[Cryptography] The Laws (was the principles) of secure information systems design

[John Denker]

On 07/21/2016 04:08 AM, Jerry Leichter wrote:

> I'm not sure of the continuing validity of Morris's analysis.  Big
> data, sophisticated analysis of metadata, all kinds of correlation
> analyses, have made it very difficult to know exactly what data I can
> safely "not care if its known".

Good point.

I would go farther and say that Morris analysis was naïve
when first formulated and even worse now.  Most non-experts
seriously underestimate how important it is to protect
little details.

Let me suggest another aphorism for the collection.  As I've
been saying for years:

 137)  Metadata is data.

Corollaries include:

 137a)  Any system that leaks metadata is a system that leaks.
 137b)  Stealing metadata is stealing.

As another way of saying almost the same thing:

 *) Traffic analysis is a Big Deal. 
    Ignore it at your peril.

A great deal of what is nowadays called "big data" is traffic
analysis and variations on that theme.  It's been going on for
eons.  The main recent change is an increase in the number of 
people doing it.

It is embarrassing how badly this is handled by typical internet
protocols, especially considering how long the problem has been
allowed to fester.

It affects not just primeval protocols such as TCP but even
relatively recent things.  For example, SNI came out in 2003
with no provision for encrypting the virtual hosthame.

There's been some chit-chat about extending TCP to include
encrypted headers and cover traffic, but this doesn't seem
to have had much impact.