Blogs

The intersection of body camera video with CJIS guidelines and privacy

General Manager, Safer Planet, IBM
Director of Public Safety, Law Enforcement, Emergency Management Solutions, IBM

When community relations and criminal data collide—as is the case with the video content captured by body-worn cameras—it takes only one video frame to capture valuable criminal justice information. For example, video and multimedia data may become law enforcement data, and potentially be governed by Criminal Justice Information Services (CJIS) guidelines, in some of the following scenarios:

  • An officer’s radio transmission is picked up on a camera audio track, and the information being relayed is about a background check and contains reference to felony data. 
  • An officer’s body-worn camera captures images of a crime being committed.
  • In the car, a body-worn camera captures images of the officer’s laptop screen with CJIS data displayed.
  • Either video or audio captures personally identifiable information (PII)—for example, when an officer reviews a driver’s license produced upon request.
  • Images of crime victims or minors are captured when officers respond to an incident.
  • The images and audio captured when the camera is operating within an area where the subjects have a reasonable expectation of privacy—for example, when officers respond to domestic violence calls.

Following CJIS guidelines adds complexity and cost to managing body-worn camera video. For example, when the public, media or attorneys request to view the video, the owning agency then has two legal requirements to juggle: deliver the content as may be required by regulations such as the US Freedom of Information Act (FOIA) and meeting CJIS guidelines. Addressing these two somewhat conflicting requirements can cause a major strain on agency resources that seems to far outweigh the costs of acquiring the cameras in the first place. 

What are some organizations doing to handle criminal justice information content?

Some law enforcement organizations seem to be undecided as to whether the video content from body-worn cameras should be considered criminal justice data and, therefore, if the CJIS guidelines apply at all. At this time, assessment is generally a state or local government policy decision. Through conversation with informed members of law enforcement, body-worn camera manufacturers, cloud providers and IT specialists in law enforcement, it is our observation that many agencies have concluded that body-worn camera video content should be treated as if it is law enforcement data and, therefore, managed within the same CJIS considerations as other evidentiary materials.  

If the video is to be stored in the agency’s own data center, which they believe to be following the CJIS guidelines, then very likely no additional data security issues need to be considered. However, the volume of video data that body-worn cameras generate, especially where department or state policy dictates longer retention cycles, means that storing and managing video content can place a huge strain on in-house facilities. That strain may drive the need for significant capital expenditure (CAPEX) for new servers and storage.

Consider that 30 minutes of video consumes somewhere in the neighborhood of 800 MB of data storage. Further, the average body-worn camera generally records about three hours of content per shift. Even with a relatively small number of cameras deployed, the storage and processing requirements can explode very quickly. Given the difficulty in obtaining grants for anything other than the initial acquisition of the cameras, this capacity requirement can add immense financial challenges to an already onerous CAPEX process for any agency.

To offset CAPEX issues, many agencies, and the body-worn camera vendors themselves, are looking to cloud services companies to provide services for storing and making body-worn camera footage accessible. The cloud providers deliver a service that is billed monthly and therefore supports an operational expenditure (OPEX) financial model. The OPEX model is the direction many agencies want or need to take to be able to afford to deploy body-worn camera technology for any meaningful percentage of their sworn officers.

Therefore, ensuring that cloud services providers are maintaining an environment that follows the CJIS guidelines to host all this potential and probable digital evidence becomes imperative. The natural conservatism of law enforcement agencies has caused some hesitation to accept that some of the cloud service providers can and do provide services that follow the CJIS guidelines. Agencies may want to consider requesting third party reviews of CJIS processes when considering cloud service platforms.

What about privacy considerations?

In some ways, understanding the trade-off between CAPEX and OPEX financial models for storage, implementing retention policies, and incorporating CJIS guidelines is the easy part.

Once the video from body-worn cameras is captured and stored, agencies need to be able to find it again in response to requests from the public, media or attorneys. The Freedom of Information Act (FOIA) and many states’ Public Information Acts provide for government transparency and public access to information. The availability for body-worn camera video is already proving to generate a high number of access requests. And as agencies deal with video libraries that grow rapidly camera by camera, shift by shift, the challenge can become to find needles in haystacks.

For example, according to the LA Times, the Los Angeles Police Department (LAPD) calculated that they needed over 100 extra personnel to manage video content from the approximately 7,00 body-worn cameras they were planning to deploy. The article reported that the additional officers would focus on] such tasks as training and poring over footage of use-of-force incidents and police pursuits. And [Police Chief] Beck said it would be “imprudent and irresponsible” to collect the body camera video and not review evidence that could be used in criminal cases or internal investigations.1 Clearly, this project scale is a significant cost and resource issue for LAPD.

As noted earlier, other types of content exist—both audio and video—that is expected to require redaction to help ensure privacy. The work to review, redact and prepare a video clip for public release has generally been performed by a time-consuming, manual, frame-by-frame process. This method is hardly productive police work in itself, and at the same time it is very expensive in terms of labor.

Fortunately, vision computing technology, or video analytics as it is more generally known, is becoming available to help reduce these workloads and to allow computers to assist in the work. IBM video analytics software has been used for the last 10 years to help search video from static cameras used for security, traffic monitoring, surveillance, and so on. With these types of cameras, which have a fixed field of view, the analytics do a ‘subtraction’ of the static, known background and focus on analyzing anything left in the frame as people and objects of potential interest.

Traditional video analytics algorithms can become ineffective when the whole camera field of view is moving.

A new generation of video analytics

IBM has recently broken through the moving camera barrier with a new class of detection base algorithms that can be effective when the camera is in motion, as it is with body-worn cameras. The new generation of video analytics identifies images and categorizes them (for example, person, vehicle, baggage) and then creates appropriate metadata related to those objects. The metadata can then be used to provide enhanced and faster searching into the content.

A new IBM solution, Intelligent Video Analytics on Cloud, uses the metadata to drive automated redaction2. If the identified object is a person, Intelligent Video Analytics on Cloud can find the frames with the best view of that person’s face, personal characteristics, clothing color and so on. That metadata is then used to drive automated redaction of that person’s image across one or multiple video files. Redaction uses masking techniques that include blurring, pixilation, blank mask, and gradient map. With Intelligent Video Analytics on Cloud, you can redact one or more faces from a video segment as well as redact all detected faces or specific objects or all pixels of a video segment. As a result, this process is expected to reduce the human labor factor and the cost of managing body-worn camera inventory.

Given the potential evidentiary nature, privacy issues and the growing volume of body-worn camera video being captured and stored, agencies should develop their local policies with consideration of state and federal guidelines and regulations. Some of the resources cited below can help shape the decisions about storage and management of video from body-worn cameras. And, a tool like IBM Intelligent Video Analytics on Cloud may help agencies manage the time and resource costs so that they can get the most out of their camera investments.  

Learn how to maximize body-worn camera video

Resources

Endnotes

  1. http://www.latimes.com/local/lanow/la-me-ln-body-cameras-police-officers-los-angeles-20151214-story.html (December, 2015)
  2. IBM Intelligent Video Analytics on Cloud release date in United States: June 28, 2016.