C4W1 Deploying models on mobile devices

I can see the advantages of deploying models on mobile devices (latency etc.). Question: what is the security of the model? Can any one not re-engineer it and use it for her/himself? Do the ml infrastructures and models (e.g. tensorflow lite) have a facility for avoiding this?

Consider android. It’s possible to make reverse engineering harder by turning off debug mode when deploying a build in production and let google play store manage your bundle along with the signing key. Other than that, I don’t see lite specific options. See if model obfuscation can be done on your target platform.

Thanks so much. I will check this.