# Processing with AI ## Exploration : IA & Ethics Name: > Clara de Veyrac > Subject: > Detect Student in class using Face Recognition >[TOC] ## Design brief ### Biais If we don't source our dataset with enough rigor, the following biais might appear: >1. Not able to recognize student everyday (change of haistyle, glasses, etc.) >2. Exchanging student that are looking alike (brother and sisters, same origins, etc.) >3. Recognize a picture as being someone alive We will ensure that our model is not biaised by: >1. Sourcing our data from the student themselves >2. Adding data on people looking alike that are not student to make it really precise >3. Making sure our data take into account that we need the face to be 3D ### Overfitting *I am not sure that the model can really overfit. I think that most of the threats are more about it being not good enough or being good but misused.* We will make sure our model does not overfit by > Checking that the model does not keep in mind all the faces it doesn't know. We want it to learn from the faces it alredy knows (student) but we do not want it to learn new faces. ### Misuse >We have to remind ourselves that our application could be misused by **the school** to **check the stuents when they're not in class**. It could also be a mean to always know where is someone when he's in school.In tht case the school would be over controlling the students (where they are, when, with who, etc.). ### Data leakage >In a catastrophic scenario, where all of our training dataset were stolen or recovered from our model, the risk would be that the thiefs use the technology to do something else than detect student presence.They could also use all the data about the student faces and either sell it or use it to create all those things like porn etc. It could be also a mean to blackmail them when they will have bid jobs, responsability or something that can interest the thiefs. ### Hacking > If someone found a way to "cheat" our model and make it make any prediction that it want instead of the real one, the risk would be that the system would not work at all for him. They could be mark as present to all courses and never come. He could also give the "hack" to everyone he knows. It can be for classes but also for exams etc.