How could an attacker poision the training data?

bake_betty

I came across the following definition of Backdoor attack (in this paper ): These attacks are accomplished in two steps. First, special patterns are embedded in the targeted model during the training phase, which is typically achieved by poisoning training data . How could the training data be poisoned? Isn't the training data local to the software developer who is developing the ML algorithm? And won't he train the data on his local machine...

Originally asked by Amit wadhwa on Stack Exchange (ai) (CC BY-SA 4.0).

uphilltech

Good thread on this — worth a read before you pay anyone to build it.

Discussion: https://ai.stackexchange.com/questions/28228/how-could-an-attacker-poision-the-training-data.

Want it set up for you? UphillTech.