The idea of ​​creating the “National video surveillance platform” information system is being discussed in Russia. The system will unite city video surveillance systems throughout the country and become a unified framework for analyzing incoming video data. The model in which cameras will be equipped with computing modules and independently recognize and record incidents, and only after that send the video stream to data centers, will be taken as the basis for the system.

Pros and cons

The technique, where surveillance cameras with built-in “on-board” video analytics modules are responsible for primary video processing before sending it to the server, is called “edge analytics” - product director of Ivideon, a cloud surveillance systems developer, Zaur Abutalimov explains. In general, experts mentioned both advantages and disadvantages of this approach.

If we speak about the advantages, first of all, the use of such a system reduces the load on servers. Secondly, in the long term, this approach can help reduce the operational costs of both the data center and communication channels. Thirdly, the video surveillance and video analytics systems are being accelerated, they become faster. This is achieved due to the fact that the video stream comes to the camera directly from the matrix, it is not compressed or additionally processed.

Zaur Abutalimov

Product director of Ivideon

Other market participants have given similar arguments.

If we consider video cameras networks at the regional level, the introduction of AI cameras will have a cumulative effect on the speed of response processing and will help to save money on the volume of storage on the central node.

Igor Afonin

Director of Multimedia Center and Situation Centers, “T1 Integration”

Despite the higher cost of such cameras, the cost of equipment in the data center will decrease, and in the long term and in case of scaling, the project can be cost-effective.

The proposed solution will significantly increase the number of cameras and reduce the cost of the entire system.

Nikita Mokhnatkin

Head of the information systems security sector, Innopolis University

Moreover, since the decision to transmit the video signal to the control center will be made by the camera itself, the recognition algorithms must ensure the highest accuracy in detecting illegal actions, Mokhnatkin points out. However, today, according to ISS, it is still too early to talk about it.

Nevertheless, the company representatives add that the quality of processing by built-in video analytics algorithms “on board” the camera can still be higher, and it can be feasible to use the system if narrow data transmission channels are available, or during short-term projects, for example, at public events (festivals, concerts, etc. etc.).

Market participants mentioned information security risks, the inability to conduct retrospective video analysis, the use of video from one camera to detect different types of incidents, and the necessity to use the products of certain vendors among the main disadvantages of using cameras with computing modules.

Another challenge in using such devices is the need to modernize and update the databases in each of them. For example, in a conventional video surveillance system, it is enough to replace one camera with another with similar or superior characteristics and, if necessary, separately improve the server capacity - the data will be processed in the data center regardless of which source sent it. But this is not the case with a smart camera.

Pavel Lobanovsky

Expert of the engineering systems unit of the IT outsourcing department, KORUS Consulting Group

The proposed model has both positive and negative features, so it can be assumed that hybrid solutions will be preferable.

Speaking about the national video surveillance platform, it is necessary to take into account the fact that a huge number of video surveillance networks using cameras with different functionalities have already been deployed in the regions. What approach can prevail in this case is not very clear yet. I believe that a hybrid option will most likely be chosen - in this case new networks will use AI cameras when the need arises, and old networks will be integrated via cheaper communication channels with the central core of the platform.

Igor Afonin

Director of Multimedia Center and Situation Centers, “T1 Integration”

Alexander Sivolobov from Skoltech believes that there is no alternative to smart IP cameras. But in this context the question is what functionality will be entrusted to the cameras, and what functions will be performed by centralized data centers.

It is most likely that the main question will be what kind of architecture the solution will have. Probably, it will be a divided unit, where there will be a separate camera from a well-known manufacturer, and next to it there will be a small server with a system for recognizing faces, cars, etc.

Alexander Sivolobov

Deputy Head of the Skoltech-based NTI Center of Excellence in Wireless Technologies and the Internet of Things

Security issues

Information security risks - possible data leaks and unauthorized access by intruders - are the most widely discussed issue of the potential use of cameras with computing modules.

It is true that additional information security risks are emerging. Smart cameras will be of much greater interest to cybercriminals: the cameras will probably contain confidential data, ranging from the accounts using which they will be connected to the data center to provide remote control or to download updated databases, which will be used for search, to the data itself and the software.

Sergey Nenakhov

Head of information security audit department, Infosecurity (a Softline Company)

Special attention is paid to the potential risks of compromising biometric data.

The use of biometric data is becoming more popular, however, in my opinion, protective measures for the collection, storage and processing of such data are not enough. Today, the only good thing that protects us from large-scale incidents is the fact that biometric data is used just as an additional security factor.

Sergey Zolotukhin

Information security specialist, Group-IB

Sergey Nenakhov from Softline believes that due to the relative physical accessibility of individual devices it will be necessary to pay special attention to protecting channels of communication with the data center, and protecting data on the device, for example, using encryption. Andrey Arsentiev from InfoWatch Group says that cybercriminals will be interested in the data in central storages. According to the expert, it will be necessary “to implement measures to protect data from copying and transmission by internal violators”.

Nikita Mokhnatkin from Innopolis University has a different point of view. He believes that the proposed model is more secure.

The decentralized architecture will ensure information security of personal data. It is easier to protect one server than hundreds of cameras. At the same time, if the intruders manage to hack a classic data center, they will get access to all information at once, and in the case of separate surveillance cameras only a small part of information will be available to violators.

Nikita Mokhnatkin

Head of the information systems security sector, Innopolis University

Experience of the first ones

Global surveillance systems are being implemented all over the world, they are widely used in Asian countries. But the use of cameras with built-in computing modules will be an innovation.

I am not aware of any significant experience in using such systems. The implementation of advanced technologies for detecting illegal actions is “at the cutting edge” of technologies, these are the latest scientific achievements, and it will not be possible to “adopt someone’s experience”. These are innovations, and in this case Russia is in the first echelon.

Nikita Mokhnatkin

Head of the information systems security sector, Innopolis University

Both ISS and T1 Integration representatives mention this point.

I have not seen materials on the implementation of large video surveillance systems using AI cameras with a detailed description of the functionality in the open sources.

Igor Afonin

Director of Multimedia Center and Situation Centers, “T1 Integration”

However, Zaur Abutalimov from Ivideon notes that the use of cameras with built-in video analytics modules in local projects is “a worldwide practice for Safe City projects, which has already replaced the outdated continuous streaming method, which is less secure due to possible failures resulting from increased load on the network and servers”. The expert mentions the example of using facial recognition at the check-in counters in US airports, and the example of using technology for identification in Singapore.

In other words, edge analytics technology is currently used only in local projects in some countries and is at the initial stage of development. The implementation of such a video surveillance project at the national level may set a great precedent on the world stage.

Original (in Russian)