The existing security methods in the cloud such as encryption

access control

and VM isolation can not guarantee end-to-end data security. To address this problem

a double-layer information flow control model is proposed. The definition of key element

centralized and decentralized information flow rules

capability adjustment rules

label propagation rules

and declassification rules of the model are presented. Then

taking the advantages of dynamic taint tracking and virtual machine introspection technologies

a prototype system named IFCloud are designed and implemented. IFCloud achieves information flow tracking and controlling as a service for cloud tenant and provides detection methods against common system attacks such as stack and buffer overflow attack for the cloud provider. Finally

IFCloud is demonstrated to be a flexible and accurate system that tracks and controls the sensitive data flow in the cloud at runtime from the function test results

and it can be applied to protect data security at a fine-grained level for the software as a service cloud.