ALPR and Digital Civil Rights

Once again my fight for digital civil rights has landed on the front page of the Times Colonist, this time in relation to the ALPR (Automatic License Plate Recognition) surveillance system. I highly recommend reading the commissioner's report, which you can find at

The report goes into great detail about the ALPR program and is derived from a lot of information that our research group has not been able to obtain under the freedom of information access processes -- this despite repeated requests for all documents of all types relating to the ALPR program. (Rob Wipond reports that he currently has 6 complaints before the federal information commissioner)

The report has learned that non-hit data (data including the movement patterns of innocent Canadians) is being acquired and shared outside of the BC jurisdiction. It also makes crystal clear that where local police collect the information, they are in custody of the information and are subject to FIPPA regulations on their handling of that data. This includes not storing and not sharing any data that is, after-scanning and comparison to an on-board hotlist, no longer useful for policing purposes.

The Commissioner's report also reveals a new data point which we were unable to access. Obsolete Hits. These are hits that are valid at some point in the database, but that are no longer valid when the vehicle is scanned. The commissioner's report suggests that these false-hits cannot be shared with the RCMP either. This requirement alone is a huge win for accountability of this program, as it will mandate the review of each and every hit produced by the ALPR system before it is shared with the RCMP or used for secondary purposes. This should return the ALPR system to being a useful convenience tool for police plate scanning, but will remove the dragnet surveillance capability of the system as it will likely necessitate manual review of the data produced.

That said, I was disappointed that the commissioner did not engage in an analysis of the confidence rating of the system as a whole. With accuracy rates claimed in the 70-95% range for ALPR systems more generally, they have the potential to generate tremendous amounts of false, incorrect, information that will be used against people. The commissioner's report gives us two data points that are hugely valuable in this regard however. For every 100 scans, only 1 is a hit. In a 95% accurate scanning system, 5 scans in 100 will be inaccurate. The report also states that 4% of hits are obsolete hits, further reducing the confidence rating of the resulting data produced. A Bayes Theorem Analysis of the overall system's data confidence rating is definitely needed, but will require significant resources and access to do properly. The initial data however, suggests that the system may produce significant volumes of incorrect data, and confidence ratings may be low enough to call into question the entire program, even when only discussing the hit-data context. Certainly ALPR's use as an evidence-generating tool for court-purposes will be easily challenged and investigations that start from ALPR data may be subject to the poisonous tree doctrine in certain jurisdictions.

Overall, I'm thrilled with the report. It validates what I and my colleagues have been saying about police use of surveillance tools and is an incredible study into what these programs actually look like in practice. The data in the other-pointer-vehicle category, as just one example, shows just how broadly these programs are being applied. It also draws into question many previous statements by the authorities on the scope of the ALPR program.

I look forward to Victoria Police and the province fulfilling the report's recommendations on disclosure and access to information. Sunlight is always the best disinfectant.