Request demo
Contact Us

FAQ . .. ...

bluegen.live Synthetic Data FAQ

Category: General
Category: Data Quality
Category: Technical

General.

Synthetic data is artificially generated data that looks and behaves exactly like real data.

Read more

The most common use cases for synthetic data are to use it as a privacy-safe alternative to real data. For example when training machine learning models, performing exploratory analysis and collaborative research, or in your software development and testing process.

Besides being privacy-safe, you have full control over the feature characteristics of the synthetic data you generate. This opens new possibilities for conditioning, augmentation, balancing and scaling which real data does not allow.

Read more

An intrinsic property of synthetic data is that personally identifiable information isn’t inherited from the real data set. So you can not find and use specific individual records that exist in the real world.

Read more

Synthetic data overcomes your data privacy, security, and utility barriers and enables faster, cheaper, and broader access to the value hidden inside your data. The question we ask ourselves at bluegen.live is why you are still being at risk and wasting resources by using real data.

Read more

We use various deep learning and data processing techniques in an autonomous system that generates high-quality synthetic data. The generative core of our software runs on state-of-the-art diffusion models and is also ready to be combined with differential privacy and federated learning, to offer more privacy and secure deployments for the most demanding use cases.

Deep learning uses artificial neural networks to discover complex patterns in large amounts of data. Like ChatGPT which is trained on massive amounts of unstructured text so it is able to write new text by predicting word combinations it has discovered..

We train our neural networks specifically to learn from structured data sets and create high quality synthetic data sets.

  1. Download and install the software in your environment
  2. Import your data set into the software
  3. Configure the model based on your use case requirements
  4. Train the model and review its output quality
  5. Generate your own synthetic data on demand
Load more

Data Quality.

We use both metric- and attack based evaluations to calculate the remaining privacy leakage risks in the synthetic data for singling out, linkability and inference. For the most demanding use cases and tough attacker profiles we offer the addition of differential privacy during the training process which results in a mathematically proven privacy guarantee at the cost of some resemblance and utility of the synthetic data.

By positioning a generative model between the real and synthetic data it inherently breaks the 1-on-1 relationship between them. This is fundamentally different from (pseudo)anonymization techniques which only alter the real data in various ways and have either lower utility or higher privacy risks. This is also why for example pseudonymized data is still considered personal data by the GDPR, while synthetic data is not.

In addition to our extensive default evaluation on resemblance, utility and privacy, we advise you to compare the outcomes of the real and synthetic data for various relevant domain specific business questions. This is the proof of the pudding and helps with communicating the quality to your business peers.

When our software is trained to generate your synthetic data, it automatically creates a comprehensible PDF evaluation report which you can easily share. To support the scores on resemblance, utility and privacy, the report comes with all kinds of graphs and performance indicators such as distributions, percentiles, distances, correlation, precision, and recall.

Technical.

The bluegen.live software resides in your own secure environment of choice, either in the cloud or on-premise so the real data stays safe where it is.

As an add-on, the bluegen.live software offers various API’s to integrate with your IT environment.

Once the system is trained, bluegen.live can generate thousands of synthetic data rows per second.

Depending on the available computing power and amount of real data, the training process can take from an hour up to a day in the most complex cases.

We haven’t run into our limits yet regarding the amount of real data. But take into account that training time will increase with larger and more complex data sets.

bluegen.live needs at least a thousand rows of data to train itself properly. And the more columns there are in a dataset, the more rows are required to learn all its statistics, correlations, and relationships.

Read more

When our self-learning system has trained itself to create synthetic data, it only needs to run periodically to keep up with the changing characteristics of your real data. To save time, you can do this incrementally and over different time intervals, depending on your use case.

Read more

Yes, if the data in your data lake is structured so that bluegen.live can do its work efficiently.
bluegen.live supports all kinds of single table and time series data, such as in database systems like SQL, Oracle, and MongoDB.
Category

General.

Synthetic data is artificially generated data that looks and behaves exactly like real data.

Read more

The most common use cases for synthetic data are to use it as a privacy-safe alternative to real data. For example when training machine learning models, performing exploratory analysis and collaborative research, or in your software development and testing process.

Besides being privacy-safe, you have full control over the feature characteristics of the synthetic data you generate. This opens new possibilities for conditioning, augmentation, balancing and scaling which real data does not allow.

Read more

An intrinsic property of synthetic data is that personally identifiable information isn’t inherited from the real data set. So you can not find and use specific individual records that exist in the real world.

Read more

Synthetic data overcomes your data privacy, security, and utility barriers and enables faster, cheaper, and broader access to the value hidden inside your data. The question we ask ourselves at bluegen.live is why you are still being at risk and wasting resources by using real data.

Read more

We use various deep learning and data processing techniques in an autonomous system that generates high-quality synthetic data. The generative core of our software runs on state-of-the-art diffusion models and is also ready to be combined with differential privacy and federated learning, to offer more privacy and secure deployments for the most demanding use cases.

Deep learning uses artificial neural networks to discover complex patterns in large amounts of data. Like ChatGPT which is trained on massive amounts of unstructured text so it is able to write new text by predicting word combinations it has discovered..

We train our neural networks specifically to learn from structured data sets and create high quality synthetic data sets.

  1. Download and install the software in your environment
  2. Import your data set into the software
  3. Configure the model based on your use case requirements
  4. Train the model and review its output quality
  5. Generate your own synthetic data on demand
Load more
Category

Data Quality.

We use both metric- and attack based evaluations to calculate the remaining privacy leakage risks in the synthetic data for singling out, linkability and inference. For the most demanding use cases and tough attacker profiles we offer the addition of differential privacy during the training process which results in a mathematically proven privacy guarantee at the cost of some resemblance and utility of the synthetic data.

By positioning a generative model between the real and synthetic data it inherently breaks the 1-on-1 relationship between them. This is fundamentally different from (pseudo)anonymization techniques which only alter the real data in various ways and have either lower utility or higher privacy risks. This is also why for example pseudonymized data is still considered personal data by the GDPR, while synthetic data is not.

In addition to our extensive default evaluation on resemblance, utility and privacy, we advise you to compare the outcomes of the real and synthetic data for various relevant domain specific business questions. This is the proof of the pudding and helps with communicating the quality to your business peers.

When our software is trained to generate your synthetic data, it automatically creates a comprehensible PDF evaluation report which you can easily share. To support the scores on resemblance, utility and privacy, the report comes with all kinds of graphs and performance indicators such as distributions, percentiles, distances, correlation, precision, and recall.

Category

Technical.

The bluegen.live software resides in your own secure environment of choice, either in the cloud or on-premise so the real data stays safe where it is.

As an add-on, the bluegen.live software offers various API’s to integrate with your IT environment.

Once the system is trained, bluegen.live can generate thousands of synthetic data rows per second.

Depending on the available computing power and amount of real data, the training process can take from an hour up to a day in the most complex cases.

We haven’t run into our limits yet regarding the amount of real data. But take into account that training time will increase with larger and more complex data sets.

bluegen.live needs at least a thousand rows of data to train itself properly. And the more columns there are in a dataset, the more rows are required to learn all its statistics, correlations, and relationships.

Read more

When our self-learning system has trained itself to create synthetic data, it only needs to run periodically to keep up with the changing characteristics of your real data. To save time, you can do this incrementally and over different time intervals, depending on your use case.

Read more

Yes, if the data in your data lake is structured so that bluegen.live can do its work efficiently.
bluegen.live supports all kinds of single table and time series data, such as in database systems like SQL, Oracle, and MongoDB.
Category

General.

bluegen.live needs at least a thousand rows of data to train itself properly. And the more columns there are in a dataset, the more rows are required to learn all its statistics, correlations, and relationships.

bluegen.live needs at least a thousand rows of data to train itself properly. And the more columns there are in a dataset, the more rows are required to learn all its statistics, correlations, and relationships.

Category

Data Quality.

bluegen.live needs at least a thousand rows of data to train itself properly. And the more columns there are in a dataset, the more rows are required to learn all its statistics, correlations, and relationships.

bluegen.live needs at least a thousand rows of data to train itself properly. And the more columns there are in a dataset, the more rows are required to learn all its statistics, correlations, and relationships.

Category

Technical.

bluegen.live needs at least a thousand rows of data to train itself properly. And the more columns there are in a dataset, the more rows are required to learn all its statistics, correlations, and relationships.

bluegen.live needs at least a thousand rows of data to train itself properly. And the more columns there are in a dataset, the more rows are required to learn all its statistics, correlations, and relationships.

Where do you want to unlock the innovation power of your data?

High quality test data for software testing and development

Training and testing data for accurate ML models

Privacy-compliant data for sharing and collaboration

See all use cases