Data Obfuscation (Supply Chain)

Obfuscation is a method to remove all sensitive data, typically private data, from a dataset while preserving the statistical patterns of interest for a given task, for example supply chain optimization.

Time-series data are sometimes highly sensitive. You may even think that your data are too sensitive to be entrusted to an online third party such as Lokad. We agree that data sensitivity is an issue but we believe that it is not an insurmountable one.

Obfuscation can be used to protect your most sensitive time-series data. Intuitively, obfuscation is extremely simple: without proper contextual information, raw time-series data are totally unreadable. Without a time-series description (What is being measured by this time-series?) and a time-series unit (How are the time-series measurements defined?), a time-series is nothing but a meaningless list of numbers. By obfuscating your data, you make sure that neither the descriptions nor the units can be guessed for the time-series data stored in your Lokad account.

Obfuscating your time-series data is easy: choose arbitrary time-series names such as T1, T2, T3, ...

Also, on top of raw time-series data, Lokad also supports tags and events. Those meta-data can be easily obfuscated following the very same principle. Human-readable tags should be replaced by identifiers. Since Lokad has a pure statistical approach, such substitution has no impact on the forecasting results.

By sticking to those two guidelines, you ensure that not even your fiercest competitor will be able to exploit the data contained in your Lokad account. The first guideline ensures that no description is provided, while the second ensures that the time-series cannot be reverse-engineered by exploiting key numbers that can be recognized as such.

Obfuscating your time-series data has no impact on the Lokad forecasts because our methods are purely statistical. Time-series names are only provided for user convenience and play no part into the forecast process.