Schneier on Security

Clive Robinson • June 16, 2018 4:44 AM

@ moz,

IOT comes to padlocks. The result will make you smile.

The device appears to have been designed not by engineers of any kind, but by a Marketing Dept “subing it out” to a “China Knock Off”[1] house. None of whom appear to understand what even base physical security is about.

As anyobe with even a mear smidgen of knowledge knows the metal chosen yeilds to pocket power tools in seconds as will any “security pins”… Likewise “security screws” used in consumer products are generaly a compleat joke.

Having trained as a toolmaker many many years ago as part of my education I know no matter what funny shape they use for the screw/bolt head you can repurpose an existing screwdriver or similar hand tool like hex wrenches to fit in a very short time with just good quality needle files. I do it quite frequently when repairing electronic household items etc for friends and others. These screws/bolts are usually not designed to do anything other than keep untrained people out when the consumer device goes wrong. They are certainly not designed for security because they are virtually never “one way” due to rework and servicing under warranty requirments.

Even the supposadly “one way” screw/Bolts used in the likes of security fencing can have their heads repurposed with a grinding tool you can carry in a pocket, then an ordinary flat head screwdriver will work on it. The trick generally used to improve their security is to have a “Ny-lock” or similar nut on the “secure side” so that after half a turn or so of the screw/bolt from the “attacker side” the nut just spins and will not go further down the thread, even if a split tipped leaver –think small crow bar– generally insufficient force can be applied under the screw/bolt head for there to be sufficient friction on the nut to enable it to be unscrewed further. Thus from an attackers point of view “chemical” solutions[2] against the actual fence rather than it’s fixings are required.

[1] The name “China Knock Off” has been around for many years and originally applied to Taiwan that worked under “Made in the Republic of China” (ROC). It is only more recently with mainland theft of companies and IP has it been applied to the Communist Party controled mainland.

[2] By chemical I’m not limiting it to acids and similar corrosives, special forces tend to find a loop ot RDX / Detcord on small stand offs generaly punches a nice man sized hole in even concrete based fences and walls… Failing that an anti tank missile will work against even hardend command posts. Though as some one I used to know chearfully used to say “Shaped charges are your friend when you forget your keys” followed by a big smile and one of those thousand mile stares of fond memories. As a University Prof and researcher he was realy liked if not loved for his mad cap ways.

Clive Robinson • June 16, 2018 6:33 PM

@ Humdee, bttb, et al

Foldering is a legitimate security technique that isn’t “underhanded” any more than encryption is underhanded.

A little more on foldering[0],

Firstly and more importantly than a security technique at it’s simplest level it is a legitimate ad hoc way of doing “collaborative working” and faster than bouncing emails back and forth. Importantly it is a “pull” not a “push” method, so does not require the participents to run or have run on their behalf Mail Transport Agents (MTA) or permanent Internet connections.

Think of it historicaly as “Cloud working” before the “Cloud”. That is in times past systems admins in small widely distributed organisations might set up a “group account” on an email system as it was a lot lot cheaper than setting up a shared network folder. As a shared email account would work with “dial up” Intetnet accounts, where as a shared folder system would require atleast one permanent server and Internet connection which had other security concerns. You could look on it as being the predecessor of Google’s collaborative “cloud” systems for the likes of tablet/pad users.

Secondly with encrypted links to and from the server being standard these days if might appear falsely to offer a slightly higher degree of “confidentiality”. The reality is it does not for the colabarative asspect, because in most cases that requires the use of “plaintext”[1]. Which means the confidentiality depends on trusting[2] the server it’s owner and operators and anyone who comes into contact with their systems and backups etc… Which with NSL’s and other legislation and surveillance methods they can not be trusted these days in anyway shape or form, nomatter how trustworthy the individuals may appear to be[3]. Therefore any idea that you might get increased confidentiality by foldering for collaborative or other plaintext work on the server is entirely false.

Therefor if privacy or secrecy are required foldering should only be viewed as a “communications” method not a “privacy” or “security” method (which encryption can be but is not always).

Not understanding the difference between a communications method and a privacy or secrecy method is likely to give you a lot of grief later if you don’t understand the differences.

However foldering if used as a “Storage” method for time shifted communications then in effect it becomes a “drop box” so full ciphering or coding of a files contents and it’s file name meta-data can be easily achived as no extra support is required on the server. So it can be used to increase “privacy” but not “secrecy” on it’s own.

To understand the difference you need to think about it in terms of old school OpSec fieldcraft. A dropbox is in effect like a mail box in shared accomadation or a post office. It’s location and who uses it are in effect public knowledge. Whilst it might provide limited privacy of message content the meta-data of the communications is on public display, so it is not a secret communications method even though it gives privacy of message content. To get communications secrecy you need to avoid “Traffic Analysis”, which in old school fieldcraft involved setting up a “dead letter drop” in an anonymous and often changing location prefereably accessed via “cut outs” and controled by covert indicators.

For “Foldering” this requires several email servers from entirely unrelated service providers to give multiple “drop boxes”, that should also be set up anonymously. To make these drop boxes into “dead letter drops” you need to access them anonymously as well, not just on the human level but the network level as well. That is you need to access it from different IP addresses[4] unrelated to you at different times of the day or days of the week / month etc, and importantly only when there is something to exchange or if “cover is being run”[5].

Another issue is the recomended use of “cut outs” back when Tor and other privacy network systems were not blocked by service providers mix nets and onion routers provided reasonably good cut outs. However due to the perceived need of “low latency only” such systems were still susceptable to traffic analysis and some other attacks.

There is of course a further problem which is how to also run a covert signalling system. I’ve addressed this in the past on this blog when talking about how to run botnets in a headless manner[6], so don’t need to cover it again here.

Not carrying out all these actions will almost inevitably result in your secrecy being rolled back, which in turn can give rise to a whole heap of issues, only some of which can be mitigated.

Thus we have the old problems of OpSec / Fieldcraft is hard, and inexperienced people usually implement insecure systems often to their cost. It requires further a good deal of “hinky thinking” backed by considerable experience to set up new secure ideas, and as our host @Bruce has been known to note “thinking hinky” is a rare skill. It also means that any quick description such as this will contain insufficient information as well as the potential for ambiguity, neither of which are good if you decide to build your own system.

Oh and further as @Bruce also notes attack methods move forward all the time, so what was once secure, given time in all probability will become insecure. So all security systems need to be not just reviewed regularly, they also must not be dependent at any point on a single method. For instance encryption we know has a habit of failing long before expected (thirty years appears to be the tops). Thus you would be well advised to use two encryption systems in series, provided the underlying methods used are not common.

[0] In this case should “foldering” be a noun? Even a proper noun?…

[1] The reason for this is in most cases colaborative working means information “processing” at the server. Currently there is no easy way to “process” information that is encrypted into ciphertext and it is extreamly unlikely that a server owner would alow it any way even if it could be easily done. Which means that you would need to do any processing locally, which means downloading, decrypting, modifying, encrypting and uploading files. Therefor the server is realy being used as a “store and retrieve” communications node. Both storage and communication of information can be done with any type of file including fully encrypted ciphertext, so foldering will give privacy when used this way.

[2] There are two types of trust, the human type and that of trusted systems. They are in effect the opposite of each other. Human trust is basically where you are crossing your fingers and hoping other people will not betray you[3]. Whilst trusted systems are designed against a security policy and when opperating correctly
enforce that system’s security policy. Further the design of a trusted system is usually such that even if it develops faults or becomes subject to attack or in some cases even being seized by an adversary then it will still maintain the systems security policy.

[3] There is a very old notion, that the only way you can trust living people is by having more dirt on them than they have dirt on you. Whilst that might have been true at one point in the distant past, “Turning Kings Evidence” and in more modern times “Witness Protection Schemes” for entire families have renderd that method of trust fairly moot. As crime syndicates have found the LEOs can have even more persuasive argument and be very forgiving even to those who have committed multiple murders and worse. Thus the old “Three can keep a secret providing two are dead” starts to take on additional meaning in the modern world[7].

[4] These different IP adresses should however all be from an apparently localised area such as coffee shops / bars / etc in and around one business district. As this minimizes the “drawing of attention” to the account. Nothing says Foldering to an email service provider like two IP addresses in different countries “ping-ponging” fairly rapidly, other than perhaps a Tor exit node or known VPN provider…

[5] Running cover is a whole other subject, but in this case it referes to using the account as what appears to be a legitimate email account such that the signal to noise ratio on the Foldering or other private / secret messages puts it well down in the noise floor. There is a lot more to it than just traffic volume ratios, but it is a very lengthy subject in it’s own right involving amongst others coding / stenography, compression encryption, steganography and cover stories / legands.

[6] Most recently at, https://www.schneier.com/blog/archives/2018/06/router_vulnerab.html#c6776643

[7] During WWII it was considered a very real possability that Britain might be invaded. Thus the idea of “stay behind” or “behind enemy lines” irregular forces were considered in depth. It was realised that there was an issue via training, in that it ment people who should not know about each other would thus information could be tourtured out of them if caught. So a solution was arived at by those in charge. All training of group members and all communications with the agency would be carried out by only one member of the group, the leader. For the number two of the group the leader would pass down “sealed orders” the leader had never seen. The orders were only to be opened if the country was invaded and the groups location over run by the enemy. The first order in those sealed orders was that the number two should kill the leader and assume the leadership role. Oh unbeknown to those that came up with that idea and those at a senior level in the organisation, trainers and those who implemented the idea there were also “Kill as required” orders… Bringing a whole new meaning to “Don’t worry, I’ve got your back”.

Clive Robinson • June 16, 2018 7:48 PM

@ bttb,

thegrugq on “foldering”

I gave up on “thegrugq” some time ago, for someone who talks the talk on security his practices (use of twitter, Tumbler, Medium) are far from secure as they are data mining organisations.

Trying to use either Twitter or Tumbler witout “full tracking” methods (javascript cookies) enabled does not work. Medium can be used soet of but it’s a mess, trying to find the index is at best a chance operation by the majority of potential readers. Whilst some of what thegrugq does is “in the moment” in the way some social media works, much of his stuff if properly indexed and accessable to mobile users would be a usefull resource.

Both he and other “security gurus” should realy know one heck of a lot better. In fact one way to rank those who have some security status is just how securely they treat those they diseminate information to and gow easy to use the information when not the current article. On that scoring thegrugq is way way down the ranking currently.

He realy should “get with the program” especialy since the Europran Directive came into force… There realy is no excuse.

Clive Robinson • June 18, 2018 9:13 AM

@ All,

A silly sounding World Cup factoid…,

Apparently when Mexico won against Germany enough people in Mexico jumped up and down to set of earthquake detectors…

Why is this “silly sounding” but actually of “security interest”?

Well many moons ago somebody postulated that China if it could get all it’s Billion citizens to jump up and down in a synchronized way, could use them as an “earthquake weapon”. This was immediately “poo pooed” at the time with various expert –talking heads– rolled out to say it was not possible…

Well now we know that in part they were wrong, enough sychronised human jumping up and down can set of detectors.

Now whilst they may not be capable of doing damage, it could be used as a form of DoS weapon causing false triggering thus panic etc.

Sometimes the most unlikely of things can have security implications…

Clive Robinson • June 18, 2018 12:35 PM

@ MarkH,

I looked in on this thread, to see any discussion on how the FBI purportedly retrieved messages exchanged via the Signal app.

Put simply they did not retrieve any signals in transit as far as I’m aware.

However as I’ve repeatedly pointed out the securiry end point with all these apps is in the wrong place. Thus all the FBI has to do is get at the User Interface in a number of ways and it’s “game over”.

Thus there are a number of options available to them if they can be bothered to put some effort in.

1, Use the Over The Air interface to “end run” attack the Apps User Interface.

2, Place a key logger of some kind on the device which can be done in a number of ways to get at “user secrets” at the User Interface.

3, Get direct access to the device and use some form of coercion on the user, or appropriate forensics devices/code.

So these apps are not secure, and having them on your device makes you “guilty” in the prosecutions eyes irrespective. And judges appear to give the prosecution way way to much leaway without any supporting evidence. The prosecution just waves it’s arms and says “here be big scary monsters” or “think of the children” or some such nonsense.

The aim being to force the defendent to plee bargin or bankruptcy or some such. Basically to deny a fair trial. Such prosecuters need to be permanently removed from such positions, preferably by being made totally bankrupt themselves.

Clive Robinson • June 18, 2018 4:32 PM

@ Mark H,

I quickly ascertained that such a standard was infeasible.

The feasability of finding evidence at a given level is not the issue.

What is at issue is the response that follows on from that evidence.

Let me put it this way the US has seriously talked about “Kinetic Response” to “Cyber-Atacks”. Thus if you are going to bomb the crap out of some location by drone etc what level of evidence would you want?

Because the one thing we do knoe with very little doubt is that false attribution is easy way to easy. Thus creating a situation with false attribution being the aim would not be very difficult.

So ask yourself the question what would you feel if Russia sent in a cruise missile into a US city and blew up a block of a business district, because they thought they had “evidence” of people attempting to hack the Russian infrastructure from a computer in that location?

As I said it’s not the evidence that concerns me but the actions taken on that evidence. If they eant to have diplomats being rude to each other or being kicked out of the country fine that generally is just so much ruffled feathers. But going kinetic where people will die and infrastructure will be destroyed thats altogether different. We know from the ME that the US has got it wrong numerous times with drone attacks already and many innocent people have suffered as a result. What do you think would happen if the US did the same to a city block in downtown Moscow, Beijing or other states with the capability to retaliate not just in kind but more so?

I get the feeling that the leaders of several super powers don’t believe in MAD these days. Thus the danger of escalating into all out war is on the table…

That is why I insist on high evidentiary standards and others should as well, especially if it’s members of their families and friends that are going to be the cannon fodder.

Clive Robinson • June 21, 2018 1:49 AM

@ vas pup,

We have not too much knowledge about such synchronized mental/biological power yet.

Just a few years ago science had kind of ruled out ESP abilities and had also ruled out other EM relayed activities with regards the brain of most living cretures.

However the “plants have feelings” crowd gave science a bit of a kick up the backside by demonstrating that sick plants had different electrical fields and that other close by plants would respond to the sick plant. Whilst cause and effect were not seperated from what I remember it did start people investigating.

Roger Penrose went out on quite a long limb over quantum structurrs in the brain. Whilst none have yet been found there they have in other biologicals. The recent discovery of “lower than red” photosynthesis has generated yet more renewed interest in such quantum effects in biologicals.

But backto the brain we now have proof that the brain will respond to EM signals and “mag pulse” systems are starting to be used in medical therapy trials to try and ease if not stop the biggest epidemic heading most of our ways which is dementia.

So I would not rule out further research showing us even more interesting results…

Clive Robinson • June 22, 2018 3:49 AM

@ Bruce,

It’s Cephalopod Week! “Three hearts, eight arms, can’t lose.”

Yet the “Grim Reaper” collects the harvest…

The solution to this would appear to be “brains” whereby you invent your way to a longer life.

Speaking of brains some cephalopods are somewhat disadvantaged in this respect, in that they have a “hole in the middle” of theirs. Worse we talk of “heart burn” as a symptom of excess eating, those brains with a whole in the middle have the equivalent of the esophagus going through, do they suffer from “brain burn” as a consequence?

Clive Robinson • June 22, 2018 6:00 PM

@ Bob Paddock,

Any insights here Clive?

As an engineer by training as I did not have the money to put myself through University at the usual age to become a scientist, I tend to make decisions based on what I can measure in a manner that is of experimental use.

As I tell trainee engineers and scientists, the most important thing they need to learn is “testing techniques”, not only is it a fundemental part of the scientific method, you do not learn anything new when things do not go wrong or you don’t fix them when they do.

It’s one of the reasons “ITSec” gets my goat, as nobody wants to come up with rational and testable measurands (it would wipe out the lucrative “Snake Oil Market” ITSec products have become). Even information theory is lacking in appropriate measurands for more than just it’s statistical observations.

As for those who write programs, and give the testing to others, “nugh said already” 😉

Thus I can usually tell when measurands are missing and can sometimes describe the required properties they should have to move onto the next stage of observational verification. Worse though I can spot a number of “experimental errors” or poor methodology.

The most famous example I can think of this is the “Rat Maze Experiments debunking”. As you are probably aware there were a very large number of papers written about rats food treats and mazes. A researcher from another field was skeptical of what was being claimed, so repeated the experiments. To cut a longish story short the researcher found that all the tests had significant failings in methodology, and when the errors were eliminated, unsuprisingly the experimental results were not the same as all those previous experiments that were written up and published…

Thus I am cautious about any paper written where recognised measurands and methodologies are not used, as there is little of predictive use, or to aid in repeating tests by others to confirm the observational data.

That is not to say that the use of new methods or measures is verboten, just that they should go through the “mill of acceptance” prior to their use as much more than observations and hypotheses. If they do survive the mill then they can be used for verification and then prediction.

As many know we actually realise that we know less than there is to know, and there is a finite limit on what can be known. Thus there will always be things that are unknown to us, and with time what we know now may well become forgotten knowledge in the future. That is, in the same way that history shows there is now forgotten knowledge from earlier human history some of which we rediscover.

Thus a number of simple predictions can be made but of sufficient generality that they could be considered either banal or obvious. One such is that our understanding of not just the theory of mind put the practicalities of the bio-mechanics of the brain will improve. Of less obviousness is that at some point we will find out if the human brain does have quantum abilities as some other biological processes are now known to do.

But will we find out if human creativiry is purely based on complex yet determanistic rules or on random processes or a combination of both?

We may never know due to some fundemental questions we may never be able to answer.

It’s currently not possible to say as we do not have a testable method of correctly identifing by observation of the outout alone the difference between determanistic and random processes, and in all probability we may never be able to.

Which gives rise to the possibility that “random” does not realy exist. If it does not then we might have to consider the possability that we have determanistic but chaotic systems that are in effect one way. That is they are so sensitive to initial conditions at any given point in time that they quickly progress beyond a point where they can be “wound back”.

Theory has an axiom that a closed process has memory of all previous states/interactions. But the reality is it is not possible to measure a system with sufficient accuracy to carry out an experiment to verify this (lesser flea / turtles all the way down issue).

It is also dependent on if the universe is truly analog at all levels or quantitative at some level below which you can not go. If the latter is true then it is not possible for a system closed or open to have “memory” of all previous states/interactions (see “Cantor’s diagonal argument” that was used by both Gödel and Turing in the 1930’s to answer certain fundamental questions).

One thing we are aware of is that “madness has stalked early contemplaters of infinity by rigours process”, but not by later mathmeticians. Various suggestions have been made for this including that of inheritable genetics. Which if true has other implications we would probably rather were not answered.

Which brings us to an awkward philosophical question. We know that there are limits on our knowledge based on our current understanding of the physical finite universe. Thus the question of “What do we want to know and what do we not want to know arises?” there is a certain logical reasoning that says we have no choice in the matter…

What we can reasonably safely assume is that what we currently understand as “human” will be gone long long before the universe reaches even close to maximum entropy. But that actually is saying little as we may continue to evolve into forms we consider not human by our current standards.

Anyway it’s midnight in the UK and I have to get up early tomorow/today 😉