Review: Alcotest 9510 source code flawed, challenged in courts

The Alcotest 9510, a breath testing machine used by numerous police agencies, has been challenged in court in several states. Now, a report from two software engineering consultants who reviewed its source code has been distributed. That report is no longer available for intellectual property reasons, but the consultants expressed serious concerns about the reliability of the test results.

In 2008, the New Jersey Supreme Court tested an Alcotest breath machine using the same underlying code. Although the court found it “generally scientifically reliable,” it ordered law enforcement to make some critical changes to the configuration. Although defense attorneys contend that the changes were never made, the court reaffirmed five years later that the machine was reliable and its results were admissible in court.

Last year, however, Massachusetts found that all but two of 392 machines examined statewide had not been properly calibrated. Several years’ worth of tests were ruled presumptively unreliable, threatening thousands of DUI prosecutions.

A Massachusetts defense lawyer and a computer forensic expert then obtained the Alcotest 9510’s source code and analyzed it. They found that, even properly calibrated, the machine could produce inflated results under certain circumstances. They also found a variety of programming errors in the code, including evidence that a fuel cell used for one sensor degrades over time and the source code fails to correct for that — risking inflated results. A court rejected those findings and held that the machines, properly calibrated, were reliable enough to be used in court.

This latest report came about from a 2015 DUI case in Washington State. In that case, two software engineering consultants identified a number of source code errors that could lead to inflated test results. They noted the fuel cell degradation issue like the experts in Massachusetts had done. They also found several other errors that could tip breath tests over the legal limit.

For example, it is known that a change in breath temperature of 1 degree Celsius over standard can skew the reading upwards by six percent. In Washington State, however, the state patrol had not even installed a component to test breath temperature. Moreover, a quadratic formula in the source code meant to correct for the breath temperature problem “may be insufficient” because the formula appeared faulty.

These events highlight a major issue with breath testing machines overall. Because they are manufactured for profit, their source code is a closely held secret and the machines are not widely studied or validated by peer review.

The Alcotest 9510 is not the only breath testing machine being challenged over possibly inaccurate results. People accused of drunk driving should work with attorneys who know how to challenge breath tests.