Processing data through the G2M platform involves multiple steps. Data security is built at the heart of the platform so your data is secure every step of the way. When creating a new model you can choose one of three security modes. Once the model is created the security mode can be downgraded but cannot be upgraded. We recommend using the encoded data mode, which provides two layers of security while also allowing you to share or integrate your model. In all cases your data is encrypted in transit using banking-grade TLS encryption. The three security modes are as follows:
Unencoded data. In this mode your data is still encrypted in transit using the same TLS encryption banks use, but the data is temporarily stored unencoded during processing. The modeling results and supporting metadata are also stored unencoded. This may be necessary if you are planning to integrate your G2M platform API tenant with other systems, e.g., ETL orchestration or CRM, and these systems cannot access the G2M model encryption keys.
Encoded data (recommended) . In this mode in addition to TLS encryption in transit, the data is also encoded at the source when it is ingested by the G2M platform application. All categorical variables and all numerical variables will be encoded using the G2M platform's patent-pending homomorphic encryption. Note that in the case of A/B testing analysis (propensity score matching), individual numerical values for the outcome variable will not be encoded at rest. Encoding keys will be stored both on the user's local machine and on the G2M platform backend if you need to share your model with other users or access your model from multiple machines. The modeling results and supporting metadata are stored encoded and in a backend that is segregated from the staging area where encoding keys might be shared.
Enhanced security mode. In the enhanced security mode, all data is encoded as described above. Sharing is disabled and encoding keys are stored on the user's local machine at the time of model creation. The model will not be visible or accessible from another machine. Any personally identifiable information (PII) contained in categorical or numerical fields in the data will be confined to the original data source and the local machine paired with the model at the time of creation. No PII will be stored in, temporarily cached in, or transmitted to the G2M platform.