Breathalyzer source code must be disclosed

Florida police can't use electronic breathalyzers as courtroom evidence against drivers unless the innards are disclosed, a state court ruled Wednesday.

A three-judge panel in Sarasota County said that a defense expert must have access to the source code--the secret step-by-step software instructions--used by the Intoxilyzer 5000. It's a simple computer with 168KB of RAM (random access memory) that's manufactured by CMI of Owensboro, Ky.

"Unless the defense can see how the breathalyzer works," the judges wrote, the device amounts to "nothing more than a 'mystical machine' used to establish an accused's guilt."

The case, one of the first to test whether source code used in such devices will be divulged, could influence the outcome of hundreds of drunk-driving prosecutions in the state. So far, Florida courts have been split on the topic, with some tossing out cases involving breath alcohol tests and others concluding that the information about the machine's workings should remain a trade secret.

In one similar 1988 case, Florida defense attorneys discovered that the police had mechanically modified a breath test machine so much that its results were no longer valid and could not be admitted as evidence in a prosecution.

The Sarasota judges didn't require the public disclosure of the source code. Rather, they ordered that it must be given to a defense expert who will keep it in confidence and return it when his analysis is complete. That analysis could show bugs or reveal that the code was modified after the Intoxilyzer was certified for use by the state--meaning the device's output could not be used in court.

[yrrahxob]

Good idea but where will it end?

The thing we now need to look out for is when the defense wants the arresting officer's brain turned over to some so-called "Specialist" to see if the source code for his brain has been tampered with.

[nicmart]

Not to worry

If the defense were to investigate a police officer's brain, in most <br />cases they would find nothing.

[Hernys]

Why?

I don't understand the rationale.<br />If something can be proven to work on a double blind, audited test, what's teh point in knowing how it works? <br />If we declare double blind tests aren't enough prove then we'll have to demonstrate theorethically the workings of every simple drug on the market (something that's just plain impossible for even the simplest drugs).<br />This is especially futile since a good coder can code things so they appear normal to even the best programmers and still hide malicious behavior. The fact that it can be done on a short, 20 line program (see <a class="jive-link-external" href="http://graphics.stanford.edu/~danielrh/vote/vote.html" target="_newWindow">http://graphics.stanford.edu/~danielrh/vote/vote.html</a> for some examples) indicates that this should be even easier to do in a more complex program.<br />This seems to me like the consequence of a defense lawyer questioning the system as the means to take some drunk driver out of jail.

[thanhvn]

Defense has right to challenge judgment

whether that judgment is made by a human or a machine. If it was an officer who made the final judgment on whether someone is DUI or not, I bet most people would agree that it is reasonable for the defense to cross-examine that officer on how he/she came to such a judgment. Now, if it is a machine that made that same judgment, why can't the defense "cross-examine" that machine to see how it came to such a judgment?<br /><br />Most people would agree that it is difficult, if not impossible, to test a non-trivial piece software for ALL possibilities. Thus, just because a non-trivial piece of software has never failed a single test so far doesn't mean that we have 100% confidence that it is correct 100% of the time. With a finite chance for the machine to make error, no matter how small, I believe it is reasonable to allow the defense a chance to "cross-examine" the machine as we normally would if it were an officer.<br /><br />H. Sagues draws parallels between software and medicine. It's true that it is difficult, if not impossible, to prove that a program/drug is 100% correct/safe, and requiring proofs beyond a certain limit, such as double blind tests, is simply not practical. However, we are not requiring the vendor to prove that its machine is 100% correct; only that we allow the defense to prove that the machine is not 100% correct. Just one failed test (empirically or theoretically) will prove it, and thus it is within practical limit. (Whether one or more failed test is enough to establish reasonable doubt is up to the jury, and thus outside the scope of this argument.) <br /><br />Thus, allowing the defense to examine the source code not only serves justice and but also is within practical limit. Now, most defendants, whether truly guilty or not (that's up to the defendant and God to know), use the same tactics to win a case: tearing down the plaintiff's argument(s) and supporting evidence(s). Denying a defendant a fair chance to do that to make his/her case, no matter how guilty he/she may appear, definitely does not serve justice.

[VI Joker]

Because

...it is not about the originally built machine. Its about those instances where an officer has made a modification to the breathalyzer that allows them to arrest someone who is not under the influence. I am not an advocate for drinking and driving, but I would not want to be arrested and my BAC is below the limit. The defense has every right to review and analyze the methods, pratices, and tools used to arrest their client. However, I the lawyer should reveal anything outside of faults and tampering with the device.

[zaznet]

Reasonable Doubt...

I agree they need to establish a way to test that does NOT involve disecting the code. The method is fairly simple. You have the device provide a visible and audible signal that it has tested "zero" before the test is administered. You test zero on video (installed in most police cars). Then administer the test. The results are also registered on video with visual and audible reading. The police officer then reads alloud the test results, again for the camera. This is ended with another zero of the device to show nothing is interfering with the test. The device needs to zero, without a zero'd reading you can't prove that the device was not defective or tampered with.<br /><br />The thought that a Police Officer could some how tamper with the source code (that he doesn't have access to) in order to produce "false positive" results should be countered as "NOT reasonable". Possible, sure in 1 of 10 million cases, but that is hardly reasonable for a basis of defense.

[heystoopid]

Logical decision

the only logical decision possible!, for we don't want another diebold fiasco do we!

How is this related to Diebold?

I agree with the need for transparency, esp. in the criminal justice process where potential deprivation of liberty is at stake, but I fail to see how the Diebold 'debacle' is related - drunk driving is hardly as fundamental right as important as the right to vote.

[parich1776]

Bah!

The software *will* have bugs of some kind. This is a statistical fact and a characteristic of software in general. And a Defense *expert* can always argue that if the code had been written that way rather than this, the device would be more accurate/bug free/reliable/blah. The next obvious question is whether or not the bugs (read imperfections) are relevant, if so how relevant, etc. Meanwhile at least 1/12 of the jury, most or all of whom regard software as an obscure black art, will drift off into utter confusion. And this is the real objective.<br /><br />Device manufacturers cannot reasonably be expected to exhaustively test and provide lifetime certification for every single unit produced. Nor can the State be expected to pay for endless trials in which each unit must be submitted to an *expert* to get an opinion on whether or not it is correctly constructed, especially if it contains software. And if a new version of the software is introduced into the field, does the comedy begin all over again?<br /><br />The software question is a red herring. The relevant question is, Does the unit perform within specifications? And there should be a rapid and inexpensive means of answering this question. Speed guns have been down a similar road. Then the argument can shift to specifications which pertain to all units and away from ambiguous questions pertaining to a single unit.<br /><br />What we have here is variant n on the OJ defense regarding DNA. "What? You mean there is a possibility that the DNA belonged to someone other than the accused? How can you say there isn't some doubt?"<br /><br />The law imposes a vague, undefined term called *reasonable doubt* as a basis for jury verdicts. Smart defense lawyers, and not so smart, have learned that if they can confuse just part a jury sufficiently - not so difficult these days - they can parlay the confusion into doubt and a not-guilty verdict or hung jury. Defense then becomes primarily a matter of introducing a sufficient amount of chaff into the proceedings to create confusion. <br /><br />The core issue in a jury trial is never really detecting *true* guilt or innocence. It's effective manipulation of the jury. That's a principal reason why defendants request jury trial in the first place rather than submitting arguments to a judge. So here we have just another instance of tired, but often effective, tactics disguised as a search for truth and justice.

[zaznet]

You are right, but...

You are right, but there still needs to exist a way to kill "reasonable doubt" when it comes to the device in question. They need a test of the device, NOT a review of the instructions written in some programming language.<br /><br />Few on a jury will be able to reasonably understand the implications of any software imperfection that could be found. Most will understand a self-testing device and be able to see that evidence in a video taped test using the device.<br /><br />Lacking video evidence of testing for a drunk driver in any trial when such video cameras are very common is just asking to have the case thrown out.<br /><br />Let's face it, he didn't get pulled over because he was a safe driver. He most likely did something that got the officer suspecting alcohol to begin with.

[zaznet]

Source code or proof it works?

Judges should not be forcing open source code to outside "experts" working to discredit a system.<br /><br />They should be forcing those using the systems or manufacturing them to be able to demonstrate that they WORK. In such a demonstration they should show the principles behind how it works and that will show logic flow of the program, which should be enough.<br /><br />Next time I get a speeding ticket I'll ask that the "source code" to the radar gun be disclosed to my lawyer... It just opens up a BIG can of worms and sets an entirely new precident that we don't want to see bog down the courts.

[Earl Benser]

Three techncally blind mice...

... currently labeled as 'judges', were stupid enough to fall for <br />the defense attorneys' desperate gambit. The source code is <br />irrelevant; what counts is performance. Is the breathalyzer <br />accurate or not? There's where the attention should have been <br />placed, <br /><br />I suppose that DWI drivers need legal rights, like the rest of us. <br />But DWI's should be treated more seriously, with loss of the <br />automobile and temporary suspension of a driver's license on <br />the first offense, loss of driver's license and automobile on the <br />second offense, and loss of the automobile and a long jail time <br />on the third. <br /><br />Too bad we can't hang the defense attorneys who get a habitual <br />DWI driver off the hook.