This blog was originally published on LinkedIn by Phil Jang, VMware’s senior manager of physical security systems & technology.
Traditionally, the primary use of security video has been for investigations. The video was stored for a limited amount of time and used to review and investigate security incidents and potentially as evidence. Innovations in AI and computer vision, however, are changing the value of video data – and demand that we rethink why and for how long we store video data.
When it comes to video investigations, one of the biggest questions has been: how long should you store your video data? Most organizations store data for 30, 60 or 90 days. But why are these the accepted periods? What is the value of storing data for longer or shorter periods of time? Does it lose value after a certain amount of time?
These seemingly arbitrary storage periods are loosely based on risk. When video data was only valuable for investigations into physical security incidents, there wasn’t a need to keep data long after a security incident had occurred. But what if the data was more valuable than this?
In this blog, I’m going to discuss the role of video data in physical security, why innovations in AI and computer vision have made that data more valuable, and why these changes are demanding a new, smarter approach to storing security video data.
Security video: An untouched goldmine
Innovations in AI and computer vision have led to a breakthrough in physical security called “computer vision intelligence.” Now video data isn’t just temporarily valuable. With computer vision intelligence, video data is the new goldmine for physical security leaders – and the organization at large.
That data provides valuable contextual data that security leaders can use to make smarter operational security decisions, like whether making an investment into revolving doors will actually deter tailgating. That video data can also provide insights beyond the realm of physical security, including accurately understanding wait times and peak times at retail locations or gauging whether employees are actually using collaboration spaces or whether that space can be better used in another way in offices. When viewed through this broader lens, storing video data is no longer a cost center but a valuable source of insights for the entire company. In this new paradigm, arbitrarily storing data for 30, 60 or 90 days may not make a lot of sense.
Enter Smart Storage
Enter the new model: I predict storage of physical security camera data will be changed to Smart Storage – no longer loosely based on risk but based on risk and business intelligence needs. In security, this video data can now be used to make strategic security decisions that impact the safety and security of an organization beyond one incident or the next. Business intelligence use-cases have more to do with spatial intelligence. Spatial intelligence goes further than occupancy to not only tell you that a person is in a space but to give you insights into their behaviors. The applications for this spatial intelligence on businesses can be far reaching.
“Computer vision intelligence-enabled video data is already proving to have use cases beyond physical security. From providing insights into how people are interacting with their environment to understanding how employees are using office space. For instance, are employees really using collaboration space as intended to which coffee stations need to be restocked the most often? The sky is really the limit on the possible use-cases.” – Rob Kay, Manager, Technical Security Engineering & Reliability at Rivian
Smart Storage isn’t simply about retaining video data longer to enable more use cases with computer vision intelligence. The storage itself should also change. Right now, video data is a temporary asset to be dumped after a set period of time. Because the value of this video data is limited, we typically buy the cheapest storage possible. Large form factor spinning hard drives in a RAID configuration that is relatively inexpensive at $10 per terabyte. You tend to have some failures and less availability than more expensive storage, but the cost and quality is appropriate for the limited use-cases.
When that video data has been mined for security and business information by the computer vision intelligence, it becomes more valuable. Once that video data is more valuable, it is logical that we’ll also want better storage with no loss of data, full redundancy, and high availability (think virtual storage or cloud-based storage). The cost is about 10X greater than the storage we use now, but the data is also much more valuable – and the data will only get increasingly more valuable
Quick note: A side benefit of having computer vision intelligence mining this data, is that it filters out unnecessary video and has the potential to decrease the overall size of video storage requirements. We might throw the raw video data out sooner, but keep the mined data indefinitely. For example, I won’t need to save any raw video just because there was motion detected but would save the video if it was associated with some kind of security threat, whereas I would keep the mined behavioral data and save it indefinitely.
The greater value of video data for security and business use cases along with innovations in 5G and the need for less camera infrastructure will mean that organizations are incentivized to add more cameras. More cameras means more video data and in turn increases the need for high quality storage that can support AI workloads like those done with computer vision intelligence. Fortunately, the storage industry is innovating to meet these demands.
“Traditionally, hard drives for physical security had one limited purpose. The demands on storage weren’t great. Now, however, new applications for AI are changing the requirements for storing security video, increasing the importance of high-availability storage that is purpose built for AI use cases.” – Caleb Lau, Program Manager, global Security at Western Digital
AI is turning video data that we would normally delete after 30, 60, or 90 days into a valuable commodity that can have a huge impact on physical security – and the business at large. Computer vision intelligence has been the catalyst for this paradigm shift, and it requires a new approach to how we store security video data.