Patent analysis on key technologies for the security and trustworthiness of AI

Source：iprdaily.com

Author：IPRdaily

The original title：Patent analysis on key technologies for the security and trustworthiness of AI

Introduction：Considering that AI robustness, AI data privacy protection, AI interpretability and fairness are the key issues affecting the adoption of AI technology, we conducted this global patent data retrieval and analysis from these four dimensions.

Artificial intelligence (AI) is not a new technical term. It plays an essential role in multiple aspects of everyday life, including clothing, food, housing, and transportation.

AI is the driving force for the scientific and technological revolution. It has rapidly stimulated the improvement and transformation of industries in various fields. However, there has been concerns regarding security, ethics, privacy, and other associated risks of this technology. It is therefore important to foresee, identify and effectively manage such risks while considering technical solutions to these problems. 

With the increasing trend towards AI adoption, "relying on AI to deal with the security problems under the digital wave" and "ensuring its trusted application" have become an important issues for consideration by academia, industry and policy-makers.  

AI technology has many diverse applications. The main challenge is to ensure it can operate safely in different environments and conditions. 

Enhancing the robustness of these systems will help to improve the stability of AI technology, which will be key to unlocking the benefits of the digital economy. 

Data is an important pillar of the development of AI models. Technical solutions will have to be used to overcome concerns around extensive data collection, misuse, and embezzlement. This has implications not just for industry research, but for user rights. There are considerable differences in regional legislation and regulation which govern data ownership. Using privacy or security-enhancing technologies has become a driver for the use and commercialization of data. 

AI decision-making is based on trust with the user. How should these systems give visibility and transparency on decision-making processes and include these as a design feature for the model? How do we ensure algorithmic fairness and prevent discrimination in these decisions? Building trust with the public is therefore a key consideration. 

Considering that AI robustness, AI data privacy protection, AI interpretability and fairness are the key issues affecting the adoption of AI technology, we conducted this global patent data retrieval and analysis from these four dimensions. Is the industry focusing on these technical problems? Who is committed to ensuring the trust and security of AI technology? This paper provides an overview based on publicly available patent data[1].

1.Innovation output of key technologies for the security and trustworthiness of AI

The analysis results in Figure1[2] show that there are 8,038 DWPI Families[3] related to key technologies for the security and trustworthiness of AI, corresponding to 13,776 patent applications. It can be seen that the key technologies for the security and trustworthiness of AI has gained significant attraction and has produced a number of innovative technologies. 

Among the 8038 innovative technologies, more than 70% are related to AI data privacy security and more than 20% are related to AI robustness. On the other hand, there are fewer innovations related to AI interpretability and fairness.

Based on the region of the first application shown in Figure 2, 53% of the innovations in this field originated from China and 25% from the United States. These two countries account for more than half of the world's innovations and focusing on the domain of AI security and trusted technology.

Figure 1 Innovation output of key technologies for AI security and trustworthiness technologies

Figure 2 Priority country analysis

2. Top innovation entities of key technologies for the security and trustworthiness of AI

The Top 21 innovation entities in AI security and trustworthy key technology area are based on the number of DWPI families. Ant Group (Alipay) leads in the list with 409 DWPI families, followed by Tencent with 248 DWP families, and IBM with 236 DWP families.

Figure 3 Top innovation entities-based on number of DWPI families

3. The number of granted patents of the top innovation entities in this area

The results show that the top 3 players are Ant Group (Alipay), NTT and IBM. These 3 companies all have more than 100 grants. This was based on further analysis of the number of DWPI family grants of the Top innovation entities in AI security and trustworthiness key technology area (if one patent in the DWPI family is granted, this family will be regarded as a granted family) and rank according to the number of grants.

Figure 4 Number of granted patents of the top innovation entities

4. Markets of interest of the Top innovation entities in key technologies for AI security and trustworthiness

Generally, the countries / regions where patents are filed are determined according to the market strategy of an entity, and they will generally choose the more important and confident patented technologies to file in multiple target markets. Therefore, by analyzing the filing countries / regions of the enterprise's patent layout, we can understand the enterprise's target market and technical confidence. By analyzing the overall target market of the patent layout in the field of AI security and trusted technology, we find that, except for Xi'an University of Electronic Science and Technology only focusing on China, Tencent filed in three markets (China, the United States and China Hong Kong) Ping An technology has filed in 5 markets (China, the United States, Europe, Singapore, China Hongkong), other Top innovation entities have filed in at least 8 or more markets.  Ant Group(Alipay) has filed in 17 markets which leads the broad filing in China entities.

Figure 5 Markets of interest of the Top innovation entities

Through this study, we identified that AI security and trustworthiness are key technologies which have attracted significant attention and already produced several important innovations. Among them, China's Ant Group (Alipay) and Tencent have played a leading role, followed by IBM, NTT, MICROSOFT and other electronic giants. It is believed that as AI becomes more widely applied, security and trustworthiness will attract more research attention.

[1]Patent system is designed to encourage invention and creation and facilitate technology dissemination. Moreover, because of the first-to-file principle of patent system, any innovative technology solution will normally file a patent first unless it is a tradetechnical secret. Therefore, patents are the most comprehensive, timely and objective information carrier, which can be used to understand objective technology, marketmarket, and competitive intelligence.
[2]The search deadline of this research is February 23, 2021, covering global patent applications and / or granted patents published from the earliest priority date January 1, 2000 onwards.
[3]DWPI is the most reliable value-added patent data worldwide. Patent offices in more than 40 jurisdictions are using DWPI for official patent retrieval and examination. In the process of data processing, the Derwent world patent index (DWPI) will merge multiple patents of the same technical subject into a Derwent patent family(DWPI family), and the multiple patents of this family relate to the same technical subject. In this paper, the technology analysis is based on the Derwent patent family. In this way, we can avoid that a technical topic is calculated multiple times because there are many patent family members. When it comes to the filing country analysis, we use a separate patent as the statistical unit.

 

Source：iprdaily.com

Author：IPRdaily

Editor：IPRdaily-Vapor