Abstract Many statistical agencies face the challenge of maintaining the confidentiality of respondents while providing as much analytical value as possible from their data. Datasets relating to businesses present particular difficulties because they are likely to contain information about large enterprises that dominate industries and may be more easily identified. Agencies therefore tend to take a cautious approach to releasing business data (e.g., trusted access, remote access and synthetic data). The Australian Bureau of Statistics has developed a remote server, called TableBuilder, which has the capability to allow users to specify and request tables created from business microdata. The tables are confidentialised automatically by perturbing cell values, and the results are returned quickly to the users. The perturbation method is designed to protect against attacks, which are attempts to undo the confidentialisation, such as the well-known differencing attack. This paper considers the risk and utility trade-off when releasing three Australian Bureau of Statistics business collections via its TableBuilder product.

中文翻译：

---

通过远程服务器释放总数时保护业务微数据的前景

摘要许多统计机构都面临着维护受访者机密性的挑战，同时要从其数据中提供尽可能多的分析价值。与企业相关的数据集存在特殊的困难，因为它们可能包含有关主导行业的大型企业的信息，并且可能更容易识别。因此，代理倾向于采取谨慎的方式来发布业务数据（例如，受信任的访问，远程访问和合成数据）。澳大利亚统计局已经开发了一个名为TableBuilder的远程服务器，该服务器具有允许用户指定和请求从业务微数据创建的表的功能。通过扰动单元格值来自动对表进行保密，并将结果快速返回给用户。摄动方法旨在防止攻击，即试图取消机密性的攻击，例如众所周知的差异攻击。本文考虑了通过TableBuilder产品发布三个澳大利亚统计局业务集合时的风险和效用折衷。