Subscribe to Our Newsletter
Sign up for our bi-weekly newsletter to learn about computer vision trends and insights.
How to Preserve Data Privacy With Computer Vision
Elizabeth Alves
Data collection is crucial for computer vision systems to operate successfully. However, the images captured, videos recorded, and data stored by computer vision systems can present challenges for the privacy protection of personally identifiable information (PII). Issues of privacy and the use of PII are not new when it comes to Big Data.
This article describes how artificial intelligence and computer vision use data, how people may encounter computer vision, how businesses and other entities can use the technology, and how companies can use platforms like alwaysAI to create secure, privacy-preserving computer vision applications.
Privacy in the Age of AI
Artificial intelligence and computer vision are becoming increasingly ubiquitous with government and public sector applications as they have many benefits. For instance, governments use computer vision applications to build smart cities to monitor traffic, road conditions, and public events. However, in the age of AI, computer vision systems also pose a risk to users because governments and law enforcement can misuse them to violate people's privacy.
For both the public and private sector industries, the impact of computer vision is so prevalent that the public might not even be aware that they are being exposed to computer vision. The use of AI-guided camera systems is everywhere. For instance, we encounter a surveillance camera tracking our movements whenever we visit a convenience store, restaurant, or supermarket. However, it’s likely that people already expect to be captured on video when out in public.
Such camera systems are becoming increasingly sophisticated, with new technologies providing greater video analytics and business intelligence. Businesses want to offer people better and more personalized experiences, ultimately helping the company succeed and generate more revenue. To accomplish that goal, computer vision needs to continuously amass more real-time data, including human behavior, to provide deeper insights and predict future behavior.
Many employees also encounter computer vision and video analytics tools in their work environment. Businesses use these tools to monitor and improve their business operations, so they must consider how employee privacy is affected and how to manage that data.
The responsible use of computer vision can greatly improve people’s day-to-day lives. But, without addressing the risks to data privacy, AI and computer vision may have negative impacts.
Concerns about data privacy should not stop any business from implementing this game-changing technology. It just means developers need to take a proactive approach to preserve privacy. That’s a huge benefit of using a computer vision platform like alwaysAI. We’ve developed integrated tools to help companies proactively address privacy in their computer vision applications and ensure world-class security of data. Usually, developers overlook privacy if it is difficult to implement unless it's required by law. We remove that hurdle and ensure that privacy doesn’t become an afterthought.
How Does Computer Vision Utilize Data?
Computer vision systems collect and process real-time visual data captured through cameras. The video feed is first recorded and sent to a system on-site, edge devices, or a cloud-based storage system for data processing and analysis. Then, computer vision applications perform functions such as people detection, object detection, and counting by processing the raw data using deep learning models.
In addition, computer vision platforms help analyze the data to provide more insights into business operations. For instance, a retail computer vision application could count how many people enter a store, track where they go, and how long they stop at a particular display. This information helps the stores optimize product placement, layout, staff levels, etc., to provide a better consumer experience.
There is also a big push to deploy computer vision on the edge. The use of edge computing brings a lot of benefits for both processing as well as data privacy. Not only does it run in real-time, providing more valuable data, but it also does all of the processing of data – or inferencing – locally on a device. This means no data is sent to the cloud, eliminating any possible transmission of PII, this is one of the reasons alwaysAI deploys on the edge. The only thing that could be accessed is the metadata which is completely anonymized.
However, cloud processing sends that data off-site to a central server. Third-party providers maintain many cloud storage systems, and companies do not have as much visibility on how that data can be accessed and used. Thus, data on the cloud brings privacy challenges to computer vision developers.
Additionally, data sent to the cloud is often stored for at least some time according to data retention policies or government regulations. Most companies also have policies on who can access the data and what they can do with it. Depending on such policies, third-party cloud service providers or vendors may be able to access that data, share, or sell it for marketing purposes - increasing the potential for data privacy abuses.
AI Data Privacy Concerns
Computer vision needs data to provide the real-time insight required by companies. Still, they can intentionally or accidentally collect sensitive data that can be tied back to a person and used to identify them. For instance, suppose there is a facial recognition system that can associate a face with a license plate, email address, or credit card number. In that case, that system can identify and track someone as they move in public spaces.
Some companies use facial recognition in the workplace, which could pose a risk to employees if any data breach occurs. If you are storing PII data, it is important to be concerned about how that data is stored and if other parties can access it. If the data is publicly accessible, AI tools can scrape and analyze it.
Most people expect privacy and ethical use of their data, so privacy with computer vision is a growing concern. But there is a constant push and pull with maintaining privacy and how businesses use data. Consumer privacy is the topmost priority of many companies, however, their decisions and boundaries are often still driven by regulations, with governments most often at the forefront of driving how companies can and can’t use personal data.
Many data privacy laws, such as the California Consumer Privacy Act (CCPA), require companies to acquire explicit consent from people if they use facial recognition. Thus, people must be well informed about how devices collect and process data.
Generally, it isn't easy to recognize an actual person with data, but companies must be proactive about data privacy. Moreover, companies can get a competitive advantage by being open about how they process, use, and protect visual data.
alwaysAI believes privacy protections are a key component of any good computer vision application and should be difficult to implement. Privacy laws will continue to evolve, making it difficult for companies to keep up with requirements. But developers don’t have to choose between privacy and the burden of adding those features when using alwaysAI. Our platform provides greater flexibility to continuously update privacy protections than more rigid open source, no-code, or custom computer vision solutions.
Facial blurring for privacy protection
What is Privacy-Preserving AI?
Privacy can be easily incorporated into computer vision applications using specific tools and methods to help protect people’s data.
Blurring is the most common and easiest way to anonymize visual data. You can blur just the face or the entire body, and automatic blurring is a big step in preserving privacy. Another method of object removal using redaction adds a black square over people or objects, but this is not an elegant solution compared to other methods.
Another way you can give more privacy control to the public is to let consumers opt out of data tracking. Before people enter an area that employs something like facial tracking, you can notify the consumer to stand in a certain area to let the CV system know that they don't want their information collected. This means when the system tracks the person, reidentification is disabled, so the data of everything they do in the location is not processed and analyzed.
The alwaysAI platform provides many techniques to preserve data privacy and our edgeIQ library of Python APIs lets you quickly add functions like blurring to any application and opt-out data tracking. Plus, our edge-based platform never stores data, meaning data privacy is guaranteed.
Common Computer Vision Use Cases and Privacy Protection
Computer vision is used in almost every industry, and since it involves situations with people, proactive data privacy is paramount. Let’s see how data is being collected and used in different industries as well as ways to minimize potential risks to the business.
Retail Analytics
Brick-and-mortar companies, QSRs, grocery stores, and many more companies in the retail industry heavily rely on video analytics to provide a better customer experience. They can protect customer privacy by automatically blurring faces or the entire body. These methods do not hinder the required data collection for tracking and improving checkout times, product inventory, people counting, foot traffic analysis, dwell time, and more.
Operational Improvement
Many industries track workers and their movements using computer vision to analyze and improve business operations. Contractors can better manage jobs through progress tracking and defect detection. Computer vision is used in the QSR industry for better process control and customer checkout. You can protect employee data privacy by blurring to anonymize workers while still collecting data on general productivity and how processes are being followed.
Industrial Safety
Computer vision is being used in many industries to address employee safety concerns. For example, if you're a general contractor managing a large construction site, you can use a computer vision application to detect hard hats and safety vests on every person that walks on site. You can easily maintain data privacy compliance and union or employment regulations by adding a simple blurring tool to videos and images.
Transportation
Computer vision is used to provide analytics to improve the transportation industry. For example, in-cabin analytics to count riders for capacity control or monitor rider safety. Also, it is a helpful tool for detecting a medical emergency or safety incident and alerting authorities. Blurring all the passengers' faces is an easy way to maintain their privacy.
blurring and object detection used in public transportation
In Summary
Computer vision is becoming more and more integrated into our daily lives. With that increasing presence comes the question of how to protect individuals’ privacy while still obtaining actionable, real-time data. Fortunately, alwaysAI has privacy-preserving computer vision solutions to help maintain consumer privacy. We make it easy to add privacy features from the outset so that developers don’t see it as a burden and companies don’t look at it as just additional costs to add those features.
The alwaysAI end-to-end computer vision platform can improve nearly every aspect of your business and drive huge ROI. With alwaysAI, you get an easy-to-use, data-driven solution that empowers you to reach your business goals - including protecting privacy.
alwaysAI provides developers and enterprises with a comprehensive platform for building, deploying, and managing computer vision applications on IoT devices. We make computer vision come alive on edge – where work and life happen. The alwaysAI platform offers a catalog of pre-trained models, a low-code model training toolkit, and a powerful set of APIs to help developers at all levels build and customize computer vision apps.