Schneier on Security

Clive Robinson • August 30, 2017 5:37 PM

@ veritrasheum,

Since it’s not really off topic, I’m still hoping for some input on the previous link I dropped

The authors kind of say the most important thing themselves with,

Previous work in this space quickly lead us to believe it was possible, but we weren’t sure how robust these methods were or if we’d be able to recreate them. Generally, we’ve learned to be weary with academic descriptions of attack vectors — as sometimes they only work in a lab setting, expect certain conditions, or simply aren’t as practical as their authors make them out to be

I have a basic rule of thumb about attacks which is,

If the laws of physics alow it, then someone will do it.

It’s then only a question of when which is more a technological cost issue than it is anything else.

To repurpose the Coco Chanel saying “Side channel attacks are the new crypto attacks”.

The point is as I mention occasionaly the security end point has to be beyond the communications end point. If it’s not then all an attacker has to do is an “end run” around the security end point and get at the plaintext directly.

Once you accept that it then, as with all side channels, becomes a question of “channel bandwidth” available to an attacker. Either directly in an “active attack” or indirectly as “leakage” in a “passive attack”.

Thus the unseen battle between attacker and defender is for side channels and then how to exploit them.

After a little thinking you will see why I no longer talk of the old idea of “air gapping” but the more relevant these days “energy gapping”.

Back in times past when the SigInt agencies used to design their own equipment and systems before pushed at COTS solutions by point scoring politicians there were certain design rules used. The first of which is almost a direct translation of the “KISS principle” and is “minimization” of function thus complexity. Having minimized each function within a system you encapsulate it for “segregation”. With all communications between functional sub components going through instrumented choke points.

It’s easy to see why such systems were eye wateringly expensive. Conversely it’s also easy to see why COTS systems are not going to be secure.

Which brings up the “Security -v- Efficirnce” issue I likewise go on about from time to time. As a general rule the more efficient you make a system the more you open up side channels both covert and overt. However the opposit is not necessarily true. That is a less efficient system is not generally any more secure than an efficient system unless specifficaly designed to be secure.

An example of this is power signiture analysis of an algorithm due to unbalanced in time branch instructions. It does not matter in general how efficient the algorithm is written becasue the “test and branch” behaviour is implicit in the algorithm. It’s only when you are “in the know” about the power signiture analysis that you take the required action to balance the time in the two execution paths.

It’s being “in the know” about what side channel an attacker is using and how that is the important thing. Which is why the likes of TEMPEST and EmSec techniques are usually classified so defenders don’t get to find out about vulnerabilities in their equipment. Because the “old school” passive attacks got hit by the likes of EMC in the 80’s for a while[1] new attack side channels had to be found.

But high levels of integration of function and efficiency in “smart devices” has opened up a whole raft of ttansducers thus new vectors of attack.

The thing to remember is technical cost is strongly deflationary you in effect get twice the power for the same price point every ten to eighteen months. This alone will take a theoretical attack and make it first practical to a select few then expanding geometricaly till it’s feasable to all. To see this effect in action take a look of the history of “Software Defined Radio” (SDR).

As our host has noted what is theoretical today becomes a PhD project tommorow then a common technique the day after.

And that is almost always due to the deflationary “technical cost”…

[1] Put simply prior to EMC all electronics put out lots of EM radiation into the environment around it. These “Emissions” often carried secret information thus they were “compromising emissions”. The problem was and still is that often “transducers are bidirectional” –speaker is a mic and vice verser– thus if a device or system radiates emissions the chances are it is “susceptible” to other EM radiation. This showed up most notably with consumer radio equipment, as it became more in use during and aftet the 60’s the problems became unmanagable. Thus the idea of ElectroMagnetic Compatability (EMC) where equipment not only should not have EM emmissions it should also not have susceptabity to equipment that did have EM emissions as part of it’s functionality. The EM emission / susceptibility was originally reduced by physical filtering componets like capacitors and inductors. However these are expensive so manufactures realised with digital devices they could “cheat the test” by using spread spectrum techniques to “whiten” the frequency spurs spreading their energy across multiple frequencies thus reducing the single frequency amplitude[2].

[2] The problem with spread spectrum and other “Low Probability of Intercept” (LPI) techniques, is that whilst you can spread the energy across multiple frequencies the process is reversable if you know the “spreading code”. Thus the use of simple short length linear spreading codes and no filtering components by manufactures to reduce costs in the 90’s onwards was a gift to the SigInt agencies as in effect it took the back to the good old days of high level compromising emissions and high suceptability to EM fault injection attacks.

Clive Robinson • August 31, 2017 3:34 AM

@ wael, veritrasheum,

You forgot to add the smiley at the end of,

that can bridge it.