Google Cloud-powered monitoring and surveillance solution

Introduction

Despite the rising rate of digital and distributed operations, enterprises still need to secure their real-world operations. In this regard, surveillance capabilities take on an essential part of the broader security architecture of any enterprise. With the proliferation of the internet and connected devices, remote monitoring has emerged as a popular and highly useful solution. The increasing demand for privacy and security for users makes real-time identification of suspicious behaviour an enterprise necessity.

However, this rapid growth and the sheer abundance of recording devices have also brought forth an intense workload in the form of manually sifting through hours of videos and identifying threats and other incidents. With an ever-increasing workload, the complexity and scale of ensuring accuracy and security for an agile and efficient surveillance operation require next-generation technology.

Fortunately, Searce has been driving innovations in this arena to help enterprises leverage technologies like computer vision and facial recognition, and cloud automation for stringent security surveillance across a variety of locations and environments. In this blog post, we shall discuss our approach towards deploying automated video monitoring solutions with secure cloud capabilities.

Threats to Enterprise Security

While it’s easy to categorize normal and abnormal behaviour in most public locations, this approach needs to be further refined within an automated framework where human instinct and experience are missing. An effective surveillance system must include activity identification based on risk classification, which can be ingested by machine learning algorithms to determine authorized and unauthorized persons in sensitive areas. A key requirement for this kind of automated monitoring is facial recognition.

The rapid advancement in machine learning and visual image analysis has advanced the accuracy and speed of facial recognition systems significantly. Moreover, this capability is critical in ensuring the authenticity of external video footage as well. Recent advancements in image synthesis have endowed machines with the capability to create replica videos of real people with a high degree of accuracy. These “deep fake” videos would have seemed impractical only a few years ago but today they pose a genuine threat. While the positive aspect of this innovation is applied in media and entertainment, it also adds to the threat landscape of a modern enterprise.

Authenticating such videos is a significantly more onerous task that further endangers people and organizations alike. With advanced technologies, a suspected “deep fake” video can be authenticated by handling it through a mapping algorithm that records a collection of video information to a small string of text, or “face recognition”. This data can be saved and used to authenticate the video throughout its life cycle. While the video is playing, the mapped information is recognized, and it is confirmed as authentic when it matches the initial recorded information. This system can be deployed as a highly effective tool to detect any tampering with source video and inhibit falsification as a result.

Automating Facial Recognition with Google Cloud Platform

Given the challenges above, we have devised a framework that leverages the Google Cloud platform to ensure assured performance. This is accomplished by enabling automated facial recognition capabilities that are supported by cloud services like machine learning, security, and developer tools. With the Google Cloud platform, security leaders can create and assign roles with specific, transferable permissions that can be used to detect suspicious activities. However, in case of a positive scenario, Google Cloud can authenticate with employee metadata including the physical presence, and device/ system id which is transferred to the cloud platform.

We chose the Google Cloud platform as it can handle more than 3000 simultaneous video uploads. Other additional features include scaling, authentication, analytics and dashboarding, object detection, and much more. One of the most important benefits of a hosting system in the cloud is the capability to scale resources in response to increasing traffic or expected traffic peaks which could be specific to individual businesses.

The auto-scaling feature is achieved with the Kubernetes Engine (separate from Google Kubernetes Engine i.e GKE) which is a cloud service that provides a controlled system for deploying, administering, and scaling containerized applications by employing Google infrastructure. Since Kubernetes is capable of spanning hosts across private, public or hybrid clouds, it can be considered the perfect platform for hosting cloud-based applications which necessitate fast scaling.

Kubernetes employs similar design principles which manage popular Google services and has the same benefits including automatic administration, and monitoring probes for application containers, scaling, and more. However, in certain extreme cases, even this architecture can be too slow to scale with requirements.

Conclusion

With its myriad uses across access control, customer service, and healthcare, facial recognition-based automated video surveillance is rapidly gaining a steady footing across industries. The advanced applications in the domain of safety and security are only going to be icing on the cake. And in a digitally savvy world with a bulk of personal information already on the network, employing effective and efficient face recognition will have several positive implications. With constantly improving features, it can be deployed for wide-scale usage across different domains from public governance to plant security. With a secure, resilient, and reliable cloud platform supporting it, automated video monitoring can accelerate the adoption of facial recognition-based video monitoring for greater security and safety across industries.