Urban surveillance is crucial for public health monitoring, public safety, and solving crimes. However, security cameras can easily pick up irrelevant information in the motion video feed, causing false alarms and a host of privacy concerns. A vital security tool used for safeguarding these sensitive data is privacy masking.
In this blog post, I’ll explain what privacy masking is in camera surveillance and the two main masking methods used.
Types of Surveillance Cameras
Video cameras can gather, transmit, and/or process data depending on application needs. Surveillance cameras come in two main categories:
- Analog cameras are used mostly in Closed-Circuit Television (CCTV) with a small digital footprint and few connectivity options. Analog cameras require the use of a Digital Video Recorder (DVR), which is the older version of Network Video Recorders (NVRs).
- Internet Protocol (IP)-connected cameras have become the staple for the surveillance market. IP cameras can connect to a Local Area Network (LAN), providing superior Internet of Things (IoT) versatility and the ability to be part of a unified security and analytics platform. LAN connectivity also significantly increases an IP security camera’s cyberthreat threshold.
What Is Privacy Masking in CCTV Cameras?
Privacy masking is a video analytics feature that applies one or more privacy filters to raw surveillance video streams or recordings from security cameras. Privacy masks are a means of protecting the identity of people, objects used for identity (e.g., license plates), or any activities captured by video cameras. With cam masking, the video recording or live stream is processed with a blur over objects, people, or environments.
Why Is Privacy Masking Necessary?
The goal of masking cameras is to enhance identity security. The end result of camera masking is capturing and tracking all essential objectives without infringing upon peoples’ privacy rights. For example, privacy masking can be used for General Data Protection Regulation (GDPR)-compliant CCTV services in public surveillance.
Moreover, camera masking saves storage space because it uses much less video data. This is especially important for operators that frequently run into surveillance capacity issues.
By 2030, ABI Research forecasts that 1.4 billion CCTV cameras will be used worldwide. As cities become smarter and install more security cameras, the role of privacy masking will only grow in prominence.
Camera Masking at the Source Endpoint
The first way to mask surveillance cameras is at the source level. In other words, privacy masking is done at either the endpoint video recording or the NVR. This method leverages the processing chip to process the raw video feed and blur any unwanted features that are pre-defined.
When masking a camera at the NVR level, the video recorder will transmit the processed video stream to a Video Management System (VMS), typically on the on-site server. This provides human operators with a live view of the surveillance footage, or the recording can be used later for video analytics.
When privacy masking of cameras is done at the source level, the following points must be weighed:
- A camera masked at the source endpoint involves altering the video feed. Therefore, the original unmasked version of the camera footage cannot be accessed by internal or external stakeholders, including law enforcement.
- All endpoint software must use the same hashing/filtering algorithm.
- The resulting format must be the same among all connected cameras across all locations and can be aggregated under the same VMS.
Camera Masking at the Video Management Server
The second method of privacy masking in surveillance cameras is performing the mask at the on-premises video management server. With this approach to privacy masking, the video recorder obtains raw footage from all connected cameras and creates two copies.
The first copy is the surveillance video recording with all the filtering elements over disputed objects, people, and features. This version is available directly to human security/public safety operators via a live, yet masked, stream.
But then there’s a second copy of the camera footage, which is an unaltered version. This recording would look exactly how you’d expect it to look without any privacy masking features applied.
The second, unchanged recording is almost always stored in a local edge server for enhanced security. This surveillance footage isn’t easily accessible by human operators and self-deletes when security protocols are imposed. Alternatively, the camera footage can be retrieved when a legal, governmental, or another authorized entity requests access to the original recording.
Final Verdict and Key Considerations
Before implementing privacy masking for surveillance cameras, it’s essential to consider the processing requirements for each application. While a suitable Central Processing Unit (CPU) is technically sufficient for running the hashing/blurring algorithm, VMS vendors recommend a Graphics Processing Unit (GPU).
On the other side of the coin, camera masking at the video management server is potentially superior to masking at the source. Masking at the video management server enables implementers to lift many of the restrictions of masking at the source, and it widens the scale of video analytics by storing the original, unaltered recording.
Unfortunately, privacy masking at the video management server introduces greater complexity. This increased complexity of masking cameras stems from the added access privileges, data access, and party authorization, as well as a shifting focus of processing power from the endpoints to the edge server. This shift in focus also requires more hardware investment.
To learn some other ways to keep surveillance cameras secure, download ABI Research’s Secure Computer Vision Usage in Smart Cities research report. This research is part of the company’s IoT Cybersecurity and Smart Urban Infrastructure research services.
