Today’s security and access control systems have improved significantly from solutions in the past. They include more advanced identity badges, hand biometrics and even retinal scanners. However, these systems still require time and effort on behalf of the person entering a secure area. They still have to swipe a badge, place a hand on a biometric scanner or look into a retina device. This process can be extended when a security guard has to verify their identity.
IT security and facilities personnel are faced with a challenge. They must balance entry to a facility being as easy as possible while maintaining a strong security posture.
World Wide Technology (WWT) and Amazon Web Services (AWS) believe that in the next couple of years technologies such as the Internet of Things (IoT), computer vision, artificial intelligence (AI) and machine learning (ML) will disrupt access to facilities.
The concept of frictionless entry focuses on leveraging these technologies, so that people can enter a space in less time and with fewer steps, such as needing to bade in.
Specifically, advancements in computer vision have created new solutions in facial recognition (FR) and license plate recognition (LPR). The use of these two disruptive ML solutions allow employees to access a facility securely and quickly without the burden of lost hardware tokens and human intervention that can cause long wait times and ultimately lead to a loss of productivity.
By binding FR and LPR together, WWT and AWS have created the Fast Lane solution, an integrated two factor authentication access solution that verifies a license plate against a face.
WWT has developed a fully integrated solution that uses the AWS platform for multi-factor authentication at a security gate, without specialized hardware tokens or other cybersecurity technologies. Leveraging advancements in advancements in computer vision, the solution combines driver (facial) and vehicle (license plate) recognition to deliver a cloud-based authorization system that ensures the driver and vehicle combination are authorized to enter the secure location. Once registered, drivers can simply pull up to the gate and proceed through without additional tools, effort or interaction with facility security.
The serverless system works by taking video feeds from the targeted location(s) directly from the camera into Amazon Simple Storage Service (S3). Once in S3, an AWS Lambda function is triggered to run still images through Amazon Rekognition or other third-party machine learning algorithms, such as OpenALPR, to pull details directly from that image to match against registered faces and license plates in seconds.
If both the driver’s face and license plate match the registered information, that driver is allowed to proceed. Any non-matching pair or low accuracy hit (defined by the user) will queue a guard for manual intervention.
The system also provides a cloud-based UI for reporting and analysis to ensure the effectiveness across multiple cameras and sites for a singular view across all entry points.
The solution was developed using the best practices of a cloud-native approach. The overall architecture was divided into three independent services all connected via application programming interfaces (APIs). The following services were developed to support the solution.
Image processing service
This service connects to the on-premise cameras and handles accessing, storing and organizing data coming from those cameras. The Image Processing Service is made up of two AWS services:
- Amazon Simple Storage Service (S3) for storing the raw and organized data. The raw data is still images coming from the cameras that trigger the Image Analysis Service. The organized data is used for time serialization and coordination across multiple cameras as well as display via the Web UI.
- AWS Lambda functions are used to coordinate with the raw data from the cameras and the Image Analysis Service.
Image analysis service
This service provides the tools for modular license plate and facial recognition based on data coming into the Image Processing Service. The Image Analysis Service can use publicly available AWS services, such as Amazon Rekognition, or third-party tools, such as OpenALPR, in manner that does not impact other services in the solution. The Image Analysis Service is made up of three core AWS services:
- Amazon Rekognition to run facial and/or license plate analysis through a managed service.
- Amazon ECS/EC2 as a managed service to run third-party facial recognition models, such as FaceNet, license plate analysis models, such as OpenALPR, or any other third-party party or custom model.
- Amazon DynamoDB to store outputs from the model for analysis and use within the UI.
The service comes with a custom, serverless UI to view live and past events within the solution. The Web UI is comprised of four AWS services:
- Amazon S3 for storing all web assets as well as cleaned, organized and analyzed data from the Image Analysis Service.
- Amazon API Gateway for accessing the web UI and coordination between services.
- AWS Lambda for handling UI requests and performing analysis against data generated by the Image Analysis Service to determine the status of drivers and license plates.
- AWS Dynamo for storing registered driver and vehicle information as well as data related to driver and vehicle match status used for reporting in the UI.
Scalable, efficient, accurate
By using the global scale of AWS, WWT is able to quickly deploy the Fast Lane solution across the U.S. with easily deployable IP cameras. No additional on-premise compute or storage resources are required. The Fast Lane solution returns results in seconds, with no human interaction allowing employees faster access to their work site while maintaining entry security and safety.