production included images of ballots or BMD printout

cards for only 168,726 of the 528,776 cast vote records: 376,863 image files are

missing. For the second machine count, Fulton County’s production included images

of ballots or BMD printout cards for 510,073 of the 527,925 cast vote records: 17,852

image files are missing.

64. Entire batches of images are missing from Fulton County’s production, for example,

images from Scanner 801 batch 117 and Scanner 801 batch 118 are referred to in the

cast vote records for the second machine count but the images were not among the

electronic records. Without additional discovery it is impossible to determine whether

the missing images are missing because of human error, programming errors (bugs), or

malware in Fulton County’s election management system (EMS). Of course, those

possibilities are not mutually exclusive.

65. It is nonetheless possible to use the produced images to show that Fulton County’s

election results included many votes more than once in the reported tabulations. The

full extent of this multiple-counting problem cannot be determined without additional

discovery, but there is ample evidence that it added thousands of bogus votes to the

reported machine-count results. That is, thousands of Fulton County voters’ votes were

included in the reported totals more than once. From the production so far, it is not

possible to determine conclusively whether any voter’s votes were omitted from the

reported totals.

66. I now describe how I established that some votes were included in the reported totals

more than once.

20
 67. Repeatedly scanning the same piece of paper generally does not produce images that

are bitwise identical, because of variations in the alignment of the paper, illumination

within the scanner, dirt on scanner lenses, etc. Similarly, a single scan can be altered

digitally to produce multiple images that look similar but are not bitwise identical.

68. Small variations in voters’ marks (e.g., not filling an oval completely or straying

outside the oval) on hand-marked paper ballots generally make it possible to tell

whether two separate scans of hand-marked paper ballots that contain the same votes

are scans of the same physical ballot.

69. It is not generally possible to tell whether two 200dpi black-and-white scans of BMD

printout cards are scans of the same piece of paper simply by looking at those two

scans, because BMD printout cards containing the same votes look the same at low

resolution in black-and-white.23 However, if both scans contain a rare write-in name or

rare combination of write-in names, that is evidence of a duplicate. Similarly, if a

series of votes is repeated in in the same order (or reverse order) in different scan

batches of BMD printout, that is also evidence that they are repeated images of the

same collection of paper. If the duplicated (or reversed) vote sequences are long and

include rare write-in names, the evidence that they are scans of the same physical

pieces of paper is particularly compelling.

70. As mentioned in paragraph 46, supra, I understand that plaintiff Donna Curling votes

in Fulton County precinct RW01. In one of the spreadsheets mentioned in paragraph

23
A sufficiently high-resolution scan might make it possible to identify differences in the
arrangement of the paper fibers. See W. Clarkson, T. Weyrich, A. Finkelstein, N. Heninger, J. A.
Halderman and E. W. Felten, 2009. Fingerprinting Blank Paper Using Commodity Scanners,
2009 30th IEEE Symposium on Security and Privacy, 301–314, doi: 10.1109/SP.2009.7

21
 58, supra, Coalition Plaintiffs identified 12 hand-marked ballots from Fulton County

precinct RW01 that were scanned twice in the first machine count (the original

election). The pairs of images are listed in the table below. The format of the numbers

is

[scanner number]_[batch number]_[image number].

pair Image A Image B

1 05162 00234 000096 05162 00235 000057
2 05162 00234 000093 05162 00235 000054
3 05162 00234 000074 05162 00235 000036
4 05162 00234 000072 05162 00235 000034
5 05162 00234 000068 05162 00235 000030
6 05162 00234 000069 05162 00235 000031
7 05162 00234 000054 05162 00235 000014
8 05162 00234 000031 05162 00235 000090
9 05162 00234 000026 05162 00235 000085
10 05162 00234 000017 05162 00235 000076
11 05162 00234 000013 05162 00235 000072
12 05162 00234 000014 05162 00235 000073
13 05162 00234 000003 05162 00235 000062
14 05162 00234 000001 05162 00235 000060

71. I wrote a program to display ballot images of ballots side by side to check whether

they look the same. The software is in Appendix 6. Appendix 7 shows these 14 pairs

of repeated images. I confirmed that they are indeed duplicated scans by visually

matching slight irregularities in the voters’ marks in each pair.

72. Coalition Plaintiffs identified at least three BMD cards from precinct RW01 that each

appear to have been scanned twice in the machine recount in RW01, based on the

votes and the order in which they were scanned in two batches. In particular, Scanner

801, batches 43 and 44—both comprising scans of advance in-person BMD printout

cards—start with images of 214 BMD cards that appear to be the same in both

batches: the same sets of votes in the same order. The two batches were scanned

within about five minutes of each other, according to the timestamps in the images.

22
 Many of the images show write-in votes24 or votes for third-party candidates, further

evidence that the similarity was no coincidence. I visually inspected25 all 214 pairs and

confirmed that they match: compelling evidence that those BMD cards were scanned

twice in the machine recount. The other 211 (214–3=211) duplicated scans are of

BMD cards from other precincts in Fulton County.

73. Coalition Plaintiffs also identified one hand-marked paper ballot that was scanned

twice in RW01 in the machine recount, and at least seven hand-marked paper ballots

that were scanned thrice in RW01 in the machine recount. I used the software in

Appendix 6 to check their work: the twenty-nine images indeed seem to represent only

eleven distinct pieces of paper, even though they contributed twenty-nine votes to

some contests, including the presidential contest. Appendix 8 shows the sets of

images. The table below lists the pairs and triples.

Multiple Image A Image B Image C

1 00801 00044 000168 00801 00043 000168

2 00801 00044 000083 00801 00043 000083
3 00801 00044 000042 00801 00043 000042
4 05160 00074 000023 05160 00067 000008
5 00794 00017 000024 00791 00026 000091 00791 00019 000010
6 00794 00017 000029 00791 00026 000086 00791 00019 000015
7 00794 00018 000001 00791 00026 000009 00791 00019 000092
8 00794 00018 000011 00791 00026 000019 00791 00019 000082
9 00794 00019 000002 00791 00026 000079 00791 00019 000022
10 00794 00019 000005 00791 00026 000076 00791 00019 000025
11 00794 00019 000006 00791 00026 000075 00791 00019 000026

24
Write-ins included votes for “Anyone,” “XXX,” “Willie Nelson,” and “Alexander Hamilton,”
as well as write-in votes for “Donald Trump” for District Attorney, Clerk of the Superior Court,
Tax Commissioner, Sheriff, Solicitor General, and Surveyor.
25
I used the software in Appendix 6 to facilitate the process.

23
 74. To confirm that the duplicate and triplicate images were included in the reported vote

tabulation, I searched the cast-vote records (CVRs) produced by Fulton County for

each image identifier among the duplicates and triplicates of images of RW01 ballots

and BMD printout cards. All twenty-four from the original count and all twenty-nine

from the machine recount were among the CVRs. I conclude that the duplicate and

triplicate votes were included in the reported machine tabulations, since the vote totals

derived from the CVRs agree with the reported vote totals, as mentioned in paragraph

60, supra.

75. For Fulton County as a whole, Coalition plaintiffs gave me a list that identified images

of 2,871 ballots and BMD printout cards that they believe were counted two or three

times in the second machine count. Some were identified by visual inspection of the

images; others were inferred to be duplicates because a sequence of cast vote records

was identical (or reversed) for long portions of two scan batches. As mentioned in

paragraph 72, supra, I confirmed that 214 of the purported duplicate scans of BMD

cards were indeed duplicates. I understand that this list of 2,871 are a sample from a

larger list of images of ballots and BMD printout cards that Coalition Plaintiffs assert

were included in the tabulation twice or more. I confirmed that all 6,118 images in

question were referenced in cast vote records in the second machine count, so all

presumably contributed to the tabulation.

76. Nine hundred sixteen (916) of the 2,871 sets of images were identified as images of

hand-marked paper ballots. I drew a random sample of 100 of those 916 using

software in Appendix 6. I set the seed for the pseudo-random number generator using

24
 ten rolls of ten-sided dice. Appendix 9 is an image of the dice with the digits they

showed, in order: 8, 6, 2, 8, 9, 2, 2, 1, 8, 4.

77. Of the 100 sets of images in the sample, 46 contained triplicate images.

78. I examined the sets of images visually, aided by software in Appendix 6. I agreed with

the Coalition Plaintiffs’ determination for 98 of the 100 sets. I disagreed with the

determination for one of the sets, and I was unable to verify one set. To be

conservative, I treat this as 98 agreements in 100 checks. The resulting 95 percent

lower confidence bound for the number of hand-marked paper ballots represented by

two or more scans is 891 ballots. That is, there is 95 percent statistical confidence that

at least 891 of the 918 claimed multiples are genuine multiples.

79. I did not have time to examine more purported replicate images of BMD printout

beyond the 214 mentioned in paragraph 72, but I might examine more before trial.

80. Based on the observations in paragraphs 58 through 78, supra, it seems that Fulton

County did not keep track of which ballots and BMD cards had been scanned and

which had not, in both the original count and in the machine recount. Alternatively or

additionally, the electronic records were altered accidentally or intentionally. The

electronic records of the election are not intact. This is a surprising lack of tracking

and protecting election materials: the most basic election safeguard is to check

whether the number of voters who participated is equal to the number of ballots and

BMD printout cards that were cast and to the number that were tabulated. Moreover, I

would expect all electronic election materials to be backed up onsite and offsite, at

least for the federally mandated retention period of twenty-two months, so the loss of

25
 hundreds of thousands of image files from the first machine count and of nearly

18,000 images from the second machine count is hard to fathom.

81. Fulton County would have noticed these errors had they simply kept track of ballots

and BMD printout cards and checked the total number against the number reported in

the electronic tabulation. It seems that Fulton County does not know how many ballots

and BMD printout cards were cast in the election, how many voters cast votes, or how

many pieces of paper were scanned—nor how those numbers compare to each other.

Absent basic ballot accounting, pollbook reconciliation, and counting of electronic

records, it is unsurprising that the two machine tallies differ so much (see the table

below paragraph 47, supra). The U.S. Election Assistance Commission has published

best practices for chain of custody.26

82. Fulton County’s chaotic, unaccountable curation and processing of cast ballots, cast

BMD printout, and electronic records make a true risk-limiting audit impossible. It is

unreasonable for voters to trust that their votes were counted at all, much less counted

correctly. Voters have good reason to believe that some votes counted more than

others: some votes were included twice or thrice in the totals. There is no way to know

how many votes were omitted from the tabulation, absent access to the physical ballots

and BMD printout and evidence that the chain of custody is intact. From the records

produced so far, it is impossible to determine whether malware, bugs,

misconfiguration, or malfeasance disenfranchised voters or altered the election results.

26
https://www.eac.gov/sites/default/files/bestpractices/Chain_of_Custody_Best_Practices.pdf
(last visited 9 March 2022)

26
 83. Based on my review of the Fulton County post-election audit, it is clear that the audit

planning, process, and controls did not detect the double and triple counting

documented above. Even if Fulton County did not rely on ballot-marking devices for

virtually all in-person voters, the lack of basic accounting controls makes it impossible

to determine who really won an election contest, even by hand counting the votes: the

record of the vote could easily be incomplete or adulterated. This remains true even if

BMDs could be relied upon to print voters’ selections accurately.

84. I have no reason to believe that problems of the kinds described above are limited to

Fulton County, but because of time constraints, I have not yet investigated other

counties. I might examine data from other Georgia counties before trial, including

comparing the tabulations based on images and cast-vote records to the ABBSs, other

RLA workpapers, and reported results.

The paragraphs below were in the version of this report submitted 11 January 2022, but

they have been renumbered.

Summary

85. A rigorous audit can provide confidence that a well-run election found the true

winner(s). But it cannot compensate for using untrustworthy technology to record

votes or for a poorly run election; in such circumstances, it distracts attention from the

real problems rather than improving election integrity and justifying confidence in

electoral outcomes. Absent a trustworthy record of the votes, no procedure can provide

affirmative evidence that the reported winner(s) really won. Georgia lacks such a

27
 record, for many reasons, including the heavy reliance on BMDs and the lack of

physical accounting of ballots, memory cards, and other election materials; lack of

pollbook and voter participation reconciliation; etc.

86. By claiming to perform risk-limiting audits when its paper trail is not trustworthy, the

State of Georgia is in effect adding stories to a building that needs its foundation

replaced. First things first.

87. To provide reasonable assurance that every voter’s selections are counted and counted

accurately requires systematic improvements to how Georgia conducts elections:

a) For every voter to be assured the right to cast an accountable vote, every voter should

have the opportunity to mark a ballot by hand, whether voting in person in advance,

in person on election day, or absentee by mail.

b) The use of ballot-marking devices should be reduced to a minimum, for reasons I

have explained in previous declarations. In particular:

i. BMDs do not necessarily print voters’ selections accurately. They can be hacked or

misconfigured, as explained in Prof. J. Alex Halderman’s testimony.

ii. A growing body of empirical work shows that few voters check the BMD printout,

and those who do rarely catch errors.

iii. There is no way for a voter to prove to an election official or anyone else that a

BMD malfunctioned. Hence, there is no way to “close the loop” to ensure that

malfunctioning devices are removed from service, even if some voters notice BMDs

misbehaving. And even if a device is caught misbehaving, there is no way to

reconstruct the correct election outcome.

28
 iv. There is no way to test BMDs adequately prior to, during, or after an election to

establish whether they altered votes, even if they altered enough votes to change

electoral outcomes. 27

c) Georgia must implement better procedures and checks on chain of custody of election

materials, especially voted ballots. Currently, Georgia is not in a position to

determine whether every validly cast ballot was included in the reported results, nor

whether there was electronic or physical “ballot-box stuffing” or votes were altered.28

Georgia needs better protocols for using and checking physical security seals on

ballots and voting equipment—and demonstrating that it has followed those

protocols. It needs to perform routine scrutiny of custody logs and surveillance video,

and to institute other related security measures.

d) Internal consistency checks and physical inventories must be performed as part of

Georgia’s canvass, including, among other things:

i. Verifying that the number of ballots sent to each polling location (and

blank paper stock for ballot-marking devices and ballot-on-demand

printers) equals the number returned voted, spoiled, or unvoted. This must

27
See note 8, supra.
28
This is evidenced by the fact that the 2020 audit found thousands of untabulated ballots. See
note 14, supra. Per the Secretary of State’s office, “[t]he audit process also led to counties
catching making mistakes they made in their original count by not uploading all memory cards.”
https://sos.ga.gov/index.php/elections/historic_first_statewide_audit_of_paper_ballots_upholds_r
esult of presidential race, last accessed 9 January 2022. Because of Georgia’s inadequate
physical accounting for voting materials, there is no way to know how many more votes validly
cast in that election have not been included in any of the reported results. Moreover, the lax
recordkeeping evidently resulted in scanning the same batches of ballots more than once.
Similarly, some ABBSs were presumably entered more than once, and as shown above, some
were not entered at all.

29
 be a physical check based on manual inventories, not on reports from the

voting system.

ii. Checking pollbooks and other voter participation records against the

number of voted ballots received, including checking whether the

appropriate number of ballots of each “style” were received.

iii. Checking whether the number of electronic vote records (“scans” and cast-

vote records) agrees with the physical inventory of ballots of each style.

e) Georgia should conduct routine “compliance” audits, a necessary precursor to

conducting risk-limiting audits. For a list of what compliance audits should include,

see, for example, Appel, A., and P.B. Stark, 2020. Evidence-Based Elections: Create

a Meaningful Paper Trail, Then Audit, Georgetown Law Technology Review, 4, 523–

541.

f) Georgia should conduct routine, genuine,29 risk-limiting audits of every contested

race in every election. The audits must have the ability to correct the reported

outcome if the outcome is wrong, before the outcome is certified. I understand that

under current Georgia law, audits take place only every other year, for only one

contest, and cannot change electoral outcome or trigger a recount—even if the audit

finds that the outcome is wrong. No matter how rigorous an audit is, an audit of one

or more contests provides no evidence that the outcome of any unaudited contest is

correct. Errors and malware may affect some contests but not others.

29
The pilots of RLA procedures in Georgia were not genuine RLAs, nor was the “full hand-
count audit.”

30
 g) A genuine RLA requires far more than Georgia has yet attempted. First and foremost,

it requires a trustworthy record of voter intent. Georgia’s records are untrustworthy

for a range of reasons, starting with the fact that all in-person voters are expected or

required to use ballot-marking devices (BMDs). As discussed at length in previous

declarations and in testimony by Prof. Andrew Appel and Prof. J. Alex Halderman,

BMD printout is not a trustworthy record of the vote. There are also issues with

Georgia’s verification of voter eligibility and voter participation. But even if every

voter used a hand-marked paper ballot and there were no issues with voter eligibility,

Georgia simply does not keep track of their election materials well enough. As

discussed in my previous declarations, the foundation for a risk-limiting audit is a

ballot manifest, a physical inventory of the paper ballots describing in detail how they

are stored. This must be derived without reliance on the voting system; otherwise, the

audit is trusting the voting system to check itself. For example, if there are ballots that

were never scanned or scans that were never uploaded (as discovered during the 2020

“audit”), they will be missing from a manifest derived from voting system reports.

The ballot manifest must be based on physical inventories of the ballots, keeping

track of where the ballots are and how they are organized. Absent that, it is

impossible to account for votes reliably, and impossible to limit the risk that an

incorrect electoral outcome will be certified: applying risk-limiting audit procedures

to an untrustworthy collection of ballots is “security theater.”

88. There are additional checks that could be performed to determine the root cause of the

discrepancies among the first machine tabulation, hand count, and machine recount.

Those checks require access to the physical ballots (for instance, to determine whether

31
 every scan batch from the tabulators reflects a distinct collection of actual physical

ballots) and access to the tabulators, software, and servers (by other experts in this

matter).

89. I would like to supplement my report once the Plaintiffs have had the opportunity to

review materials that Defendants have not yet produced or provided access to,

including ballots, and to review Plaintiffs’ experts’ reports once they have inspected

the hardware and software used in the November 2020 election.

I declare under penalty of perjury, in accordance with 28 U.S.C. § 1746, that the foregoing is true

and correct.

Executed on this date, 11 January 2020 9 March 2022,

_______________________________

Philip B. Stark

32
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 2 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 3 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 4 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 5 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 6 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 7 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 8 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 9 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 10 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 11 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 12 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 13 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 14 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 15 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 16 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 17 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 18 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 19 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 20 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 21 of 22
Case 1:17-cv-02989-AT Document 1297-6 Filed 02/01/22 Page 22 of 22