- Modzy Labs
- November 15, 2019
Data Privacy: Deployment
Protecting data privacy is one of the most important issues in technology today. From data breaches, to questions about ethical data sharing, it is clear that access to data is a hot topic. This is especially true in the field of artificial intelligence (AI). AI lives and breathes with data, so it is crucial that data privacy be thoughtfully considered when implementing new AI-enabled solutions. Ethical data management is very important at Modzy and so our platform is built with our customers’ data privacy considerations in mind.
WHAT YOU NEED TO KNOW
There are two types of data associated with AI- training data and inference data. Training data is used to create or improve algorithms. Often it is labelled or tagged with accurate information so that the algorithm knows if it is behaving correctly. Inference data, on the other hand, is data that is unknown and is used in production.
Data privacy is important for both kinds of data associated with AI. The training data used plays a critical role in the functioning of an AI algorithm. In order to be useful, training datasets need to be large and specific. Unless the model is something generic like the now-famous Hot Dog or Not-Hot-Dog model, it probably contains some form of sensitive information that should be kept private. In addition, protecting training data preserves intellectual property protection for groups creating new models or making improvements to existing models. Likewise, inference data is also likely to contain sensitive information. While using an AI solution can yield a lot of benefits, extra care must be taken to ensure that sensitive data is not being unintentionally leaked or exposed.
MODZY DIFFERENTIATION
Modzy was designed to be completely agnostic of both data inputs and outputs. The API treats all incoming and outgoing data as opaque bytes, and only the models themselves need to know how to read and write the data. The API also enforces strict data access. You are only allowed to view the data you supplied, unless you have a role that allows you to see more. This is entirely controlled by an administrator.
Modzy is very careful about how data is stored. We support several deployment types to fit a wide variety of needs. You can deploy both on-premise or via a hybrid model, using your own network, databases, and file services. Modzy’s architecture is such that the components with access to training and inference data are tightly controlled and can be separated from the rest of the services. This ensures a simpler and more predictable data security setup.
WHAT THIS MEANS FOR YOU
Adopting AI into your business or organization can bring powerful new capabilities, but care must be taken to ensure that your (or your customer’s) data is kept safe and secure. Data access and storage must be thoughtfully considered when adopting new technology, especially in the case of AI, and this is a core principle at Modzy. We strive to make it easy to effectively and efficiently utilize data while simultaneously ensuring that that data is kept protected and used only in the manner in which it was intended.
Related Posts
Training Data Security
The effectiveness and predictive power of machine learning models are highly dependent on the quality of data used during the training phase. In most real-world scenarios, models are trained using domain-specific data provided by known and trusted…
- November 15, 2019
Retraining With Your Own Data
In simple terms, transfer learning is a machine learning approach where a model that is already trained on a specific dataset and developed for a specific task is reused as the starting point for training…
- November 15, 2019