Home » Blog » Qi Hui Intelligent: Applied CnosDB to Create a New Generation of AI Security Platform

Qi Hui Intelligent: Applied CnosDB to Create a New Generation of AI Security Platform

School is the “ivory tower” that people think of and what everyone thinks is the safest and most secure place. However, in recent years, there have been frequent campus safety incidents in China, and injuries such as fights, indecency, and death have frequently broken out, seriously affecting the physical and mental health and life safety of students. At the same time, problems such as insufficient campus security facilities and poor management have once again attracted attention.

In fact, at present, China’s campus security is still in its infancy, many types of schools, such as junior high schools, technical schools, universities and other security facilities are not perfect, and even some third- and fourth-tier urban schools have no security system at all. The installation of security systems on campus is a mandatory policy of the Ministry of Education. Therefore, in the face of large-scale market demand, more and more investors are paying attention to this market, especially the use of artificial intelligence technology to promote the construction of safe campuses, which is a key field promoted by the state.

Founded in 2016, Qi Hui Intelligent is an AI security solution provider, the company takes AI voice security as the entry point, successfully creates AIoT equipment for campus security, and at the same time achieves digital empowerment for campus security through a comprehensive set of digital operation system solutions that complement it to support its efficient operation management. The company’s team is composed of people with many years of experience in the industry. At the same time, the team is also the first batch of police voiceprint collection terminals and the first batch of AI voice alarm solutions on sale.

Qi Hui intelligent digital operation system solution

Business and Application Requirements

At present, the traditional campus security participants in the market are still dominated by AI video surveillance. However, according to relevant industry data, more than 50% of school violence occurs in private places such as dormitories and bathrooms. Because private places such as restrooms and dormitories cannot be equipped with surveillance cameras, it is impossible to monitor what happens inside in real time or collect evidence afterwards, so campus violence is difficult to prevent and treat.

At the same time, in recent years, with the increasing AI capabilities of IoT devices, such as voice recognition and image recognition, we are moving from the IoT era to the AIoT era, and we are also facing more serious security challenges. Because AIoT devices are generally networked, networking has the possibility of being attacked by the network and stealing information by the network. Secondly, AIoT devices generally have sensors, such as cameras and microphones, which can obtain sensitive information around them, and if these sensitive information are obtained through the network or SD cards, emmcs, etc., the security risk is high. Finally, the stability challenge of AIoT equipment is that if an exception occurs during system operation, it will cause great losses.

In the face of the difficulty of surveillance cameras to reach the monitoring blind spot and the increasingly demanding requirements of the industry for data security, intelligent voice alarms using AI voice recognition technology and high data security are the solutions in the industry. Qihui Intelligent’s AIOT product “Yinzhi Sentinel”, facing the coexistence of three major challenges of outdoor & strong reverberation, long-distance sound pickup and ultra-short wake-up words, handed over a product answer sheet that combines its own algorithms, independent and controllable hardware supply chain and network architecture, and high data security with extremely difficult market-oriented application.

Problems Under Traditional IT Architectures

In order to make the system massive data calculation more accurate and the large-scale application of AI in the industry in the future, most enterprises want to save longer-term historical data. However, in current industry applications, subject to the storage cost and memory cost of ordinary disk arrays, users often cannot save data for more than 5 years. At the same time, the database compression ratio is not high, and the storage cost is high. And its cost will continue to rise over time. It is difficult for users to save long-term data, and a large number of effective data of the device is overwritten and lost.

Taking Qihui Intelligence as an example, only the “Yinzhi Sentinel” and “Yinzhi Yingwei” series products are in urgent need of improving database capabilities. Specifically, the “Yinzhi Sentinel” product and system platform does not require too much storage space at the customer application level, but due to its AIOT architecture and the requirements of security alarm scenarios, the front-end equipment must have strong functional stability, which is the biggest challenge faced by Qi Hui, and even more AI security companies in the industry. Therefore, the background of Qi Hui Intelligent has a set of fine real-time monitoring data feedback programs for the stability of equipment hardware. For example, every five minutes the device reports heartbeat data, CPU temperature, disk occupancy, memory occupancy, etc. With the rapid development of Qihui’s intelligent business and the increase in equipment deployment points, that is, the number of digital intelligent operation system devices on the cloud has increased rapidly, Qicihui urgently needs a database with more powerful data throughput capabilities to replace the PostgreSQL or MySQL databases that can only be selected by users of its original IoT platform.

In the view of Qihui Intelligence, taking every 6 regional-level platforms and 5,12 devices per platform as an example, each device needs to access <> smart sensors, collect data every <> seconds on average, and there will be an average of about <>,<> writes per second, that is, <>,<>-level QPS, and with the rapid growth of the number of devices, the writing pressure of future data will continue to increase. Therefore, under the future data scale, traditional relational databases represented by MySQL and Oracle cannot complete the rapid and continuous storage of massive time series data at the database design level due to the limitation of storage structure. Secondly, Qihui Intelligent also needs to store the monitoring data for a long time to optimize the system analysis reference to ensure the stable performance of the system in various scenarios, regions and different magnitude of use. The storage cost of the past database is too high to retain longer-term historical data more economically, which is an unknown loss for system optimization efforts.

In addition, the illegal leakage, alteration or destruction of data is also an absolutely unacceptable situation for Qihui Intelligence, so the security of the database system is also a problem that Qihui Intelligent Platform continues to focus on solving.

Connect to CnosDB to Solve the Problem

For a new generation of AI security manufacturers such as Qihui Intelligent that deploy AI computing power terminals at massive points, with the growth of both markets and scenarios, the amount of data has begun to grow rapidly, and its requirements for data storage are getting higher and higher. In view of the contradiction between the burst of data volume, application requirements, and technical matching, Qiahui intelligently applies the new generation of time series database CnosDB to improve storage throughput and reduce storage costs.

CnosDB is a time series database that is very suitable for IoT scenarios, with strong write throughput and efficient compression ratio. Compared with traditional relational databases, it can support larger measurement points, and can support millions of QPS write requirements on ordinary servers. At the same time, through the new architecture design, CnosDB has a good performance in data compression, which can minimize storage costs and make it possible to retain longer-term data. According to the experience of Qi Hui Intelligence, the storage cost of its platform has been reduced by two-thirds after using CnosDB.

In terms of security, CnosDB has taken several measures to ensure the security and confidentiality of data. First, it supports authentication and access control at high levels of encryption. Secondly, CnosDB provides data encryption function, which can encrypt and store data to prevent unauthorized access. In addition, CnosDB also has data backup and recovery capabilities, which can recover data in case of data loss or corruption. CnosDB has multi-layered protection measures in terms of security, which can effectively help Qishui intelligently protect the security of data.

New Platform Architecture

The Future

Through the AIoT platform of Qihui Intelligence, the combination of technology and security software and hardware is realized to achieve the AI security control scheme of pre-prevention, early warning and post-event tracking, which not only saves time and effort, but also solves the pain point that traditional security can only collect evidence after the fact, and it is difficult to collect evidence.

Qi Hui Intelligent has always been at the forefront of industry exploration, through the application of various cutting-edge new technologies, its AI security platform has entered the stage of digitalization, integration and intelligence. Helping users realize the automation level of their security monitoring, improving users’ intelligent analysis capabilities, and lightweight operation and maintenance has always been the never-ending pursuit of Qi Hui. Through the in-depth cooperation with CnosDB, we believe that in the future, we will give a pure land to the campus and a blue sky for the children.