- Someone listening
- A copy of your traffic changes nothing the phone can see. Encryption is the defense, not detection.
- A full relay with a valid certificate
- If the attacker passes your connection through to the real site, or holds a valid certificate for the real name, the phone sees what a browser sees. A site owner watching Certificate Transparency logs could notice. The phone can't.
- An attacker who never lies to the checker
- Every probe Owl sends can in principle be told apart from a real app's. I can raise the cost a lot, and I can't make it impossible.
- Who is behind the Wi-Fi name
- Android hides the access point's name and address from apps without the location permission, and Owl never asks for it. A copy of your Wi-Fi with the same password looks like the original.
- The neighbor table
- Android 10 and later block apps from reading which devices sit on the local network, so an ARP spoof can't be seen directly.
- Fake cell towers
- I found no public Android way for an app to see which tower it is on, or to read the phone's warnings about it, and tower identity needs permissions Owl won't hold.
- A route hijack far away
- A hijack of someone else's network is visible to the phone only through symptoms Owl already tests, like a certificate mismatch.
- A person clicking through a warning
- Owl can say a name is fake. It can't stop a tap on a certificate warning, and that is how the 2026 router campaign worked.
- TunnelVision
- This attack hides traffic with routes pushed by DHCP option 121. Android's DHCP client ignores that option, so it doesn't apply to the phone.