Submitted by Darren Smith, Guest Blogger
Recently the FBI’s “Next Generation Identification” project was funded by congress to bring significantly enhanced identification and recognition capabilities to government agencies. The system relies for the most part on individual data collected from local law enforcement and state departments of motor vehicles or licensing, that is fingerprint, booking photos, and most recently driver license photographs.
With the advent of greater storage, computational, and recognition technologies the ease of facial recognition has increased. Coupled with the impetus on the federal level, reportedly, to identify terrorists and criminals the technology is now being utilized under a fundamentally different approach.
Identification data has historically been collected for decades by local law enforcement agencies and jails and forwarded to the U.S. Government. Most of this has traditionally been fingerprint cards. The cards formerly were classified by general fingerprint characteristics and then a paper hardcopy was kept on file. The AFIS (Automated Fingerprint Identification System) later enhanced this ability by permitting individual fingerprints to be both categorized uniquely and stored electronically.
The main limitation on whether a latent fingerprint submitted for analysis was whether the matching person had been previously fingerprinted either for a crime, had applied for a security clearance, for a position as an employee of a government agency, or had been arrested for a criminal offense meeting the FBI’s submission requirements. Other factors are a measure of time and quality of the latent print submitted.
When the primary identification was persons’ fingerprints it was effectively impossible to identify an individual who was merely in public view. The ability is now there by simply photographing the face of this individual. But rather than collecting fingerprints, of certain security cleared individuals or arrestees, all persons subject to having a driver license or state ID card now are to be gathered.
One facet that needs to be understood is that there is a general disconnect between federal and state agencies. Presently, the federal government only directly collects photographs on individuals when they apply for a passport, an immigration card, military, government employees, and a few others the general public is not subject to requirements of. Yet, the individual largely has photographs collected by state departments of motor vehicles or licensing.
Another example of a disconnect was state governments did not often collect social security numbers for driver license data that the federal government had. The federal government mandated the states collect social security numbers for each driver license applicant. This had an auxiliary (or intentional depending on opinion) benefit of states being able to have social security information for parents absconding on child support and provided linking information for the federal government to state licensing data.
State governments have comparative advantage and resources to collect driver’s photographs the federal government lacks this but has vast funds and resources to process these photographs.
Looking at the laws of one state as an example, Washington provides for a facial recognition matching system to be operated by the Department of Licensing.
Pertinent excerpts of the subject code (RCW 46.20.037) are as follows:
(1) The department may implement a facial recognition matching system for drivers’ licenses, permits, and identicards. Any facial recognition matching system selected by the department must be used only to verify the identity of an applicant for or holder of a driver’s license, permit, or identicard to determine whether the person has been issued a driver’s license, permit, or identicard under a different name or names.
(4) Results from the facial recognition matching system:
(a) Are not available for public inspection and copying under chapter 42.56 RCW;
(b) May only be disclosed when authorized by a court order;
(c) May only be disclosed to a federal government agency if specifically required under federal law;
The purported reasons for gathering facial data for all persons applying for a driver license or identification card is to ensure that an individual does not fraudulently apply for a driver license assuming a false name or identity.
Yet one could ask if there was more to this system than meets the eye.
One system could instead require citizens to provide a biometric retina scan, unique to each individual, to ensure the person applying for or renewing a license is the intended applicant rather than a publicly recognizable datum such as their entire face. Would the retinal scan suffice for the purposes of driver license security or could it be more all-encompassing?
We could look at what a state agency is required to comply with to obtain a recognition metric: a court order. A court order would have to comply with a narrow tailoring, specific to an individual for a specific purpose as required by any other court order such as those for search warrants and the like. Yet, with regard to the federal government it is construed broadly where the requirement is simply “federal law.” Could this be that the requirement is the United States Code or could it be administrative laws such as agency rules? Under the Revised Code of Washington it is open-ended because the federal law can simply be enacted to require “All licensing photographs of all persons must be submitted to a designated federal repository.”
A short primer on facial recognition might be helpful to understand what is involved. An image of a face is sent through an algorithm to locate certain aspects of a human face, such as proportionality of eyes to nose size, color of eyes, facial structure and such. This data is hashed into a string of data that represents the structure of this face. This creates a framework where using other matching algorithms can be used to compare a subject face to a database of hashes to determine probability of match between persons having similar hashes. The complexity of which can grow with time and accuracy will improve, even accounting for noise such as blurriness of the photograph, and various positions and attitudes of the image.
There might also be differing encoding systems that could produce hashes or data types. Using different systems could produce artifacts between each other where a person might match one set of criteria but not another in a different system. How is the system going to protect the individual against false positives or negatives?
As media and data collection increase the accuracy and prevalence of systems will increase sharply as Moore’s Law has shown through computational power generally. The technology will eventually come to fruition where a federal agency can take images, either as video or digital photographic in nature and place these into facial recognition software for quick identification of each person in a crowd.
It is rather easily understood the system is being created that can have the ability to take gigapixel photographs of large crowds of people and identify nearly every person in the photo having their face visible to the photographer. Is this something a person should have to be concerned with? It can be argued does the government have a true interest in identifying each person in a crowd of persons, whatever lawful activity they might be engaging in? Could this be a chilling effect?
The U.S. Secret Service routinely takes extremely detailed photographs of crowds of people who attend speeches and events their protected persons are hosting, such as presidential addresses and such. The service attempts to scan these crowds for persons who have made threats to presidents or provide security risks. That same practice can be greatly enhanced by the facial recognition system that can identify most of the persons in the images and analyze the names of who has attended and look for patterns with other events.
Our society is going to have to ask itself some fundamental questions on the role of government and how much citizens have liberty and privacy when our government has made a significant effort to obtain the likenesses of every person possible. Do we as individuals want to be held to draconian standards where persons could be questioned for “why were you here” or “why were you talking with X”?
Formerly it was a common Hollywood movie saying to “round up the usual suspects”; that is criminal elements likely to be involved in a particular crime. Are we as citizens now collectively “the usual suspects” since we eventually can all be identified and tracked? It will certainly be useful to apprehend actual criminals but watching the watchers is going to be more important as well.