Innovations in Artificial Intelligence (AI) has become wide-ranging. The lifelines of AI are big data and big computing machines. However, the concern for cyber-security can present a threat to advancement of AI. The greatest AI innovations happened in a distributed manner where many teams solve a single AI problem. This is possible only when data are shared openly. In healthcare applications, such sharing of data is impractical. Medical institutions are increasingly not able to share their data due to tightening of data protection laws. Current method of sharing through remote server is cumbersome, insecure and inefficient. This way of sharing does not allow for distributed innovations and utilization of large super-computing facilities.

This project develops a secure data sharing solution for remote AI developments. The novel solution is built on top of Cloak Apps platform, together with A*STAR’s Bioinformatics Institute (BII)’s technologies for AI on clinical data.


The resulting technologies are available via Cloak Encrypt, a free-to-use privacy protection solution with the following advanced security specifications:

1.Encryption: client-end encryption, client-end key generation, unique cryptographic key for each user, AES encryption (256 bit), and RSA encryption (2048 bit)
2. Single Sign-On (Google, Apple, Facebook, Dropbox, Box)
3. Enterprise Single Sign-On (Active Directory through *Cloak Gateway)
4. Multi-Factor Authentication (with *Cloak Gateway)
5. Key Management Tools for lost certificate recovery, endpoint migration, and user revocation
6. Sharing with no file password required, pre-defined sharing list, owner notification and approval of file access request

The Cloak Encrypt app is available from:
• Web app: Cloak website
• Mobile app: iOS or Android

A histopathology inference web application has been developed to demonstrate how trained model can benefit the user, while enforcing good data security. The technologies remain in use by BII in their research projects to securely operate on biomedical data.


Marcus Tan: