AI video analytics: What are some user concerns?

Needless to say, artificial intelligence has become one of the major topics in security. While AI and advanced analytics do have various benefits, they also create certain issues that cause user concerns, which can be addressed via transparency and clear communications between the stakeholders.
 
That was according to Anna Sliwon, Research Analyst for Security and Building Technologies at IHS Markit.
 
AI has gained more market acceptance due to the user problems that it can help solve. Indeed, the benefits of AI are manifold, Sliwon said.
 
“A device using artificial intelligence will process information through neural networks and other statistical methods which will improve the accuracy of the process’ output,” she said. “Current systems with AI capabilities can help in police investigations, and they can help security managers analyze vulnerable points around their facilities or perimeters. Some other applications use AI analytics to spot shoplifters. Deep learning can be used for perimeter protection to identify trespassers trying to breach the facilities’ perimeter, or to analyze dwell time of vehicles or objects, or for direction tracking, and abandoned object detection. This may help detect people loitering around key infrastructure or government buildings, and help alert the authorities to this seemingly unthreatening behavior.”
 

User concerns

 
However there are certain factors that cause users to be hesitant to deploy these technologies. According to Sliwon these can be summarized as follows.

• Lack of familiarity with how intelligent security systems work.

• Lack of IT expertise among security installers. “Many security installers are not IT experts, and they consider the deployment of intelligent security systems as risky or cumbersome,” Sliwon said. “Installation of intelligent systems requires enhanced understanding of various system integration tools and connections, as well as systems’ operational features in order to be able to provide a full range of technical support and maintenance services, in a timely manner.”

• Cybersecurity concerns. As critical insights about patterns of activity within a building or around its perimeter are much easier to generate, this has given rise to a great concern over security of the processes through which data is sourced and analyzed as well as the locations where such data and insights are stored, and the connections used to transmit them, Sliwon said.

• Privacy. According to Sliwon, many countries have adopted regulations that restrict the use of CCTV systems with requirements for transparency of what images are being collected and for what purposes. “GDPR agreed in the European Union has further strengthened the case for higher standards of data protection across all industries and applications where personal data is processed in any form,” she said.

• The cost. These systems are much more expensive to maintain and install, as each intelligent security camera has to be set up with a specific role in mind, Sliwon said.

 
Below are some ways to address these user concerns, according to Sliwon. 

• Increased education efforts directed at potential customers about various aspects of AI-powered systems – their benefits, risks and the way they work.

• Manufacturers should communicate clearly to installers and end users where relevant, about the weak points in their systems so that users are aware or are able to monitor these more closely.

• Clear communication by the party which owns the intelligent system about the purposes for which the data is being collected and how it is being used. This is in line with GDPR agreed by the European Union, which can provide a useful reference point when designing company’ privacy policies.

 
“The public education efforts would also have to carefully explain how the system would be used, so as to not create an impression of a police state where even a seemingly innocent behavior of a person could be categorized by the intelligent system as suspicious,” Sliwon said. “This also means that for the foreseeable future, AI-powered systems will require regular inputs from the operator to minimize false positives and that there are checks and balances in place.”