Police Blotter: Court won't release breathalyzer source code

Police Blotter is a regular CNET News report on the intersection of technology and the law.

What: A Minnesota man accused of drunk driving says he should be able to review the source code of the Intoxilyzer 5000EN breathalyzer used by police.

When: Minnesota Court of Appeals rules on January 13.

Outcome: Request for source code rejected.

What happened, according to court documents and other sources:
If police think you're driving while intoxicated, they'll probably demand that you blow into a tube connected to a simple portable computer. Your breath flows into a chamber with an infrared light and a sensor that's designed to detect alcohol vapor through a process called infrared spectrometry.

The problem is that the calculations depend on what assumptions the programmer made, including the relationship between your blood alcohol content and alcohol in your breath, which can vary with the individual. In addition, the programmer could have made coding errors along the way.

Which brings us to a question that has bedeviled dozens of judges for the last few years: Should a defendant have the right to review the source code--the step-by-step programming instructions--to a breathalyzer that police and prosecutors are using against him?

To computer scientists, there's little reason not to disclose the source code to a defense expert who can review the logic that the device employs; otherwise, it's a little like being convicted based on the word of an inscrutable black box. Defense attorneys have been zealously pressing that point in court. But prosecutors say it would be too disruptive and burdensome.

Nicholas Swendra is one of those DUI defendants asking prosecutors to hand over the source code to the Minnesota version of the Intoxilyzer 5000EN, made by a private company called CMI of Owensboro, Ky. The Intoxilyzer relies on technology that would have been state of the art in the early 1980s: a dot matrix fluorescent display, an RS-232 port, a Z80 microprocessor programmed in assembly language, and less than 100KB of RAM. CMI insists the source code is a trade secret.

The Minnesota government submitted an affidavit from a forensic scientist saying the source code was accurate, while Swendra's attorney countered by pointing out congressional testimony and a flurry of other court opinions authorizing a peek at the Intoxilyzer's instructions. (Defense counsel's goal is to get the Intoxilyzer evidence tossed out of court, or to persuade prosecutors to drop charges.)

A district judge denied Swendra's request for additional information about the Intoxilyzer, saying that turning over the source code would be unreasonably burdensome and that the defendant could offer only speculation to support his argument that the source code is relevant. On January 13, a Minnesota appeals court agreed.

This will not be the end of the overall issue. Swendra could appeal, and other cases have reached different conclusions. For example, a July 2007 ruling from the Minnesota Supreme Court in a case involving accused drunk driver Dale Underdahl--reported by Police Blotter at the time--allowed the defendant to review the source code. A Florida court said in 2005 that state police can't use electronic breathalyzers as courtroom evidence against drivers. And another Florida judge ruled last month that the results of breathalyzer tests in more than 100 drunk driving cases will not be allowed to be introduced as evidence during trials.

Excerpts from Minnesota Court of Appeals' ruling:
The results of the Intoxilyzer breath test are presumed reliable under Minnesota statute. Because appellant stipulated to the Intoxilyzer's reliable administration, he bore the burden of presenting the court with "some evidence beyond mere speculation that questions the trustworthiness of the Intoxilyzer report." See Kramer v. Comm'r of Pub. Safety, 706 N.W.2d 231, 236 (Minn.App.2005).

In (the 2008 case called) Underdahl II, this court held that a finding of relevancy respecting the source code "must be premised on a showing that an examination of the instrument's software would show defects in its operation or at least would be necessary to determine whether a defect exists." We found that the party seeking discovery in Underdahl II failed to show relevancy because he furnished no evidence that the code would reveal testing deficiencies or that any deficiencies would affect test results and no evidence that the driver, who is permitted by law to test the machine, was deprived of access to any information that might be revealed by the source code.

As in Underdahl II, appellant made inadequate showings in the district court on the relevancy of the source code. Appellant theorized that, "[w]ithout the source code, the Intoxilyzer 5000 is just a machine that does nothing. The source code itself is absolutely relevant and material to the question of whether the test record in this case is accurate and valid." The district court found appellant's evidence (testimony on voting machine source codes, a New Jersey source-code report, and trial court opinions) unhelpful and concluded that "[p]etitioner has provided only speculation that the information provided by the Intoxilyzer 5000EN is even at issue in this case.... In this case and on this record, [p]etitioner is not entitled to additional discovery."

The district court acted within the scope of its wide discretion in its relevancy decision. It follows that we have no occasion to reach appellant's remaining argument that the court erred in addressing respondent's alleged possession of the source code.