Data security, the core lifeblood of security data labeling industry

Time:2020-1-23

With the rapid development of artificial intelligence, we are enjoying the security brought by intelligent security.

In addition to real-time monitoring of various situations, intelligent security can also analyze and predict the content, extract key information (such as license plate, face, action, etc.), and play a role of “prevention in advance”.

Under the background of “landing difficulty” in AI industry, intelligent security is a typical example of AI landing application, which provides substantive reference for AI scene landing application in other industries.

However, unlike other industries, which pay more attention to model and data quality, intelligent security field pays more attention to data security, especially the security of labeled data.

As the basis of artificial intelligence industry, data annotation industry provides a large number of annotation data sets for AI scene landing, which is reflected in the field of security, such as face data, identity information, license plate information, etc. Once these sensitive data are leaked, it will expose people’s daily life and basic information in an all-round way, causing harassment and even fraud risks.

Specific to the data annotation industry, from data collection, data annotation to data storage and transmission, each link must ensure that these sensitive data are not leaked or stolen. However, there are many data security risks in the current data annotation industry.

On the one hand, a considerable proportion of annotation teams do not have a self-developed annotation platform. They still use open-source tools or slightly modified upgraded versions based on open-source tools. Using this open platform has a great risk of data disclosure;

On the other hand, in the process of labeling and storing data, many teams still use public servers due to the cost and technical factors, so they can’t do private cloud deployment, which also has many security risks.

Data security issues have greatly affected the process of AI’s commercial implementation in the field of security.

In view of the above industry problems and based on the actual needs of the enterprise, MF technology has launched its own data annotation solution in the field of AI security. It maintains the security of annotation data from six dimensions of “deployment, network, storage, authority, system, risk control” in an all-round way, and builds a solid shield for the security data annotation industry:

I. deployment

Internal and external network isolation and network scanning rejection;

Container process isolation to improve defensive;

Support private cloud deployment and physical isolation.

Two, network

HTTPS asymmetric encryption, no man in the middle attack.

Three, storage

Multi enterprise data separation, reducing the coupling;

Data asymmetry encryption and storage to prevent data leakage;

Resource access restriction (IP, time control, etc.) to solve malicious access;

Cold and hot data backup for disaster recovery in different places to improve robustness.

Four. Authority

The gateway authority control, refuse to operate beyond the authority;

Separation of enterprise and project authority, mixed refusal authority;

Large batch and abnormal export data authorization to reduce risk.

Five, system

Monitor vulnerability platforms at home and abroad, quickly respond to repair 0day vulnerabilities, and strengthen security protection.

Six. Wind control

Abnormal interception of gateway, anti professional crawler;

Monitor data flow in 7 * 24 hours and inform customers quickly.

Through the above efforts, MF technology is committed to providing the most secure data security services for AI security users and creating the most secure data annotation service experience.

In addition, MF technology’s data security labeling and solutions in the field of security are also fully applied in other industries, such as automatic driving, VR / AR, UAV, new retail, AI education, industrial robots and other fields. MF technology expects to forge the most powerful core lifeline of the data labeling industry with the most stringent data security standards.