IN THE COURT OF APPEAL OF THE STATE OF CALIFORNIA

FIFTH APPELLATE DISTRICT

THE PEOPLE OF THE STATE OF

CALIFORNIA,

Plaintiff-Respondent, Court of Appeal No.F071640

v. Superior Court No. BF151825A

BILLY RAY JOHNSON,

Defendant-Appellant.

Appeal From the Superior Court of Kern County

Honorable Gary T. Friedman, Judge

BRIEF OF THE INNOCENCE PROJECT, INC., THE CALIFORNIA

INNOCENCE PROJECT, THE NORTHERN CALIFORNIA INNOCENCE
PROJECT, AND THE LOYOLA LAW SCHOOLS PROJECT FOR THE
INNOCENT AS AMICI CURIAE
IN SUPPORT OF DEFENDANT-APPELLANT

Gerald B. Hrycyszyn SBN: 227814 Dana M. Delger

Michael A. Albert INNOCENCE PROJECT, INC.

Received by Fifth District Court of Appeal

Justin C. Colannino 40 Worth Street, Suite 701
WOLF, GREENFIELD & SACKS, P.C. New York, NY 10013
600 Atlantic Avenue Tel: 212.364.5340
Boston, MA 02210
Tel: 617.646.8000
Counsel for Amici

Linda Starr Justin Brooks Paula M. Mitchell

NORTHERN CALIFORNIA Alexander Simpson LOYOLA LAW SCHOOL
INNOCENCE PROJECT AT SANTA CALIFORNIA INNOCENCE PROJECT FOR THE
CLARA UNIVERSITY SCHOOL PROJECT INNOCENT
OF LAW 225 Cedar St. 919 Albany St.
900 Lafayette Street, Suite 105 San Diego, CA 92101 Los Angeles, CA 90015
Santa Clara, CA 95050 Tel.: 619.515.1525 Tel: 213.736.8143
Tel: 408.554.4790

1
 TABLE OF CONTENTS

TABLE OF CONTENTS................................................................................................................ 2
TABLE OF AUTHORITIES .......................................................................................................... 3
I. INTRODUCTION ................................................................................................................. 7
II. BACKGROUND ................................................................................................................... 9
A. Procedural History ......................................................................................................... 9
B. TrueAlleles Secret Source Code ................................................................................. 10
1. TrueAlleles Undisclosed Source Code Analyzes The Physical Evidence
And Spits Out Match Data Whose Relevance To The Input Data Is
Shielded From Examination. ................................................................................ 10
2. TrueAlleles Inconsistent Match Statistic Values Are Often Astronomical
Compared To Those Of Its Competitors, And the Reasons Why Are
Hidden In The Source Code.................................................................................. 12
III. THE CONSTITUTIONAL GUARANTEE OF A MEANINGFUL
OPPORTUNITY TO DEFEND AGAINST THE STATES ACCUSATIONS ................. 14
IV. ARGUMENT ...................................................................................................................... 15
A. Software Errors, Biases, Or Inconsistencies Can Be Uncovered Only By
Examining How The Source Code Actually Operates................................................. 17
1. Source Code Errors Are Commonplace And Are Likely To Be Present In
TrueAllele ............................................................................................................. 17
2. TrueAlleles Source Code Is Statistically Expected To Have Errors
Impacting The Match Statistic .......................................................................... 19
B. Source Code Is Needed To Test The Reliability Of The Governments Case............. 20
C. The Trial Court Erred In Allowing Trade Secret Protections To Outweigh Due
Process ......................................................................................................................... 22
1. It Was Error To Put TrueAlleles Commercial Interests Before Mr.
Johnsons Right To Challenge The Evidence Against Him ................................. 22
2. TrueAlleles Commercial Interests Can Be Protected Without Violence To
Mr. Johnsons Constitutional Rights..................................................................... 25
V. CONCLUSION ................................................................................................................... 27
CERTIFICATE OF WORD COUNT ........................................................................................... 28

2
 TABLE OF AUTHORITIES

CASES

Chambers v. Mississippi,
410 U.S. 284 (1973) ............................................................................................................ 14, 23

Commonwealth v. Foley,
38 A.3d 882 (Pa. Super. Ct. 2012) ............................................................................................ 13

Commonwealth v. Robinson,
No. CP-02-CR-0007777-2013 (C.P. Allegheny Cty. Nov. 19, 2015) ...................................... 21

Crane v. Kentucky,
476 U.S. 683 (1986) .................................................................................................................. 14

Davis v. Alaska,
415 U.S. 308 (1974) ............................................................................................................ 23, 24

E-Contact Techs., LLC v. Apple, Inc.,

2012 WL 11924448 (E.D. Tex. June 19, 2012) ........................................................................ 26

Hinton v. Alabama,
134 S. Ct. 1081 (2014) ........................................................................................................ 16, 20

In re Avena,
12 Cal. 4th 694 (1996) .............................................................................................................. 17

In re Brown,
17 Cal. 4th 873 (1998) .............................................................................................................. 15

In re Newbern
175 Cal. App. 2d 862 (Dist. Ct. App. 1959) ............................................................................. 14

In re Sodersten,
146 Cal. App. 4th 1163 (Ct. App. 2007)............................................................................. 15, 20

In re Source Code,
816 N.W.2d 525 (Minn. 2012) ................................................................................................. 18

Kyles v. Whitley,
514 U.S. 419 (1995) ............................................................................................................ 15, 20

Melendez-Diaz v. Massachusetts,
557 U.S. 305 (2009) .................................................................................................................. 16

MobileMedia Ideas LLC v. HTC Corp.,

No. 2:10-CV-112-JRG, 2012 WL 7783403 (E.D. Tex. Oct. 19, 2012).................................... 26

3
Ohio v. Shaw,
No. CR-13-575691 (Ohio Ct. Com. Pl. Cuyahoga Cty. Oct. 10, 2014) ................................... 12

Pennsylvania v. Ritchie,
480 U.S. 39 (1987) .............................................................................................................. 24, 25

People v. Cunningham,
25 Cal. 4th 926 (Cal. 2001)....................................................................................................... 23

People v. Hansel,
1 Cal. 4th 1211 (1992) .............................................................................................................. 14

People v. Hernandez,
53 Cal. 4th 1095 (2012) ........................................................................................................ 9, 14

People v. Hilary,
Ind. #: 2015-15 (St. Lawrence Cty., New York, August 26, 2016) .......................................... 12

Perry v. New Hampshire,

565 U.S. 228 (2012) .................................................................................................................. 23

Rockstar Consortium U.S. LP v. Google Inc.,

No. 2:13-CV-893, 2014 WL 5831041 (E.D. Tex. June 19, 2014) ........................................... 26

Spencer v. Texas,
385 U.S. 554 (1967) .................................................................................................................. 14

State Farm Fire & Cas. Co. v. Superior Court,

54 Cal. App. 4th 625 (Cal. App. 1997) ..................................................................................... 27

State v. Chun,
943 A.2d 114 (N.J. 2008) ......................................................................................................... 18

U.S. v. Cronic,
466 U.S. 648 (1984) ........................................................................................................ 8, 14, 16

U.S. v. Johnson,
15-CR-565, D.I. 67 (S.D.N.Y. July 18, 2016) .................................................................... 19, 27

U.S. v. Nixon,
418 U.S. 683 (1974) .................................................................................................................. 15

U.S. v. Nobles,
422 U.S. 225 (1975) .................................................................................................................. 24

U.S. v. Scheffer,
523 U.S. 303 (1998) ............................................................................................................ 23, 27

4
Washington v. Texas,
388 U.S. 14 (1967) .............................................................................................................. 17, 21

Youngblood v. West Virginia,

547 U.S. 867 (2006) ............................................................................................................ 15, 20

STATUTES

California Evidence Code 1060 ........................................................................................... 10, 27

OTHER AUTHORITIES

Ars Techncia, A year of digging through code yields smoking gun on VW, Fiat diesel cheats,
May 28, 2017 ............................................................................................................................ 18

Charles Fishman, They Write The Right Stuff, FastCompany (Dec. 31, 1996)............................. 20

David E. Anderson et al., Author-Initiated Retraction: Anderson et al, Induced Alpha Rhythms
Track the Content and Quality of Visual Working Memory Representations with High
Temporal Precision, 35 J. Neurosci. 2838, 2838 (2015) .......................................................... 18

David Murray, Queensland Authorities Confirm Miscode Affects DNA Evidence in Criminal
Cases, Courier Mail (Mar. 20, 2015) .................................................................................... 9, 19

ESR, Access to STRmixTM Software by Defence Legal Teams (2016) ...................................... 9, 19

James Gleick, A Bug and a Crash, N.Y. Times Mag., Dec. 1, 1996 ........................................... 17

Mark W. Perlin, Cybergenetics TrueAllele Technology Enables Objective Analysis of Previously

Unusable DNA Evidence, MathWorks (Apr. 17, 2013) ........................................................... 19

N.Y. Times, Experts Find a Door Ajar in an Internet Security Method Thought Safe, April 8,
2014 .......................................................................................................................................... 17

N.Y. Times, Traces of Crime: How New Yorks DNA Techniques Became Tainted, Sept. 4, 2017
............................................................................................................................................ 19, 27

National Academy of Sciences, Committee on Identifying the Needs of the Forensic Sciences
Community, Strengthening Forensic Science in the United States: a Path Forward, 7 (2009) 7

Richard P. Mann et al., Retraction: Multi-scale Interference of Interaction Rules in Animal

Groups Using Bayesian Model Selection, 8 PLOS Comput. Biol., No. 8 (Aug. 2012) ........... 18

Shuhuan Zhao & Zheng-ping Hu, Retracted: Occluded Face Recognition Based on Double
Layers Module Sparsity Difference, 2015 Advances in Electronics 210160 (2015) ................ 18

Signal Processing Toolbox Release Notes, MathWorks ............................................................... 20

Statistics and Machine Learning Toolbox Release Notes, MathWorks ........................................ 20

5
William C. Thompson et al., Forensic DNA Statistics: Still Controversial in Some Cases, The
Champion, at 20 (Dec. 2012) .................................................................................................... 13

RULES

United States District Court for the Eastern District of Texas Local Patent Rule 3-4.................. 26

6
I. INTRODUCTION

DNA evidence can be powerfulindeed, often conclusiveevidence of guilt or

innocence. An individuals DNAif properly handled, analyzed, and interpretedhas
been rigorously shown to have the capacity to consistently, and with a high degree of
certainty, demonstrate a connection between evidence and a specific individual or
source. National Academy of Sciences, Committee on Identifying the Needs of the
Forensic Sciences Community, Strengthening Forensic Science in the United States: a
Path Forward, 7 (2009). The phenomenal statistical accuracy and scientific imprimatur
conveyed by the presence of DNA evidence, and the resulting tendency for such evidence
to overwhelm any doubts a factfinder may have, requires that its use in court be
rigorously tested by the adversarial process. Much the way a chain of custody must be
subject to rigorous examination and disclosureeven if the item has only changed hands
three or four timesso here the data that purports to reflect whether a DNA match has
been identified must be subjected to similar disclosure. In this case, the relevant chain
of custody is that of the data as it worked its way through TrueAlleles secret computer
program upon which the State relied to convict Mr. Johnson. That chain may be dozens
or even hundreds of steps long (the computer code itself measuring some 170,000 lines).
Yet the trial court refused to allow defense counsel to examine that chain to see if
distortions occurred as a result of errors in that code errors that, as discussed below, are
extraordinarily likely to be present.
Here, in short, the government seeks to turn the light-throwing power of DNA into
darkness by using a computer program called TrueAllele Casework (TrueAllele), that
operates in a manner unknown even to the government labs that purchase it. How the
DNA match statistic that the State presented to the jury is calculated is known only to
TrueAlleles source code programmers,1 and is withheld from the accused.
This non-disclosure turns what might otherwise be a reliable piece of evidence into

1
Source code is the human-readable instructions that control the calculations performed
by a computer program. As the State indicates, without reading the source code, it is
nearly impossible to know how a given piece of software will operate. (4 CT 934-35).

7
an untestable oracle of innocence or guilt. Indeed, the company that makes TrueAllele,
Cybergenetics, openly proclaims that its undisclosed method [t]urns inconclusive data
into a match statistic strong enough for court that is typically a million times stronger
than manual interpretation of the same physical evidence. TrueAllele Forensic E-
Brochure (E-Brochure) at 4.2 For inconclusive data to suddenly gain the aura of
scientific certainty in an unverifiable manner should hardly reassure anyone least of all
courts charged with ensuring fundamental fairness and adversarial openness to one facing
a lifetime of incarceration. Yet, because the trial court refused to order the disclosure of
TrueAlleles source code, Mr. Johnson was denied the opportunity even to confront, let
alone understand, evaluate, or challenge, how the match presented to the jury was
calculated in his case.
Let there be no mistake about the interests at issue here. TrueAllele, a private for-
profit company, is selling data-analysis services to the People of California, and claiming
that it can generate results with accuracy strong enough for court. Yet it is refusing to
allow a defense expert to test that proposition by reviewing its source code under any
circumstanceseven under terms that courts would routinely order if TrueAllele were
accused of patent infringementi.e., if only money were at stake. But Mr. Johnson
stands to lose his freedom for the rest of his life without a meaningful opportunity to
challenge the most significant evidence against him. The balance is not close.
The trial courts ruling permitting TrueAlleles author, Dr. Mark Perlin, to present
conclusions derived by his software without allowing Mr. Johnson to evaluate how
TrueAllele came to those conclusions violates Mr. Johnsons due process and
Compulsory Process Clause rights. Simply put, the Constitution demands that the
accused be permitted to learn how the software turns the inconclusive data into a
match statistic3 so that he can present a full defense and subject the method to the
crucible of meaningful adversarial testing. U.S. v. Cronic, 466 U.S. 648, 656-57 (1984);
2
Available at: https://www.cybgen.com/products/casework/forensic_e-brochure.pdf.
3
TrueAlleles match statistic purports to be the probability greater than coincidence
that a DNA sample contains a given persons DNA. 23 RT 4115-4116, 4118.

8
People v. Hernandez, 53 Cal. 4th 1095, 1104 (2012). Mr. Johnsons conviction must be
reversed so he can open the TrueAllele black box to properly confront the DNA match
statistic the Government introduced. Without such access, it is impossible to determine
whether the match statistics heresome of which were based on the same physical
evidence found inconclusive by manual analysisare scientifically sound, or the result
of a programming error, deliberate tampering, or an unsound algorithm. Doubts about
TrueAlleles accuracy are plausiblea competing program, STRmix, that performs the
same type of analysis as TrueAllele, was found to have coding errors that impacted the
statistics presented in court. David Murray, Queensland Authorities Confirm Miscode
Affects DNA Evidence in Criminal Cases, Courier Mail (Mar. 20, 2015).4 STRmixs
source code is now made available to defendants. ESR, Access to STRmixTM Software by
Defence Legal Teams (2016).5 Without access to TrueAlleles code, Mr. Johnson is
deprived of a meaningful ability to challenge the States case.
Due process and compulsory process demand that the Court require TrueAllele to
provide the defense with access to its source code, subject to reasonable restrictions
tailored to protect any trade secret interests TrueAllele may demonstrate. This Court
should reverse Mr. Johnsons conviction because he was denied a fair opportunity to
challenge the prosecutions case.

II. BACKGROUND

A. Procedural History

Both prior to and again during trial, Mr. Johnson moved the trial court for access
to TrueAlleles source code. (4 CT 902; 4 RT 430-491; 23 RT 4090-4111). His counsel
argued that due process required disclosure of the source code so that Mr. Johnson could
appropriately identify and challenge the formula used to calculate the match statistic.
4
Available at: http://www.couriermail.com.au/news/queensland/queensland-authorities-
confirm-miscode-affects-dna-evidence-in-criminal-cases/news-
story/833c580d3f1c59039efd1a2ef55af92b.
5
Available at: https://strmix.esr.cri.nz/assets/Uploads/Defence-Access-to-STRmix-April-
2016.pdf.

9
E.g., (4 RT 432:3-27) (What [the defense is] asking for is essentially for [TrueAllele] to
show [its] work, to show us the formula, to show how the formula works, so we can
verify.... [N]obody else knows [the formula]. Nobody else has ever seen the formula.
Nobody else has ever gone through the formula. So we dont know if when we give this
to a statistician they will confirm and say yes, yes, [TrueAlleles] math is good.).
The prosecution opposed, arguing (i) that the materials sought by Defendant were
privileged pursuant to California Evidence Code 1060, and (ii) the source code was
immaterial. (4 CT 930-31; 23 RT 4097-111). The trial court denied Mr. Johnsons pre-
trial motion, relying on 1060. (4 RT 496:18-20) (The source code is a trade secret. I
dont think adequate showing has been made to justify the breech [sic] of that privilege.)
The trial court reiterated this holding at trial: Theres been no particularized showing
under Evidence Code 1060, et seq, as to how the TrueAllele source code is necessary to
defenses ability to test the reliability of its results. (23 RT 4111:8-12).
Dr. Perlin testified at trial and presented the various match statistics computed
by the software (23 RT 3971-4268), resulting in Mr. Johnsons conviction.

B. TrueAlleles Secret Source Code

1. TrueAlleles Undisclosed Source Code Analyzes The Physical

Evidence And Spits Out Match Data Whose Relevance To The
Input Data Is Shielded From Examination.

TrueAllele purports to use the same workflow (i.e. sequence of steps) as the
DNA analysis techniques of its competitors. This is clear from Dr. Mark Perlins
testimony. (23 RT 4056:17-4058:1). For example, before analysis by TrueAllele, DNA
samples must first be collected and amplified. These amplified samples are then run on a
genetic analyzer which measures patterns at well-defined and routinely-used locations in
the human genome. Dr. Perlin made clear that TrueAllele has nothing to do with this
gathering of the physical evidence i.e. the genetic analyzer. Rather, TrueAllele
mathematically interprets the data produced by the genetic analyzer to generate a match
statistic. (23 RT 4058:2-5). In short, TrueAllele operates on the same of physical
evidence as is its competitors. Yet it advertises materially different results.

10
 How does TrueAllele achieve this remarkable level of match statistics that other
companies do not? Neither defense counsel nor this Court is currently in a position to
know, and it is impossible to know without the disclosure of the source code. As with
other systems, TrueAlleles source code encodes rules, or assumptions, that generate its
match statistic conclusions as reported in court. A YouTube tutorial by Cybergenetics
outlines the TrueAllele workflow. Cybergenetics, TrueAllele Process Overview
(TrueAllele Overview).6 First, the data generated by the genetic analyzer is sent to
Cybergenetics for interpretation. Id. After receiving the raw data for interpretation, the
TrueAllele source code assesses the quality of the data. TrueAllele alerts the user to any
potential data issues . . . when a [TrueAllele] rule has detected a problem. Id. at 1:20-
1:45. After a few minutes, an analyst has reviewed all the quality check data. Id.
The analyst then inputs multiple parameters, or assumptions, that are utilized in
the TrueAllele analysis. Dr. Perlin indicated that these parameters include assumptions
input by the operator concerning the number of contributors to the DNA sample. (23 RT
4260:22 4261:17). Because TrueAlleles source code has not been revealed, it is
unclear how the analysts assumptions interact with the raw data and TrueAlleles rules
to generate a so-called match statistic. As Dr. Perlin testified, an incorrect assumption
regarding the number of DNA contributors can inflate the [match] statistic by a few
zeros i.e. by one hundred or one thousand times. Id. In other words, changing that
assumption could turn an admittedly inconclusive match statistic of 100 into a match
of 10,000 or 100,000. Id.
Cybergenetics online tutorial indicates that after the analysts assumptions are
entered, these assumptions and the raw data are subjected to additional TrueAllele rules,
separate from those used in the preliminary analysis, resulting in a match statistic that
Cybergenetics invites the analyst to simply paste into their case report. TrueAllele
Overview at 8:35-8:50. In other words, TrueAlleles source code conceals substantive
choices that are insulated from adversarial examination choices, for example, about

6
Available at: https://www.youtube.com/watch?v=OU29b5sW88Y.

11
what data issues are worthy of being reported to the analyst during the preliminary
quality analysis, or how the analysts assumptions are merged with the raw data to
generate a match statistic. The methodology behind each of these determinations is
hidden within the source code, thus concealing from defense counsel potential data
distortions at any step along the way, be they part of the algorithm or simply arising from
a coding error.

2. TrueAlleles Inconsistent Match Statistic Values Are Often

Astronomical Compared To Those Of Its Competitors, And the
Reasons Why Are Hidden In The Source Code.

TrueAllele is a computer program developed by Cybergenetics. In its advertising,

Cybergenetics claims that the program [t]urns inconclusive data into a match statistic
strong enough for court. E-Brochure at 4.7 These advertisements claim that
TrueAlleles raw data interpretations are typically a million times stronger than
competing interpretations of the same physical evidence. Id. How TrueAllele is able to
produce data a million times stronger is hidden in TrueAlleles source code.
TrueAlleles match statistics often purport to provide inculpating evidence that
its more modest competitors are unable to provide. In various instances, TrueAlleles
competitors found DNA analysis data too unreliable to generate an interpretable match
statistic, yet TrueAllele generated match statistics of astounding confidence. See e.g.,
Ohio v. Shaw, No. CR-13-575691, at *2 (Ohio Ct. Com. Pl. Cuyahoga Cty. Oct. 10,
2014) (while analyses by two independent competitors were inconclusive, TrueAlleles
analysis of the same raw data generated a purportedly conclusive match statistic).8
And even when TrueAlleles competitors do return a match, TrueAlleles

7
Available at: https://www.cybgen.com/products/casework/forensic_e-brochure.pdf.
8
The reverse has also happened: a TrueAllele competitor returned conclusive results
when TrueAllele did not. People v. Hilary, Ind. #: 2015-15 (St. Lawrence Cty., New
York, August 26, 2016) (noting that TrueAllele found a genetic sample inconclusive,
while a competitor program found that same sample conclusive). Such inconsistent
results between programs only underscore the need for defendants to be able to access
source code.

12
corresponding results for the same data often claim a match rate several orders of
magnitude larger. See e.g., Commonwealth v. Foley, 38 A.3d 882, 887, 890 (Pa. Super.
Ct. 2012) (while competitors found match statistics of 13,000 and 23,000,000, TrueAllele
generated a match statistic of 189,000,000,000over 14,000,000 times higher than one
of its competitors). It could be that TrueAllele is actually more precise than similar
programs, but that assumption cannot be tested in the absence of the source code.
Worse, TrueAlleles own repeat analyses of the same data have resulted in vastly
different outcomes. William C. Thompson et al., Forensic DNA Statistics: Still
Controversial in Some Cases, The Champion, at 20 (Dec. 2012)9 (indicating that in one
case, TrueAllele analyzed a data set four independent times, generating match statistics of
389 million, 1.9 billion, 6.03 billion, and 17.8 billion a forty-five-fold difference
between the lowest and highest likelihood ratio, despite operating on the very same
physical evidence). Just as with the differences between TrueAlleles results and its
competitors, the underlying reasons for these divergences is hidden from the defense.
This case itself demonstrates TrueAlleles internally inconsistent results when
operating on the exact same data sample. Two stains on pants from the Antonia scene
were deemed inconclusive by Kern County Lab criminalist Mr. Sugimotos manual
analysis. When Dr. Perlin operated TrueAllele, the software took this same inconclusive
physical stain evidence, and generated match statistics of 5,400,000 and 1,000,000,
respectively. Later, Kern Countys Mr. Sugimoto ran TrueAllele on the same physical
stains, but generated match statics of 2,100,000 and 426,000 about 50% less than Dr.
Perlins results using the same program.
This is the software on which the government relied to convict and incarcerate Mr.
Johnson. The Court need not conclusively determine the unreliability of TrueAllele to
acknowledge that, at a minimum, such inconsistent results buttress Mr. Johnsons right to
confront and examine the methodology used by that software.

9
Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2214459.

13
III. THE CONSTITUTIONAL GUARANTEE OF A MEANINGFUL
OPPORTUNITY TO DEFEND AGAINST THE STATES ACCUSATIONS

Whether rooted directly in the Due Process Clause of the Fourteenth

Amendment, or in the Compulsory Process or Confrontation clauses of the Sixth
Amendment, the Constitution guarantees criminal defendants a meaningful opportunity
to present a complete defense. Crane v. Kentucky, 476 U.S. 683, 690, (1986) (internal
citations omitted). Indeed, the Due Process Clause guarantees the fundamental elements
of fairness in a criminal trial. Spencer v. Texas, 385 U.S. 554, 56364 (1967).
Similarly, the spirit and the purpose of the right to due process under the California
Constitution is to assure to everyone a full and ample opportunity to be heard before he
can be deprived of his liberty or his property. People v. Hansel, 1 Cal. 4th 1211, 1219
(1992) (quoting In re Newbern 175 Cal. App. 2d 862, 865 (Dist. Ct. App. 1959)). This
right requires that a defendant be given a fair opportunity to defend against the States
accusations. Chambers v. Mississippi, 410 U.S. 284, 294 (1973). Confrontation and
cross-examination both serve this purpose: The defense has a fair opportunity to defend
only if it is permitted to subject the States evidence to the crucible of meaningful
adversarial testing. Cronic, 466 U.S. at 656; Hernandez, 53 Cal. 4th at 1104 (same); see
also Crane, 476 U.S. at 690 ([E]xclusion of [evidence that is central to the defendants
claim of innocence] deprives a defendant of the basic right to have the prosecutors case
encounter and survive the crucible of meaningful adversarial testing.) (quoting Cronic,
466 U.S. at 656).
The Compulsory Process Clause likewise protects the right to have a meaningful
opportunity to obtain, and bring before the trial court, evidence that is either
independently exculpatory or that calls into question inculpatory evidence. As the
Supreme Court has explained,

We have elected to employ an adversary system of criminal justice in

which the parties contest all issues before a court of law. The need to
develop all relevant facts in the adversary system is both fundamental and
comprehensive. The ends of criminal justice would be defeated if
judgments were to be founded on a partial or speculative presentation of the

14
 facts. The very integrity of the judicial system and public confidence in the
system depend on full disclosure of all the facts, within the framework of
the rules of evidence. To ensure that justice is done, it is imperative to the
function of courts that compulsory process be available for the production
of evidence needed either by the prosecution or by the defense.

U.S. v. Nixon, 418 U.S. 683, 709 (1974).

Due process also requires disclosure of impeachment evidence in the possession of
others acting on the governments behalf in the case. Kyles v. Whitley, 514 U.S. 419,
437-38 (1995); In re Brown, 17 Cal. 4th 873, 881 (1998) (prosecutors failure to review
crime lab files and to disclose exculpatory evidence would be fundamentally at odds
with the due process imperatives of the disclosure rule); Youngblood v. West Virginia,
547 U.S. 867, 869-70 (2006) (the Brady duty extends to impeachment evidence); In re
Sodersten, 146 Cal. App. 4th 1163, 1225 (Ct. App. 2007) (same).

IV. ARGUMENT

Admitting evidence generated by an undisclosed methodology that the defense is

precluded from examining violates due process because the accused has no meaningful
opportunity to challenge the States case. Much as the government must allow a
defendants experts to examine a murder weapon for defects or other exculpatory signs,
so too must it allow defense experts to examine for defects the source code which
purports to label a defendant as having a DNA match with evidence from a crime scene.
Perhaps an even better analogy is that of chain of custody. Much as the defense has the
right to investigate and examine the chain of custody of critical prosecution evidence as it
changed hands or locations, so too must it be able to examine the process by which the
data that yields the inculpatory match statistic wends its way from module to module in
TrueAlleles software code. If a defendant has the right to question whether there was a
processing error when an item of evidence was handed off from a detective to a lab
technician, surely he has the right to examine whether a processing error crept into one or
more of the 170,000 lines of code based on which TrueAllele takes a sample and emerges
on the other end with a breathtakingly precise-sounding probability estimate regarding a

15
DNA match. Yet the trial court refused to allow defense counsel to examine that code for
errors that, as discussed below, are commonplace.
The government may not require that the court and the accused simply trust that
TrueAlleles source code performs its claimed functions correctly and accurately. Due
process requires that the defense be permitted to subject the States evidence to the
crucible of meaningful adversarial testing, Cronic, 466 U.S. at 656, which requires the
production of evidence needed . . . by the defense, Nixon, 418 U.S. at 709, including
evidence that impeaches.
That requirement was not met here. Without disclosure of TrueAlleles source
code, Mr. Johnson had no fair opportunity to mount a defense, as he was unable to
determine whether TrueAlleles computer-generated match statistic was the result of a
valid and reliable scientific methodology, as the government claims, or of an error in the
code. As discussed infra in Section IV.A.1, such errors can and all-too-frequently do
arise notwithstanding good-faith efforts by programmers to produce accurate and
unbiased code. One thus need not assume though the accused should not be forced to
disregard the possibility of a more insidious cause of error in the code, such as the self-
interested acts of a for-profit company that touts match statistics typically a million
times stronger than its competitors achieve when interpreting the same physical
evidence. (E-Brochure at 4.)
Under normal circumstances, the way to expose a dishonest, incompetent, or
simply mistaken expert is cross-examination. See Hinton v. Alabama, 134 S. Ct. 1081,
1089-90 (2014); Melendez-Diaz v. Massachusetts, 557 U.S. 305, 31719 (2009). But
without access to the code showing how the match statistic actually works, the
defenses ability to cross-examine the prosecutions experts is illusory. The defense may
lob questions at the expert but since it has no data from which to base such questions, the
defense lacks a genuine opportunity to call out the experts self-serving denials of any
error in the system. The defense is thus hamstrung. If the code contained a bug that
returned an incriminating match statistic for any defendant whose last name begins
with J, Mr. Johnson would have no way of knowing that fact. Errors, of course, are

16
likely to be far more subtle; but the trial courts denial of access would provide no
opportunity to correct even such glaring errors.
In sum, without access to TrueAlleles source code, Mr. Johnson had no
opportunity to meaningfully test the prosecutions case. Washington v. Texas, 388 U.S.
14, 18-19 (1967); In re Avena, 12 Cal. 4th 694, 77475 (1996) (adversarial testing
requires, at a minimum, timely and in-depth investigation of potential defenses and the
strengths and weaknesses of the prosecutions case). Any countervailing commercial
interest Cybergenetics may have in the secrecy of its code, moreover, can be adequately
protected by a Protective Order, as is routinely done in intellectual property litigation.

A. Software Errors, Biases, Or Inconsistencies Can Be Uncovered

Only By Examining How The Source Code Actually Operates

1. Source Code Errors Are Commonplace And Are Likely To Be

Present In TrueAllele

Put simply, source code is written by people, and people make mistakes. Bugs
mistakes made by the software engineers when programmingin source code are
commonplace. Such bugs can and do impact the reliability of underlying calculations.
Indeed, such errorsoften caused by mere oversight, though on some occasions resulting
from mischief, bias, or self-interestcan plague even multi-billion-dollar technologies.
For example, a rocket was destroyed at launch due to a source code rounding error.
James Gleick, A Bug and a Crash, N.Y. Times Mag., Dec. 1, 199610 (reporting that an
explosion that destroyed the Ariane 5, a rocket valued at $7 billion and carrying a cargo
of $500 million, was due to a simple source code error). The encryption program
OpenSSL, used by [u]p to two-thirds of all e-commerce websites was insecure for
years due to the heartbleed bug that went unnoticed by Yahoo, Facebook, Google and
Amazon Web Services. N.Y. Times, Experts Find a Door Ajar in an Internet Security
Method Thought Safe, April 8, 2014.11 Computer code can also be used to hide malicious

10
Available at: https://around.com/ariane.html.
11
Available at: https://bits.blogs.nytimes.com/2014/04/08/flaw-found-in-key-method-for-
protecting-data-on-the-internet/.
17
errors, like the Volkswagen defeat device used to cheat on emissions testing. Ars
Techncia, A year of digging through code yields smoking gun on VW, Fiat diesel
cheats, May 28, 2017.12
Even scientific peer-reviewed articles are commonly retracted due to coding
errors. See David E. Anderson et al., Author-Initiated Retraction: Anderson et al,
Induced Alpha Rhythms Track the Content and Quality of Visual Working Memory
Representations with High Temporal Precision, 35 J. Neurosci. 2838, 2838 (2015)13
(article retracted due to a source code error); Shuhuan Zhao & Zheng-ping Hu, Retracted:
Occluded Face Recognition Based on Double Layers Module Sparsity Difference, 2015
Advances in Electronics 210160 (2015)14 (same); Richard P. Mann et al., Retraction:
Multi-scale Interference of Interaction Rules in Animal Groups Using Bayesian Model
Selection, 8 PLOS Comput. Biol., No. 8 (Aug. 2012)15 (same).
Forensic source code is not immune from such errors, and defense access to other
forensic source code has revealed errors impacting reliability. In re Source Code, 816
N.W.2d 525, 528, 543 (Minn. 2012) (affirming finding, after defense access, that a
version of the source code contained errors that impacted the reliability of the Intoxilyzer,
an instrument used to measure breath alcohol levels); State v. Chun, 943 A.2d 114, 120,
123 (N.J. 2008) (finding, after defense access, that the source code for Alcotest 7110, an
instrument used to measure breath alcohol levels, required certain modifications . . . to
permit its results to be admissible or to allow it to be utilized to prove a per se violation).
Genotyping software in particular has also been found error-prone: source code
errors impacting the reliability of STRmix, a competitor to TrueAllele, materially altered
match statistics in DNA mixture analysis in over sixty cases. David Murray, Queensland
Authorities Confirm Miscode Affects DNA Evidence in Criminal Cases, Courier Mail

12
Available at: https://arstechnica.com/cars/2017/05/volkswagen-bosch-fiat-diesel-
emissions-cheats-cracked-open-in-new-research/.
13
Available at: http://www.jneurosci.org/content/jneuro/35/6/2838.full.pdf.
14
Available at: https://www.hindawi.com/archive/2015/210160/.
15
Available at: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3413371/.

18
(Mar. 20, 2015).16 To mitigate that risk, STRmix now provides access to its source code
whenever it has been used to generate scientific evidence for the prosecution in a case.
ESR, Access to STRmixTM Software by Defence Legal Teams (2016).17 More recently, a
New York Federal court permitted the inspection of the source code underlying the
Forensic Statistical Tool, or FST developed in-house by the New York City DNA
laboratory. U.S. v. Johnson, 15-CR-565, D.I. 67 (S.D.N.Y. July 18, 2016). Given access,
a computer scientist for the defense wrote in an affidavit: the correctness of the behavior
of the FST software should be seriously questioned. N.Y. Times, Traces of Crime: How
New Yorks DNA Techniques Became Tainted, Sept. 4, 2017.18 Two months after the
Court ordered access, in September 2016, the New York City laboratory circulated a
memorandum, notifying customers that it would retire FST in favor of STRmix. Id.

2. TrueAlleles Source Code Is Statistically Expected To Have

Errors Impacting The Match Statistic
It is unreasonable to simply assume that TrueAllele, unlike countless other pieces
of software, is necessarily free from such errors. In fact, TrueAllele uses two
programming toolboxes (the Signal Processing Toolbox and the Statistics and
Machine Learning Toolbox) from a program called MATLAB to perform some of its
calculations. These MATLAB toolboxes themselves have documented Bug Reports
and Bug Fixes. Mark W. Perlin, Cybergenetics TrueAllele Technology Enables
Objective Analysis of Previously Unusable DNA Evidence, MathWorks (Apr. 17, 2013)19
(delineating MATLAB toolboxes used by TrueAllele). Statistics and Machine Learning
16
Available at: http://www.couriermail.com.au/news/queensland/queensland-authorities-
confirm-miscode-affects-dna-evidence-in-criminal-cases/news-
story/833c580d3f1c59039efd1a2ef55af92b.
17
Available at: https://strmix.esr.cri.nz/assets/Uploads/Defence-Access-to-STRmix-
April-2016.pdf.
18
Available at: https://www.nytimes.com/2017/09/04/nyregion/dna-analysis-evidence-
new-york-disputed-techniques.html
19
Available at: https://www.mathworks.com/company/newsletters/articles/cybergenetics-
trueallele-technology-enables-objective-analysis-of-previously-unusable-dna-
evidence.html.

19
Toolbox Release Notes, MathWorks20 (linking to Bug Reports and Bug Fixes for the
toolbox); Signal Processing Toolbox Release Notes, MathWorks21 (same).
Moreover, the TrueAllele codebase has over 170,000 lines of code written by five
programmers. (4 CT 950 at 44). According to published estimates, such a codebase
would typically have as many as 2,000 errors. Charles Fishman, They Write The Right
Stuff, FastCompany (Dec. 31, 1996)22 (describing NASAs 420,000 line long space
shuttle software written by two-hundred-sixty programmers as the gold standard because
the codebase had just one error, while [c]ommercial programs of equivalent complexity
would have 5,000 errors.). With five programmers employed by TrueAllele instead of
the two-hundred-sixty employed by NASA, it is not merely possible, but statistically
probable that TrueAllele has undocumented bugs. And as the experience with STRMix
demonstrates, those errors could materially impact the softwares conclusions.

B. Source Code Is Needed To Test The Reliability Of The

Governments Case

Without the opportunity to review the code to see that it is actually operating as
Cybergenetics claims, any error or errors, even those made by accident or in good faith,
will go unnoticed without competent, informed, expert scrutiny and testimony to counter
the testimony of the prosecutions expert witnesses. Hinton, 134 S. Ct. at 1090.
Not only does due process require meaningful adversarial testing of the
governments evidence, it also requires disclosure of impeachment evidence in the
possession of others acting on the governments behalf in the case. Kyles, 514 U.S. at
437-38; Brown, 17 Cal. 4th at 881 (prosecutions failure to review third-party crime lab
files and to disclose exculpatory evidence would be fundamentally at odds with the due
process imperatives of the disclosure rule); Youngblood, 547 U.S. at 869-70 (the Brady
duty extends to impeachment evidence); Sodersten, 146 Cal. App. 4th at 1225 (same).

20
Available at: http://www.mathworks.com/help/stats/release-notes.html.
21
Available at: http://www.mathworks.com/help/signal/release-notes.html.
22
Available at: https://www.fastcompany.com/28121/they-write-right-stuff.

20
Cybergenetics, having conducted its analysis at the behest of the State for use in the
States case, plainly is operating as part of the prosecution team. The State therefore
cannot evade its Brady/Giglio obligations by hiding the contents of the black box that is
TrueAlleles code; nor can it, given the potential for hidden or mistaken (to say nothing
of purposeful) software bugs, simply rely on Cybergenetics assurance that no such errors
exist. As discussed above, perfection in a codebase of this complexity and size would be
an aberration, and is not something that Mr. Johnson must take on faith.
Due process in our adversarial system requires insuring that a citizen not be jailed
because of a currently undetected (or detected but undisclosed) software bug in the
TrueAllele source code. The accused cannot meaningfully test the methodology without
access to the source code itself. Washington, 388 U.S. at 1819.
Because Cybergenetics keeps the code a secret, any errors will not come out
during cross-examination, even of Cybergeneticss own technicians, nor will they be
exposed, even through careful scrutiny, by a defense expert lacking access to the code.
Moreover, the significance of minor changes in a codebase can be enormous, as
examining TrueAlleles own updates shows. For example, a recent TrueAllele update
significantly impacted the match statistics generated, causing one to jump from a match
statistic of two billion to nearly six billion. Commonwealth v. Robinson, No. CP-02-
CR-0007777-2013 (C.P. Allegheny Cty. Nov. 19, 2015) at 103:3-105:14 (prior sworn
Perlin testimony).23 Dr. Perlin acknowledged at Mr. Johnsons trial that the code has
changed over time, impacting the match statistics and introducing potential errors (23
RT 4064: 18-20) (Dr. Perlin: Every now and then it didnt work as well if we got too
creative.). The source code is simply not flawless. If there is an error impacting
TrueAlleles application to Mr. Johnsons case, no one will ever know unless this Court
reverses the conviction and permits inspection of the source code under a protective
order.24

23
Copy on file with Amici counsel.
24
The above is true even if only unintentional bugs or errors are present. More calculated
sources of error are also far from implausible, just as Volkswagen recently acknowledged
21
 Defendants need to review the source code, for any impeachment to be
meaningful, is particularly clear from Cybergenetics ads that tout TrueAlleles ability to
turn[] inconclusive data into a match statistic strong enough for court, E-Brochure at 4,
and to yield results a million times stronger than competing interpretations of the same
physical evidence. Id. As discussed in Section II.B.2 supra, the discrepancies between
TrueAlleles and other companies match statistics on the same sample and even more
tellingly, between TrueAlleles own repeat testing of the same samples, which yield
strikingly wide variations leave at the very least room for considerable skepticism. Yet
effective cross-examination regarding these figures the typical means to confront
scientific evidence would be futile without access to the underlying code. The denial of
the defenses repeated request (made both prior to and at trial) to have the opportunity to
independently verify the underpinnings of an experts analysis on an issue likely key to
the conviction is an affront to the adversarial system of justice.

C. The Trial Court Erred In Allowing Trade Secret Protections To

Outweigh Due Process
1. It Was Error To Put TrueAlleles Commercial Interests Before
Mr. Johnsons Right To Challenge The Evidence Against Him

The trial courts holding that Mr. Johnson had not articulated the materiality of the
source code was error: as described above in Section IV.B, bugs in the source code are
statistically probable, and thus, as Mr. Johnsons counsel argued, he had a due process
right to see if the math works, in other words, to verify the algorithms based on which
he was ultimately convicted. 4 RT 431:20-432:27. Depriving him of the opportunity to
check implicates both due process and the Compulsory Process Clause.
Rules restricting evidence offend due process if they are arbitrary or
disproportionate to the purposes they are designed to serve, U.S. v. Scheffer, 523 U.S.

with regard to its emissions testing. The possibility of source code manipulation to
enhance a companys image e.g. to obtain the extraordinarily inculpatory statistics
TrueAllele advertises should be fair game for discovery, but can be discovered only
through source code examination by an appropriate defense expert.

22
303, 308 (1998). Due process was denied here when the trial court weighed TrueAlleles
proprietary interest over Mr. Johnsons due process and Compulsory Process Clause
rights. People v. Cunningham, 25 Cal. 4th 926, 999 (Cal. 2001) (rules restricting
evidence must yield to a defendants due process right to a fair trial and to the right to
present all relevant evidence of significant probative value to his or her defense.);
Chambers, 410 U.S. at 295 (Of course, the right to confront and to cross-examine is not
absolute and may, in appropriate cases, bow to accommodate other legitimate interests in
the criminal trial process . . . [b]ut its denial or significant diminution calls into question
the ultimate integrity of the fact-finding process and requires that the competing interest
be closely examined.)(emphasis added).
The trial court erred in failing to give proper weight to Mr. Johnsons
constitutional right to a fair trial as compared to TrueAlleles interest in keeping its
source code secret. As discussed below, any such commercial interest could easily be
protected in a manner far less dismissive of Mr. Johnsons right to access information
from which he might persuade the jury that the evidence [is] unworthy of credit. Perry
v. New Hampshire, 565 U.S. 228, 237 (2012) (The Constitution, our decisions indicate,
protects a defendant against a conviction based on evidence of questionable reliability,
not by prohibiting introduction of the evidence, but by affording the defendant means to
persuade the jury that the evidence should be discounted as unworthy of credit.
Constitutional safeguards available to defendants to counter the States evidence include
the Sixth Amendment rights to counsel, compulsory process, and confrontation plus
cross-examination of witnesses.) (internal citations omitted).
TrueAlleles interest in the secrecy of its source code25 is not more important than
Mr. Johnsons right to a fair trial. The Supreme Court has found that interests far more
weighty than TrueAlleles monetary one must bend in favor of the protection of a
defendants constitutional rights. For example, in Davis v. Alaska, the Supreme Court

25
This interest, moreover, as discussed infra, need not be lost or even materially
impacted, as the court could enter an appropriate protective order an approach to which
the defense expressly agreed in this case.

23
held that restrictions on a defendants ability to cross-examine a key prosecution witness
about that witnesss juvenile delinquency record were unconstitutional because the right
of confrontation is paramount to the States policy of protecting a juvenile offender.
Davis v. Alaska, 415 U.S. 308, 319 (1974).26 As the Court explained, the States desire
that [its witness] fulfill his public duty to testify free from embarrassment and with his
reputation unblemished must fall before the right of petitioner to seek out the truth in the
process of defending himself. Id. at 320. U.S. v. Nobles, 422 U.S. 225, 241 (1975)
(affirming exclusion of defense investigators testimony where defense refused to turn
over investigators prior report on Fifth Amendment grounds because the exclusion
merely prevented respondent from presenting to the jury a partial view of the credibility
issue by adducing the investigators testimony and thereafter refusing to disclose the
contemporaneous report that might offer further critical insights).
And where the Supreme Court has declined to require access to potential evidence
for a defendant, it has done so in circumstances far different from those here, as when in
Pennsylvania v. Ritchie it upheld the refusal of a child protective services agency to
disclose a minors records to a defendant accused of crimes against that minor. 480 U.S.
39 (1987). Here, not only is TrueAlleles commercial interest in its source code far less
weighty than the interest inherent in the confidential reporting by children of rape . . .
[and] incest, id. at 43, but that interest remains protected upon disclosure under the
circumstances contemplated herei.e., under a protective order. Indeed, unlike the
deeply personal and damaging information at stake in Ritchie,27 the private nature of

26
As discussed above, while the trial court did not expressly forbid questioning about
errors in the source code, such a right was nugatory absent access to the source code in
order to discover such errors. As in Davis, [s]erious damage to the strength of the
States case would have been a real possibility had [Mr. Johnson] been allowed to pursue
this line of inquiry. Davis, 415 U.S. at 319.
27
It also bears note that though the Court in Ritchie found the defendants due process
rights not infringed per se by denial of his counsels access to the confidential files, it
held that he was still entitled to know whether the [child protective services] file
contains information that may have changed the outcome of his trial had it been
disclosed, that is, whether it contained Brady materialand so upheld the remand for
24
which is destroyed forever upon disclosure, TrueAlleles commercial interest in the
secrecy of its source code can be protected through a protective order, as is commonly
done in countless intellectual property cases.

2. TrueAlleles Commercial Interests Can Be Protected Without

Violence To Mr. Johnsons Constitutional Rights

A protective order (limiting the scope of TrueAlleles software disclosure to

defense counsel and their experts, limiting its use to the case at bar, imposing limits on
when and where source code could be reviewed, and requiring destruction of copies of
such code at the conclusion of the case) would have protected both Mr. Johnsons due
process rights as well as any commercial concerns Cybergenetics may have regarding its
trade secrets. Counsel for Amici practice in the area of intellectual property litigation,
where on a daily basis they, and their colleagues and adversaries, handle numerous cases
in which such protective orders are routinely entered, applied, and enforced. They see
such orders in almost every case they handle. In their personal knowledge and
experience, technology companies ranging from Fortune 100 companies to Silicon Valley
venture-backed start-ups routinely rely on the protections such orders provide for their
most sensitive trade secrets.
Disclosure of confidential informationincluding source codeis a routine
procedure in civil patent infringement cases. The Rules of Practice for Patent Cases
before the Eastern District of Texasa court that heard over 40% of all U.S. patent
infringement cases in 201528require an accused patent infringer to make available for
inspection and copying: source code, specifications, schematics, flow charts, artwork,
formulas, or other documentation sufficient to show the operation of any aspects or

in camera review by the trial court of the file. Ritchie, 480 U.S. at 61 (ordering remand to
determine materiality of undisclosed Brady material: [I]t is impossible to say whether
any information in the [sought-after] records may be relevant to [Defendant]'s claim of
innocence, because neither prosecution nor defense counsel has seen the
information.).
28
DocketNavigator Analytics 2015 Year In Review at 19 (Available at:
http://home.docketnavigator.com/year-review/).

25
elements of an Accused Instrumentality. United States District Court for the Eastern
District of Texas Local Patent Rule 3-4 (emphasis added). This is true even where the
parties are direct competitors.
Indeed, the model protective order for United States District Judge Rodney
Gilstrapthe judge that heard almost 25% of all patent cases in the United States in
201529provides special protective measures for the disclosure of source code, including
review of the code only on non-networked enabled computers and accessibility only to
third-party independent experts.30 See 10. Such a protective order would protect
Cybergenetics financial interests by preventing inadvertent disclosure of any of its trade
secrets while also preserving Mr. Johnsons Constitutional right to a fair trial. Rockstar
Consortium U.S. LP v. Google Inc., No. 2:13-CV-893, 2014 WL 5831041, at *1-12 (E.D.
Tex. June 19, 2014) (source code provisions in a protective order); MobileMedia Ideas
LLC v. HTC Corp., No. 2:10-CV-112-JRG, 2012 WL 7783403, at *1 (E.D. Tex. Oct. 19,
2012) (same); E-Contact Techs., LLC v. Apple, Inc., 2012 WL 11924448 (E.D. Tex. June
19, 2012) (same). Surely, if source code at the heart of businesses like Google and Apple
is adequately protected by such protective orders, Cybergenetics can hardly claim that
such protection would be insufficient for its interests.
Indeed, Cybergenetics must have anticipated such disclosure requirements (subject
to a suitable protective order) when it chose to bring to market and sell TrueAllele to the
People of California, at a rate of $60,000 per license, (4 CT 952 at 61), with emphasis
on its use in litigation, where such disclosures are commonplace. Dr. Perlin likewise
chose to serve as a consultant for the prosecution in this case, and as an experienced
expert witness must similarly have anticipated discovery into the processes he developed
and that he advertises as providing evidence of criminal guilt or innocence.
In sum, having developed, used, and sold software that makes its undisclosed

29
Id. at 23.
30
Available at:
http://www.txed.uscourts.gov/sites/default/files/judgeFiles/Sample_Protective_Order_Pat
ent_Cases_%28July_2015%29.docx.

26
calculations the basis of sworn testimony in court, Cybergenetics and Dr. Perlins claims
of trade secret privilege under the evidentiary code must yield to Mr. Johnsons right to a
fair trial. To read the rules any other way would offend due process. Scheffer, 523 U.S.
at 308 (rules restricting evidence offend due process if they are arbitrary or
disproportionate to the purposes they are designed to serve).
But in any event, both interests can be protected by use of a suitable protective
order. Indeed, the California Evidence Code applicable to criminal proceedings has
detailed provisions for such protective orders. Evid. Code 1061, 1062.
Similar protective orders have been entered in California civil cases. E.g., State
Farm Fire & Cas. Co. v. Superior Court, 54 Cal. App. 4th 625, 651 (Cal. App. 1997)
(requiring disclosure of trade-secret software under a protective order). As discussed
supra, a federal court recently ordered the disclosure of the source code to another
genetic software, the Forensic Statistical Tool (FST), subject to a protective order.
Johnson, 15-CR-565, D.I. 67 (S.D.N.Y. July 18, 2016). Notably, the disclosure revealed
serious questions regarding FSTs reliability. N.Y. Times, Traces of Crime: How New
Yorks DNA Techniques Became Tainted, Sept. 4, 2017.31 There is no reason such
disclosure could not be safely accomplished in this case.

V. CONCLUSION
When testimony based on the results of computer source code is admitted as
evidence, due process requires disclosure of the source code, if timely requested, so that
the accused can mount a meaningful defense. To hold otherwise would allow convictions
based on a deus ex machina, elevating error-prone software to the status of an
unchallengeable oracle of innocence or guilt. Such an unchallengeable authority
occurred in Mr. Johnsons case, to the detriment of his defense and in violation of his
rights to due process. The Court should reverse Mr. Johnsons conviction.

31
Available at: https://www.nytimes.com/2017/09/04/nyregion/dna-analysis-evidence-
new-york-disputed-techniques.html

27
 CERTIFICATE OF WORD COUNT

I, Gerald B. Hrycyszyn, counsel for Amici certify pursuant to the California Rules
of Court that, according to the computer program used to prepare this brief, the word
count for this document is 6,677 words, excluding the portions described in Rule
8.204(c)(3).
/s/ Gerald B. Hrycyszyn
Gerald B. Hrycyszyn

28
 PROOF OF SERVICE

I, the undersigned declare that: I am over the age of 18 years and not a party to the
case; I am employed in the City of Boston and County of Suffolk, Commonwealth of
Massachusetts, where the mailing occurs; and my business address is 600 Atlantic
Avenue, Boston, Massachusetts 02210.

On September 8, 2017, I caused the following document(s) to be served:

APPLICATION FOR LEAVE TO FILE AMICI CURIAE BRIEF AND

AMICI CURIAE BRIEF OF THE INNOCENCE PROJECT, INC., THE
CALIFORNIA INNOCENCE PROJECT, THE NORTHERN CALIFORNIA
INNOCENCE PROJECT, AND THE LOYOLA LAW SCHOOLS PROJECT FOR
THE INNOCENT IN SUPPORT OF DEFENDANT-APPELLANT

on the following courts and parties, through their attorneys of record, named below, and
addressed as follows:

Clerk of Court
Honorable Gary T. Friedman California Court of Appeals
1415 Truxtun Avenue
Bakersfield, CA 93301
Paul D. Cadman
T.J. Brewer
Kern County Public Defenders Office Attorneys for Defendant-Appellant
1315 Truxtun Avenue
Bakersfield, CA 93301
Cynthia J. Zimmer
District Attorneys Office Attorney for Plaintiff-Respondent
1215 Truxtun Avenue
Bakersfield, CA 93301
Bill Ray Johnson
CDCR#AW7702
P.O. Box 5248 Defendant-Appellant
Corcoran, CA 93212

Laura Schaefer
Attorney at Law Attorney for Defendant-Respondent
934 23rd St.
San Diego, CA 92102

29
( ) BY HAND DELIVERY: I instructed a delivery service to deposit the said
documents to the California Supreme Court at the addresses listed above.

( ) BY MAIL: I placed said documents in a sealed envelope to all other parties and
the trial and appellate court listed above, with said postage thereon fully prepaid for first
class mail, for collection and mailing at a metered depository in San Francisco,
California.

( ) BY EMAIL: I emailed the said documents.

( ) BY FACSIMILE: I caused the said document(s) to be transmitted by facsimile

machine to the number indicated after the address noted above.

(x) BY FEDERAL EXPRESS: I placed a true and correct copy thereof in a Federal
Express(overnight) envelope addressed to the individual named above at the address
shown, and by sealing and delivering the aforementioned documents to Federal Express
in Boston, Massachusetts to be delivered by next day, on this date.

( ) BY PERSONAL SERVICE: I caused the said document(s) to be personally served

on the attorneys identified above at the above address on this date.

Furthermore, I declare that on September 8, 2017, I electronically served from my

electronic service address of [email protected], the following entities:

Office of the Attorney General [email protected]

Central California Appellate Program [email protected]

I declare under penalty of perjury of the State of California that the foregoing is
true and correct.

Executed on September 8, 2017 at Boston, Massachusetts.

/s/ Lisa A. Woodbury

Lisa A. Woodbury

30