Hi Tim, sorry.
On 11/10/2015 12:33 AM, Tim Wilson-Brown - teor wrote:
On 11/06/2015 08:41 PM, Tim Wilson-Brown - teor wrote:
The directory authorities are generally more concerned when
they *don't* stop it afterwards, and instead keep running it,
perhaps with extra logging, packet dumps, and decryption via
seized private keys.
Is there a reliable way for an operator to detect this?
If the directory authorities have removed the relay from the
consensus, only they will see its fingerprint in their logs.
If they haven't, and it's running on the public network, the
fingerprint will be visible in Globe as a running relay.
I meant is it possible for a relay operator to detect if a snapshot of a
running VM or VPS has been taken? Asked slightly differently, if I have
a relay running as a VPS or VM, can I somehow detect if my provider took
a snapshot of the relay without informing me?
As far as I know, there are ways to detect some kinds of access, but I'm pretty sure there are exceptions.
Pedantically, I'm not even sure it's possible:
* to detect all memory snapshots on some physical hardware configurations.
* to detect whether you're running on a VM/VPS or dedicated machine (but most hypervisors will tell you).
However, it's worth noting that the Tor network security model relies on semi-trusted directory authorities and untrusted relays.
Occasional compromises can be tolerated while maintaining overall network security (of course, some specific users of those relays may have some specific activity disclosed, particularly if information from multiple sources is combined).
Following from that, are uninterrupted snapshots of running VMs possible
in all hypervisors or should we be using the provider's hypervisor
technology choice to inform how we decide which providers to use?
I don't know enough to answer this.