[Astamuse Co., Ltd.] The cutting edge of privacy enhancement technology: The future seen through research and development and patents
*Astamuse Co., Ltd.*
Press release: September 19, 2024
**
Cutting-edge privacy enhancement technology: The future seen through research, development and patents
Introduction: What is privacy-enhancing technology?
With the advancement of digitalization, the amount of data generated in every aspect of daily life and business is increasing explosively. Companies and organizations aim to utilize large amounts of data to make more accurate decisions and improve operational efficiency. In particular, advances in artificial intelligence are making it possible to efficiently analyze vast amounts of data and extract useful insights. Against this background,
Utilization of data strengthens corporate competitiveness and promotes innovation in society as a whole
I have high expectations.
While the utilization of data is progressing, the risk of privacy infringement when handling personal information has become a major concern. Collecting and storing large amounts of data that includes personal information increases the risk of misuse and unauthorized access, and may damage the trust of consumers and users. For this reason, data users must strike a balance between ensuring safe management and protection of data and how to utilize it effectively.
In particular, when it comes to personal information that is regulated by the Personal Information Protection Act, in addition to complying with legal regulations and guidelines, technical measures such as data anonymization and encryption are also important issues. By performing these properly,
There is a need to balance data utilization and privacy protection.
.
Elements of cutting-edge privacy-enhancing technology
The data utilization process can be divided into the following stages: collection, storage, processing, learning, and use, but in recent years, the processing, learning, and use stages have been receiving particular attention. In response to this trend, three technologies, secure computation, differential privacy, and federated learning, are gaining importance as technologies that enable effective use of data while strengthening data security and privacy protection.
These technologies aim to balance data privacy protection and utilization, and are attracting particular attention because they are able to reduce privacy risks and obtain highly accurate results when processing, analyzing, and sharing data. I am. We briefly explain secure computation, differential privacy, and federated learning. – *Secret calculation*
: A technology that performs calculations and processing while encrypting data. Normally, data must be decrypted for computation, but in secure computation, data can be processed directly without being decrypted.
– *Differential Privacy*
: A technology that provides aggregation and analysis results while protecting personal data. Specifically, adding noise to the data reduces the risk of individual data points being identified, while maintaining the accuracy of the overall analysis results.
– * Federated Learning *
: A technology that allows multiple devices or organizations to jointly train machine learning models using distributed data without sharing the data. Each device or organization processes its own data locally and sends only the resulting model updates to a central server. The server aggregates this and updates the overall model, but the individual data itself is not shared.
These cutting-edge privacy-enhancing technologies are generally expected to have use cases such as the sharing of medical data, the use of highly confidential data such as financial transactions, the publication of government statistical data, and the secure use of user data by companies. has been. In this report, we use our database to analyze target technologies from papers, patents, and grants, and present the results of current efforts and latest use cases in secure computation, differential privacy, and federated learning. . Research and development trends in cutting-edge privacy enhancement technology At Astamuse Co., Ltd., we conduct an analysis called “future estimation” that identifies technologies that have been growing in recent years by extracting the annual trends of characteristic keywords contained in literature, and predicts emerging fields. . By tracing the changes in keywords, it is possible to visualize technologies that have already gone out of fashion and technologies that are expected to gain prominence in the future. Through this analysis, it is possible to predict when technologies will be implemented in society and what technologies will develop in the future. It will be possible.
secret calculation
In order to understand the overall trends in secure computing, Figure 1 shows the annual changes in the number of papers, patents, and grant allocations from 2012 to 2023.
Figure 1: Annual changes in the number of papers and patents related to secure computing and the amount of grant allocation
The grant allocation amount is divided evenly over the project period and calculated based on the redistributed value for each year. For example, if the allocation amount is $30,000 and the implementation period is 3 years, $10,000 is recorded each year. However, China is excluded from the calculations because the disclosure status of grants varies greatly from year to year and does not reflect the actual situation. Regarding patents, since 1 year and 6 months pass from application to publication, the data are aggregated over the period from 2012 to 2022.
Figure 1 shows that the number of papers and patents and the amount of grants allocated are gradually increasing. Let’s take a look at some recent trends and consider 2020 as a reference. From 2020 to 2021, the number of papers and patents has changed from an increasing trend to a stagnation trend. On the other hand, it is noteworthy that the amount of grant allocations continues to increase, with the amount allocated in 2023 reaching approximately 1.9 times the amount allocated in 2020. As funding continues to be invested through grants, it can be interpreted that the stagnation in papers and patents is temporary and that secure computing will continue to advance.
Next, we will trace the changes in keywords related to grants, which have shown a clear increasing trend in recent years. Figure 2 shows the annual trends in keywords included in grant titles and summaries from 2012 to 2023. Growth is the ratio of the number of times a keyword appears between 2012 and 2023 to the number of times a keyword appears between 2019 and 2023. The closer the number is to 1, the more recently the keyword appears. This is interpreted as a high frequency of
Figure 2: Yearly trends in characteristic keywords included in the titles and summaries of grants related to secure computing
From Figure 2, keywords related to finance and cryptocurrency such as “proof-of-stake,” “contracts,” “blockchains,” “proof-of-work,” and “finance” as well as “shard (-x)” You can find specific names of startup companies working in the same field, such as “ledger” and “ledger”. From this, it is clear that finance and cryptocurrencies are being explored as applications for secure computation, and considering that startup company names have appeared, it is assumed that secure computation is close to being put into practical use. will be done.
Finally, we will introduce startups that are applying secure computation to finance and cryptocurrencies and have received the highest amount of funding.
– *Entropy*
– https://entropy.xyz
– Country of location: America
– Founding year: 2021
–
Business Overview: We provide crypto custody services that securely store private keys and make crypto assets available on demand. Specifically, we are working on building a decentralized network that utilizes secure computation technology based on cutting-edge secret sharing methods.
differential privacy
As with secure computation, in order to understand the overall trends in differential privacy, Figure 3 shows the annual changes in the number of papers and patents and the amount of grant allocation from 2012 to 2023.
Figure 3: Annual changes in the number of papers and patents related to differential privacy and the amount of grant allocations
Figure 3 shows that the number of patents and the amount of grants allocated show a monotonous increasing trend. On the other hand, the number of papers will decrease once from 2021 to 2022, but will start to increase again from 2022 to 2023. However, given the continued increase in grant allocations, the temporary decline in the number of papers can be interpreted as a coincidence. In fact, if we look at the amount of grants allocated in recent years, the amount allocated in 2023 is approximately 2.5 times that of 2020. Similar to secure computation, we can assume that research on differential privacy will continue to advance in the future.
Next, we will trace the changes in keywords regarding grants, which show a monotonous increasing trend. Figure 4 shows the annual trends in keywords included in grant titles and summaries from 2012 to 2023. Figure 4: Annual trends in characteristic keywords included in grant titles and summaries related to differential privacy
From Figure 4, we can see that “geo-graph-indistinguishability”, “pii”, “geo-indistinguishability”, “membership-inference”,
“anonymize”, “crowdsensing”, “crowdsourced(data
Keywords directly related to personal information protection and privacy protection, such as “annotation”, can be found. In addition, the names of specific startup companies working in the same field, such as “dpella” and “gdpr-compliant,” as well as regulations that contribute to the protection of personal information (GDPR is an initiative to strengthen and unify the protection of personal data within the European Union). You can also see keywords related to the rules intended for This suggests that differential privacy is in high demand in the field of personal information protection and privacy protection.
We will introduce startup companies with the highest amount of funding that apply differential privacy to personal information protection and privacy protection, similar to secure computation.
– *Sarus*
– https://sarus.tech
– Country of location: France
– Founding year: 2019
–
Business overview: We utilize differential privacy to solve problems associated with accessing and sharing personal data in data analysis and machine learning. In particular, it simplifies and automates the process of anonymizing personal data in data analysis projects, facilitating the start of projects.
federated learning
As with secure computation and differential privacy, in order to understand trends in federated learning, Figure 5 shows the annual changes in the number of papers and patents and the amount of grant allocation from 2012 to 2023.
Figure 5: Annual changes in the number of papers and patents related to federated learning and the amount of grant allocation
Figure 5 shows that federated learning is a very new technology, with very few papers or patents published from 2012 to 2018, and the amount of grants allocated to it was extremely small. However, since 2019, the number of papers and patents and the amount of grant allocations have shown significant growth. In particular, the amount of grants allocated will reach approximately nine times in 2023 compared to 2020, indicating that the demand for federated learning is extremely increasing. This suggests that active research will continue for some time to come.
Next, we will trace the evolution of keywords regarding grants, similar to secure computation and differential privacy. Figure 6 shows the annual trends in keywords included in grant titles and summaries from 2012 to 2023.
Figure 6: Yearly trends in characteristic keywords included in the titles and summaries of grants related to federated learning From Figure 6, we can see general medical-related keywords such as “health-technology,” “omop-cdm,” and “rarecarenet,” as well as keywords related to medical databases. You can also see many keywords for services and programming frameworks related to federated learning, such as “fedml”, “openfl”, and “fedn”. From this, it is clear that federated learning is already at the stage of practical application and is being widely applied mainly in the field of analyzing medical data that is siled (divided and stored).
Similarly, we’ll introduce you to some of the top-funded startups that apply federated learning to processing siled data.
– *Owkin*
– https://www.owkin.com
– Country of location: USA
– Founding year: 2016
–
Business overview: We aim to establish treatments for cancer and other diseases and discover new drug candidates by effectively utilizing data sets and artificial intelligence owned by medical institutions and pharmaceutical manufacturers. Along the way, they focus on protecting siled data sets, such as patient personal information, and leverage federated learning to accomplish this.
Future vision of cutting-edge privacy enhancement technology We analyzed the latest efforts in secure computation, differential privacy, and federated learning, which are cutting-edge privacy enhancement technologies, from papers, patents, and grants using Astamuse’s database, and introduced the results. .
In order to comprehensively understand cutting-edge privacy
enhancement technologies, we looked at the grant allocation amount (Figure 7), number of papers (Figure 8), and patents for each of the three technologies, based on 2020 (1). Figure 9 shows the annual change in the number of cases (Figure 9).
Figure 7: Annual transition in grant allocations related to three cutting-edge privacy-enhancing technologies
Figure 8: Annual transition in the number of papers related to the three technologies
Figure 9: Annual transition in the number of patents related to the three technologies
From Figures 7, 8, and 9, when comparing the number of papers and patents and the amount of grant allocation for secure computation, differential privacy, and federated learning, we find that while secure computation and differential privacy show a gradual increasing trend, It can be seen that learning has shown a remarkable increase rate since 2018. In particular, based on 2020, the number of papers published by federated learning in 2023 will be approximately 3.1 times greater, the number of patents will be approximately 2.3 times greater, and the amount of grants allocated will be approximately 9 times greater.
Also, when comparing secure computation and differential privacy, the increase rate appears to be slightly larger for differential privacy. This can be explained by the length of the history of each technology, with secure computing starting in the 1970s, differential privacy in the 2000s, and federated learning in the 2010s. , it is natural that the older the technology, the smaller the growth rate in recent years. However, both technologies are showing continuous growth and remain fields of active research and development.
The optimal use cases for each technology are still being actively explored. From Figures 2, 4, and 6, the use cases of secure
computation, differential privacy, and federated learning can be organized as follows.
– *Secret calculation*: Finance and crypto assets
– *Differential privacy*: Personal information protection and privacy protection
– *Federated learning*: Analysis of siled (divided storage) medical data The applications of each technology are, of course, not limited to the above, but also for confidential information held by companies that are not subject to personal information protection laws.
The use of cutting-edge privacy-enhancing technology to solve common issues across organizations is being discussed.
. In this way, we continue to expect the provision of new value and the creation of new services by applying cutting-edge privacy enhancement technology.
Author: Astamuse Co., Ltd. Mishenko Peter Ph.D. (Engineering) Further analysis…
At Astamuse, we conduct daily analyzes not only on technology related to “privacy enhancement” but also on various cutting-edge
technologies/advanced areas, and provide this to a variety of companies and investors.
In this report, we have published some of the analysis results. Data sources used for analysis include R&D grant data from each country to grasp cutting-edge research trends from the latest government trends, startup/venture data to grasp the latest business models, and the latest trends. There is patent/paper data etc. to support it.
Based on the results of these analyses, we conduct in-depth analysis that combines a bird’s-eye view and multiple perspectives from various time axes and player perspectives, thereby achieving the precision necessary to build R&D strategies, M&A strategies, and business strategies. It is possible to backcast and understand high medium- to long-term future predictions and the opportunities and threats they bring to your company.
In addition, in each area/theme, we not only analyze technology units and issues/value units, but also analyze players at the company level, and identify key persons as innovators as outputs that are more specific and easy to utilize in the field.
It is also possible to analyze and search for opinion leaders (KOLs) globally. If you are interested, please contact us.
– Corporate site: https://www.astamuse.co.jp/
– Inquiry form: https://www.astamuse.co.jp/contact/
– Inquiry email: inquiry@astamuse.co.jp
**