DNA has been considered the gold standard of forensic evidence for over 30 years, even as various types of unwanted science have fallen into disuse. But in recent years, police and crime labs have broadened and broadened the way they use genetic material to pin suspects – from accessing private ancestry websites to creating police sketches of the faces of suspects.
The latest practice to be examined is an obscure technique, “probabilistic genotyping,” which takes incomplete or otherwise impenetrable DNA left at a crime scene, often in tiny amounts, and performs it through software that calculates its probability. . come from a specific person. One of these programs, TrueAllele, has been used in more than 850 criminal cases over the past 20 years. The problem? No one knows if it works – the code, developed by a private company called Cybergenetics, is proprietary.
Government criminal labs that use the software do not have access to the program’s source code. Cybergenetics employees do not have access to it. Even the authors of the peer-reviewed studies of TrueAllele never had access to the code.
But now two criminal cases – one in US District Court for the Western District of Pennsylvania and another in the New Jersey Superior Court, Appeals Division – may give the world a first glimpse of TrueAllele’s secret algorithm. . Last month, the New Jersey judge ordered prosecutors to turn over the source code for TrueAllele, and a few weeks later, the Pennsylvania federal judge did the same.
Experts say the program is so complex and so hidden that software bugs are inevitable.
“It is virtually certain that there are flaws in the TrueAllele software,” wrote Mats Heimdahl and Jeanna Matthews, two computer experts, in a statement to Federal Court. “On average, there will be six vulnerabilities for every 1,000 lines of code, and TrueAllele has 170,000 lines of code.”
While the ordinances do not make the source code publicly available, they allow the defense to bring in experts to examine the software for possible flaws and inconsistencies – under strict nondisclosure agreements. . If experts find any problems, defense attorneys could try to have the DNA evidence rejected, weakening the cases against their clients and potentially leading to a wider calculation for TrueAllele.
[Read: How do you build a pet-friendly gadget? We asked experts and animal owners]
The operation of the software is much more complex and difficult to reproduce than the traditional DNA testing process, which matches a suspect with a robust type of evidence, such as blood.
TrueAllele does not match a suspect with physical evidence; he calculates the statistics probability that a person’s DNA is present in a complex mixture of DNA from several people, or in a tiny amount of DNA left behind – for example, after someone has simply touched something.
In the Pennsylvania case, federal prosecutors used the software to analyze a complex mix of DNA found on a handgun left in a car – and more specifically, to calculate the likelihood that one of them belonged to a man from Pittsburgh named Lafon Ellis. Cybergenetics experts testified that the sample likely contained DNA from four people and that it was “21.4 trillion times more likely” that Ellis’ DNA was in the mix than that of one. other random African American. Ellis ‘attorney, however, says the gun is not Ellis’ and disputes the DNA evidence that links him to it.
In addition to prosecutors, defense teams also used the software to help prove their clients’ innocence at trial or exonerate them later. Cybergenetics made the program available to both sides in all cases for testing.
According to Cybergenetics, TrueAllele’s calculations have already been admitted into evidence in 14 states, with 20 unsuccessful attempts in recent years by criminal defense teams to access the program’s source code. Prosecutors have long argued that the program was based on widely accepted mathematical concepts and that opening its source code to public scrutiny would threaten the company’s trade secrets.
Lack of software control
But experts and civil rights activists have long worried about the software’s lack of control. Their arguments may gain ground, as evidenced by the two decisions last month.
“Our justice system cannot authorize convictions based on secret evidence,” the Electronic Frontier Foundation and the American Civil Liberties Union of Pennsylvania wrote in an amicus brief to the Federal Court. “There is a long history of unwanted science employed under the guise of technological advancement in criminal matters – and of public access to this evidence and its analysis as a means of its eventual invalidation.”
From bite evidence and blood spatter analysis, ballistic testing and fingerprint matching, many forensic methods that were once common in criminal labs and courts were later called into question. or even totally demystified after being subjected to an external examination.
Life and death software – like that used in medical equipment or airplanes – must undergo independent validation and verification processes, but DNA software does not. There are few federal regulations on how police or crime labs introduce new technologies and methods into solving crime. Each state has its own rules for how the use of a new type of forensic technology is approved, and judges tend to follow past decisions when deciding whether or not to allow new scientific techniques to be approved. become evidence.
TrueAllele’s review will likely be complicated – its inventor estimates it would take a person, reading 10 lines of code per hour, about eight and a half years to complete (which defense attorneys dispute) – and how to go about it. has been another source of friction between prosecutors and defense lawyers.
Originally, the defense teams in both cases were surprised to learn that their access to the source code would be limited to a read-only iPad and that they would be allowed to use only pen and paper to take pictures. notes. They have since lobbied for access to an electronic copy that they can run and test themselves.
While TrueAllele has never been subjected to such close scrutiny before, two of its competitors have.
When a federal judge in the United States District Court for the Southern District of New York ordered source code access to New York’s proprietary DNA analysis software, Forensic Statistical Tool (FST) in 2016, the examination revealed a serious flaw that[ed] to overestimate the probability of guilt. The judge in that case ultimately lifted the software protection order, and ProPublica obtained the code and posted it on GitHub, allowing researchers and the general public to review it for themselves.
The New York City Office of the Chief Medical Examiner replaced FST with STRmix, a widely used competitor to TrueAllele. As it turns out, the makers of STRmix have also recognized software bugs that have affected 60 cases in Australia.
“Software errors are common, and forensic software has no special immunity against bugs and errors that affect software in other areas,” wrote Khasha Attaran, Deputy Federal Public Defender representing Ellis, in a statement. press release sent by email. “Constitutional principles and fairness require access because an accused must be allowed to review and challenge the accuracy of software generated evidence on which the federal government relies. “
Federal prosecutors can appeal the decision to the United States Court of Appeals for the Third Circuit, but have not yet indicated whether or not they would. The New Jersey case can go to the state Supreme Court. The Ellis District Attorney’s Office for the Western District of Pennsylvania and the Hudson County District Attorney’s Office, which brought the case to New Jersey, both declined to comment on the pending cases, just like Mark Perlin, the founder of Cybergenetics and the inventor of TrueAllele.
This article originally appeared on The Markup and has been republished under the Creative Commons Attribution-NonCommercial-NoDerivatives license.