# Dicussion with Koon Tong 22 Apr 2020

* To declare a channel secure, there needs to be ideally a 2sqrt2 violation since anything else implies an advesary having the potential to entangle her anxilliary with the channel states. 
Even when noise is taken into account when the violation is low, this could mean that Eve is controlling the noise and making you think that it's high and accept a low violation, only to use that as a cover to tap on a channel whose transmission is actually high.

* This may be overcome with this method:
The channel noise may be characterized with the forward pass.
By sending random states where some are signal, and some are characterization states, very much like the BB84 protocol, we are able to characterize the noise of the channel; Eve doesn't know when you are characterizing so she won't know when to boost the noise and when not to in order to fool you to thinking that the subset of characterizing pulses have a higher noise / loss, when the channel really doesn't, and take advantage of that for the cryptographic pulses. She could boost the noise all the time, but then it wouldn't be advantageous for her anymore.
This technique is the subject of William Ignatious's thesis (double decoy). 

* Now we think of adapting this technique to certify that if the forward channel noise / losses has been characterized and a Bell violation consistent with these values monitored for security, then that must imply that the backward channel is also secure. 
This assumes that Eve is unable to distinguish the photons by their direction, and manipulate the noise / loss characteristics in the way described before, but only on the reversed direction e.g. with a circulator. This is equivalent to using a circulator to implement an intercept-resend attack. But if we assume a symmetric channel, with symmetric characteristics, then it might be possible to sell the paper. KT thinks that this scenario is rather artificial, because it is selling something that solves a constrained problem.

	* There is also a slight issue that the round trip has more losses than the forward trip. Even if the losses are the same in both directions, it will not result in the same violation. However, obtaining the same violation for two different measurements (forward and RTT) is not the aim here: we want to know if we can have the same violation in two of the same length, but opposite direction.

	* If we receive a photon in the backward direction, must it necessarily come from backreflection? The "security statement" would be: assuming a symmetric synchronization channel, a Bell violation in the forward direction implies that the backward direction is also secure." 
	Is this statement self-evident?