As artificial and machine learning technology grows and integrates with our products and services for better per- formance and enhance the experience, with the great success of these technologies. Data privacy concerns have become a topic that needs to be addressed. With Companies and institutes trying to use such technology to help us solve real-world problems and help mankind with AI that can assist in the medical field to AI that can help moderate traffic in our busy streets the one thing that we need in common across different situations is the correct data which will help us to creates the said intelligence to helps us. But to archive such a state we have to be in the position of collaboration with our fellow people to help each other to develop. A zero trust system has to be created so that we all can benefit from technological advancements. Regarding the use of ML and AI in the heterogeneous system, we need to be able to share the intelligence that a said party or organization has acquired with others without sharing the data a key thing. To help with that a concept called federated learning has been introduced which will help them share the acquired intelligence. Although the method is a far safer alternative than sharing the data but still has some pitfalls that need to be taken care of. One such problem is the creation of false intelligence (poisoning the data from which the intelligence is derived) and that being shared with others and being used to develop an AI or ML system, this means the created system now has the false data intelligence which may reduce it overall usefulness or some time may completely be unusable. Since these heterogeneous systems like the internet, these systems are implemented we cannot know the creditability of the source of the intelligence (ML or AI model). This paper proposes a security firewall framework that will help the developers of such a system to create defense protocols to better defend the system from such data poisoning attacks..

Federated learning, Model analysis, Model poi- sonousdefence strategy