Re: AEAD in ssh
denis bider <[email protected]>
| Newsgroups | gmane.ietf.secsh |
|---|---|
| Message-ID | <[email protected]> |
If my implementation consistently sends one ethernet frame (say 1450 bytes) every 10 ms, this defeats traffic analysis at a cost of ~ 1.1 Mbps of fixed bandwidth. Netflix uses 3 - 5 Mbps of fixed bandwidth. This protects a terminal session from keystroke analysis, while providing ample bandwidth. However, it needs encrypted lengths. ----- Original Message ----- From: Peter Gutmann Sent: Wednesday, February 24, 2016 03:52 To: denis bider ; Watson Ladd ; Niels Möller Cc: Bryan Ford ; Daniel Migault ; [email protected] ; Stephen Farrell ; Mark D. Baushke ; Curdle Chairs Subject: RE: AEAD in ssh denis bider <[email protected]> writes: >That implementation, with unencrypted packet lengths, sabotages any hope of >thwarting traffic analysis, even through high-overhead SSH_MSG_IGNORE padding. With encrypted packet lengths and padding you're still not thwarting traffic analysis, you're just making the job a lot more difficult for the implementer. See my previous message. Peter.