Re: [TLS] DoS risks from draft-vkrasnov-tls-jumpstart-00

Hi,

For the first part - that is a regular amplification attack, which in that case has lower amplification factor than many other existing UDP protocols. 
I also address this in "Security considerations” chapter, where I propose a general technique to mitigate attacks of such type, that may apply to other UDP protocols as well.
For the second - the server chooses how much history to keep, and the timeout period for initiated handshakes. If the history is full, it will start ignoring incoming ClientHellos.

Under normal operation a typical serve can handle all incoming connections. Under attack it can easily fallback to legacy operation. 

Thanks,
Vlad

> On 15 May 2015, at 03:37, Watson Ladd <watsonbladd@gmail.com> wrote:
> 
> Dear all,
> 
> I'd like to explain some of the ways in which
> draft-vkrasnov-tls-jumpstart may provide opportunity for mischief. The
> first, and most obvious, is that responses are considerably larger
> than requests, leading to amplification. The second is that each
> request creates state that is supposed to be stored on the server,
> leading to exhaustion of server resources. There does not appear to be
> a cookie mechanism to mitigate this problem.
> 
> For resumed connections these problems are much less serious. It may
> be that resumption only is a better fit for this approach to latency
> reduction.
> 
> Sincerely,
> Watson Ladd
> 
> _______________________________________________
> TLS mailing list
> TLS@ietf.org
> https://www.ietf.org/mailman/listinfo/tls