[New Wave] Challenges of Big Data Utilization: 'Personal Information Infringement'

Many people have likely experienced seeing certain products they searched for with interest on mobile-based online shopping malls or commerce applications appear in the recommended lists on personal social networking service (SNS) screens unrelated to the shopping platform. Such online personalized advertising or recommender systems are representative examples of advertising and marketing strategies utilizing big data.

One application of big data is in the construction and utilization of computer algorithms. With advancements in artificial intelligence (AI), machine learning, and data construction systems, computer algorithms are replacing or assisting human decision-making processes across various fields such as politics, education, and healthcare. An algorithm refers to a logical process composed of formulas or commands designed to solve specific problems. In advertising and marketing, personalized advertising algorithms and product recommendation systems are widely used. These systems analyze and predict consumers’ accumulated consumption patterns to suggest appropriate products or present advertisements for products of interest, systematically processing a series of steps including data collection, analysis, and result presentation.

Big data provides vast amounts of information that enable the construction and use of such algorithms, but since the data is not formed for any specific purpose, it is unstructured and may contain a lot of unnecessary information. Even with the same data, the results can vary greatly depending on how each process is handled, and sometimes personal information may be unintentionally leaked or reconstructed and misused.

Handling information about individuals, that is, consumers, inevitably raises debates about the possibility of personal information infringement. Algorithms that identify the characteristics of individual consumers to recommend products or present advertisements can be helpful systems from a functional perspective. However, consumers often feel suspicious about how a site knows their interests, what personal information it holds, and how it commercially uses that information, leading to concerns about potential privacy violations.

The government is creating policies that allow consumers greater control over their information through revisions of related systems such as the Personal Information Protection Act to reduce damages from personal information leaks. On the other hand, the industry often cites privacy protection issues and the strictness of related legislation as factors hindering the development of the big data industry. Thus, there is a strong conflict between calls for deregulation of big data and opinions that prioritize privacy protection and demand stronger regulations.

Recently, federated learning technology, which has begun to attract attention, enables AI technology without the process of gathering data into a central server. For example, Google views federated learning as an alternative for big data utilization free from privacy infringement controversies and introduced a service called Gboard using this technology. Gboard is a type of keyboard software that provides various functions such as Google search, translation, and voice typing within the keyboard. When a user inputs a word, it suggests related words learned through training. Even though Gboard does not receive the user’s usage records, it can learn new words based on models updated by many people training on their own devices.

Big data-based technology has an inherent limitation in that it cannot be completely free from privacy infringement issues due to its vast information and implicit value. Nevertheless, multifaceted efforts are necessary to minimize the possibility of privacy violations and prevent incidents.

Wonjun Jeong (Associate Professor, Department of Media Communication, University of Suwon)