Only if his internet activities cause the ISP to watch your connection or local LE to raid you. Otherwise, what he does on his computer has no effect on your computer, or your security, or your Tor circuits. There is a theoretical attack where an adversary can determine which sites you are visiting by running a large number of relays, then looking for combinations of simultaneous destroy cells in the circuits passing through them. Your Tor client sends destroy cells when you shut it down. So let's say you build three circuits and are currently using one of them to visit web site A. You are unlucky enough to have picked the adversary's nodes once in each of your circuits. kmf has described it a few times on the forum. This is how it looks: Circuit1 -> entry -> middle -> bad exit -> site A Circuit2 -> entry -> bad middle -> exit Circuit3 -> bad entry -> middle -> exit You shut down your Tor client and send destroy cells to all the relays in those three circuits. The adversary notices that his relays (the "bad" ones above) get destroy cells at about the same time, so he concludes they came from the same client. Based on Circuit3, he knows who you are (your IP address). Circuit2 simply provides more confirmation in this case. Circuit1 tells him which site you are visiting, and that is how you get pwned, because he knows who you are and what you are doing. If this attack could work at all, considering the large number of Tor clients and continuous circuit destruction that they see, it would mostly work on people visiting clearnet sites. The attacker would have to be both one of your entry guards and one of the hidden service's entry guards in order to identify which hidden service you are visiting, and that is much less likely to be the case than if he was one of entry and exit nodes. The simple solution to this attack is to do exactly what you described: create a new identity, thus creating new circuits, before you shut down your Tor client. I think it is highly unlikely to happen and probably not worth doing, but that is a matter of academic debate.