Privacy enhancement technology that you cannot be unaware of regarding data collaboration

2023-12-28

On August 12th, Gartner predicted the strategic technological trends for 2021, with privacy enhancement technologies represented by federated learning among them. Is federated learning the best path to address data silos? Can privacy protection be permanent? How to maximize data privacy and practicality? Is there a gap between such technology and enterprise needs, and how will it affect the development of the enterprise?

More than a year ago, in a room where executives from the financial services industry gathered, someone asked, "Who has ever heard of federated learning?" A hand was raised from a venture capitalist in the room. None of the other executives, including the CIO and CDO, have any knowledge of this technology, although their company is likely already using federated learning technology.

Although executives do not need to understand every technology used by the company, federated learning in privacy enhancing technologies (PETs) should be quickly incorporated into strategic discussions about data collaboration when the company seeks secure and secure connections with partners. Privacy enhancement technology originated in academia and was early adopted by government agencies and highly regulated industries. Related technologies can accelerate secure data collaboration, establish customer intelligence (CI), and enhance data value without sacrificing control over data or compromising consumer privacy, and have entered a wider range of commercial application scenarios.

Due to privacy regulations and access restrictions by large technology companies, the supply of data is decreasing. With data privacy and security becoming a global trend, it is easy to see why this technology, which has been around for decades, is now appearing in the vision of some enterprises who want to strengthen their data foundation and achieve true customer centricity. In the process of upgrading data strategies, other frequently mentioned privacy enhancement technologies include differential privacy, synthetic data, Secure Multiparty Computing (SMPC), and homomorphic encryption.

What can privacy enhancement technology solve?

Privacy enhancement technology has existed for decades, mostly only working behind the scenes. Federated learning, which is currently widely known, is a type of privacy enhancement technology that can be used to store sensitive information (such as satellite locations, bank statements, and medical images), enabling licensed parties to securely access information across cloud platforms, infrastructure, and geographic locations without the need to move or copy it anywhere.

So what does federated learning serve? Give a simple example. Assuming that brands A and B with similar user groups have different data. Brand A has user tag data, while Brand B has member data and transaction data. These two companies cannot directly merge their data according to data privacy guidelines, as their respective users do not have the opportunity to authorize this.

So, the current problem is how to establish high-quality crowd models on both ends A and B. However, due to incomplete data (such as enterprise A lacking transaction data and enterprise B lacking label data), or insufficient data (insufficient data volume to establish a good model), it is possible that each end may not be able to establish a model or the effect may not be ideal.

The purpose of federated learning is to solve this problem: it can ensure that each enterprise's own data does not leave the local area, and the federated system can establish a virtual shared model through parameter exchange under encryption mechanisms, without violating data privacy protection regulations. This virtual model is like a better model built by aggregating data together.

Therefore, under such a mechanism, the data itself does not move, nor does it leak user privacy or pose potential risks to data security. The federated system has established a "win win" strategy for the ecosystem and also solved the challenge of data silos.

So, in the current digital age, is federated learning the best path to address data privacy and security?

Privacy protection is not a one-time solution

Gartner has predicted the strategic technology trends for 2021, with privacy enhancement technology among them. The uniqueness of privacy enhancement technology lies in that although all technologies have their advantages, no technology should be "better" or "worse" than another technology, because privacy protection is not a one-time solution. Enterprises hope that configurable controls can be customized to meet their needs and accelerate the establishment of cooperative relationships and the achievement of results. Most privacy enhancement technologies have different uses, and their collaboration may achieve better results depending on the business needs of the enterprise.

Technical service providers should be able to explain what methods they have used to help businesses comply with data privacy regulations and establish consumer trust, as well as the pros and cons of each method. The following are six sets of questions that companies can ask technology service providers to better understand how their technology supports their current and future needs:

Multi party support: Can I implement my own privacy standards and data controls in a multi-party environment? Or do I have to accept the actions of others? Security: Can I keep my data within a certain security range? Or, does collaborating with others require moving data outside of my data infrastructure? Flexibility: What application scenarios are supported and what are not? Speed: Will your technology slow down my analysis, querying, or processing speed? If so, is the deceleration linear (such as 10% deceleration) or exponential (such as 100 times deceleration)? Utility: Is the insight obtained by my team using available data accurate and executable?

Improve data practicality as much as possible without compromising data privacy

As Winterberry recently reported, 70% of surveyed executives in the United States and the United Kingdom are currently or plan to "share data from one party for user insights, activation, and effectiveness measurement or attribution.". Some of these companies may not yet be aware that privacy enhancement technology supports data collaboration to maximize privacy and practicality, expanding the possibilities of one party's data. Using traditional data collaboration models, sensitive information may require the removal of personal identifiers to protect privacy. However, some of this information is necessary for an accurate and unbiased understanding of the audience. Instead of providing complete data tables to authorized specific data scientists and analysts, it is better to keep the original data at a distance from data analysts who use privacy enhancement technology. This approach is faster, more effective, and most importantly, privacy conscious and customer-centric.

However, privacy enhancement technologies represented by federated learning also face problems such as long implementation cycles and high overall budgets. For brands seeking similar effects but hoping to adopt simpler and more feasible solutions, LiveRamp's identifier (ID) conversion solution - Vault - is a better choice to fill the gap between privacy enhancement technologies and enterprise needs.

Fill the gap between privacy enhancement technology and enterprise needs, achieving true customer centricity

Vault is a software service provided by Lianrui China, based on LiveRamp's strong comprehensive strength in the field of privacy protection. By efficiently and securely pseudonymizing and encrypting IDs (Identity, used for user identity in marketing and data analysis environments), it solves the challenge of cross platform interaction of marketing data under the premise of privacy compliance. Vault can handle any ID, and the ID encrypted by Vault is called RampID, which is irreversible and can be deployed in any customer specified environment. The mature key management system of Vault ensures the consistency of RampID generation and can improve the efficiency of RampID generation through clustered deployment.

Through Vault, brands can achieve multiple application scenarios of technologies such as federated learning at lower costs and with greater simplicity, enabling secure data connections and effective data collaboration:

Scenario 1: Security Third Party Data Label Supplement

The brand hopes to supplement its third-party data (such as online and offline transaction data, member phone numbers, and potential user device numbers) with tags, improve user profiles, and enhance data value. During this process, Vault generates a RampID from the encrypted brand data, which is then uploaded to a secure environment specifically designed for this brand. The RampID is then sent to a third-party data service provider. The service provider tags the matching RampID in their LiveRamp safe and ultimately returns it to an environment controlled by the brand. During this process, the matching between the two parties is based on the pseudonymized and encrypted RampID, and without a matching RampID, it is impossible to reverse engineer, identify, and use it to ensure the security of each other's data.

Scenario 2: Secure multi-party data fusion for analysis, modeling, etc

Brands need to integrate their own data with multiple sources of data to support their user profiling analysis, modeling, and other applications. During this process, the brand and other parties respectively generate their own RampIDs through Vault and upload them to the security environment set by Lianrui for the brand. Under the premise of obtaining customer authorization, they are uniformly used for data analysis and application processes such as reporting and modeling, achieving the goal of cross domain security fusion analysis of data.

Scenario 3: Safely uploading data from one party to other external platforms such as CDP

For the demand to upload data from one party to external platforms such as CDP, safe and secure data uploading is crucial. Through Vault, whether it is brand owned data or third-party data provided by LiveRamp connected data partners, RampID will be encrypted and sent to CDP. After various processing by CDP, the data is sent to the media publishing platform in the form of RampID and applied to various aspects of digital marketing, including insight, activation, measurement, and so on. By securely uploading data from one party and from two or three parties, brands can achieve secure and effective data connectivity and collaboration, enabling deeper analysis of consumers and comprehensive empowerment of data, thereby helping to improve user experience and increase data value.

In summary, LiveRamp Vault, as a convenient application of privacy enhancement technology, can achieve the same data privacy protection effect as federated learning technology, breaking down data silos and ensuring secure data connections and collaboration. On this basis, brands can strengthen their data strategy, truly achieve customer centricity, expand data application scenarios, enhance data value, and thus win the key advantage of "users" in long-term competition.

Many companies have been claiming to be customer-centric for many years, and in today's world where digital and privacy first customer-centric standards have been raised, many companies must catch up quickly to avoid falling behind. Privacy enhancement technologies are constantly proving their ability to narrow gaps, accelerate data strategies, build a true single customer view, and consistently provide the next generation experience, establishing a competitive advantage for businesses. To achieve true customer centricity, privacy enhancement technology solutions must quickly enter the data strategy and overall development strategy of enterprises.

Workplace sensor technology: helping office workers feel safe