Monthly Archives: May 2014

Effectiveness of security controls in physical security

Lately I have been giving some thought to effectiveness of security controls in relationship to physical security.

To do so requires a definition of what one will consider “effective”. If we can accept that an appropriately motivated and well-funded attacker can bypass all mitigations our definition of effectiveness must be based on the skill and motivation of our attacker.

This also means we must give thought to how quickly we can respond when we become aware of an attack.

This is in essence the same approach we take when designing secure software systems. The core difference being we also have to consider the physical properties of the space we are protecting along with the human factors of the design of the system.

On the topic of physical properties of the space we have to consider the materials it was constructed with. One of my favorite examples here is how in the movie Sneakers Robert Redford’s character bypasses a security keypad by kicking the door down.

In this case even if appropriate door hardware was in place he could have simply gone through the wall like in this case of the recent burglaries of some Best Buy stores.

The state of affairs with home security is even in worse shape. Locks on doors are often of low quality and are trivially bump-able the plethora of home alarms that are installed typically use wireless sensors that can be bypassed with just a few dollars of electronics, garage doors can often be bypassed with nothing more than a bent hanger and if that wasn’t enough just throw a rock through a window.

So given this sorry state of affairs what should we do? First we need to be realistic about what the risks, the assets we have to protect and the value to the attacker. Then we should only invest proportionally to those variables.

To do this we need to develop a solid plan of what intrusions we have a chance of detecting, when we do detect them how quickly we can do so and then what our response will be in each of those cases.

For this reason a good quality alarm system is very important but they don’t do much good if they are not activated. Once activated we need to think about how long it takes for the monitoring service to be called, once contacted what do they do and how long does it take? Also In some cases the police will refuse to respond to a call that has not been confirmed by someone on-site and if they do often times responses can take hours. With that said knowing what the response time is is invaluable to understanding how long your mitigations will need to withstand attack.

But if as they say “Locks are for honest people” why do we bother at all?

The answer is that well-thought out mitigations do act as meaningful deterrents that can significantly reduce your risk but more importantly having proactively considered the risks and built the corresponding mitigations you are positioned to reduce your exposure (see my recent posts on Why shouldn’t you use safe-deposit boxes to store Bitcoin? and Insurance and Bitcoin) and ensure that such events are survivable.


PiperWallet First Impressions

So I just got my PiperWallet. For those of you not yet familiar with it the PiperWallet is an open-source hardware bitcoin wallet based Electrum running on a RaspberryPi paired with a built in thermal printer in what looks like a 3D printed chassis.

The basic idea is that managing cold wallets is hard and it doesn’t have to be.

Even though I have only started to play with the device overall I am impressed. Here are my initial observations:

  1. It was packaged well considering the volume in which they are produced;
  2. The quality of the casing is also good considering the volume;
  3. The cut outs are a little rough and are larger than the connectors they expose;
  4. The primary “indicator LED” that is used to show that the device is booting is not terribly bright;
  5. Without reading the instructions (or waiting a sufficiently long time) it’s not  obvious when the device is ready;
  6. The print button LED is bright and of excellent quality;
  7. There is no positive feedback when the print button is pressed.

So far I am happy with the purchase though I need to do some more playing with it before I make any final conclusions.

With that said here are the things I think I would change if it were my product:

  1. Make the serial numbers on the paper wallets randomly generated; you un-necessarily leak information by using monotonically generated serials;
  2. Add tamper evident seals to the casing so that if the device is opened during shipping it is obvious;
  3. Add tamper evident seals or “plugs” over the ports exposed on the device, possibly even dummy plugs with seals so its clear nothing happened to the device as part of shipping;
  4. Add per-device fixed wallet keys to be used as a serial number to the back of each case (there is a wallet address but I believe this is an address of the Piper team);
  5. Use per device passwords shipping them on a form similar to the one I provided here;
  6. Replace the indicator LED with one with a similar brightness and quality to that used in the “print button”;
  7. Add a small LCD display that can be used to provide real-time feedback and status so it’s easier to use when headless;
  8. In the documentation included have the steps to verify what software is running on the device along with hashes to do so.

Verifying a Bitcoin Wallet Address

Before sending someone a large sum of money on the internet via a irreversible transaction you better make sure you are sending the funds to the right address.

There are a few ways to go about doing this and depending on who you are sending funds to, how accessible their keys are and what the capabilities and behavior of their wallet software is you may need to choose different solutions.

Have the recipient sign a message using their wallet key

If we assume the recipient has the key associated with the target wallet online (aka not in cold storage) and that that the software they use for that wallet supports message signing with wallet keys this can be a viable option.

Unfortunately there is not currently a standard for the format of signatures using bitcoin keys with that said thankfully there appear to only be two common formats in-use today.

The first format being in-essence no formatting; client simply present you the three values you will need to verify a message and you do with them as you see fit, for example:

  • Wallet Address: 18neTpQ5MWnXg4n4rpoK5TgxXjEVcg2MYR
  • Message: [email protected] – my voice is my passphrase authenticate me
  • Signature: G0d6BnQem1gT4nd9esfsEyn1k/GfYAxDkNJmkNvmz8wCOI2Ncw9DvIcyP7OJcEvWbUHQNIBFK3V8wYdnhEFhYHI=

This format leaves a little be desired. For one you have to pass these values independently and then you also have issues around introduction of white-space which can invalidate signatures.

There is another increasingly common format that leverages ASCII armor and some codified rules to address these issues. This style of formatting originated in a project called Privacy Enhanced Mail (PEM), it was one of the first proposals for how to sign and encrypt mail on the Internet and was later adopted by PGP (RFC

But don’t confuse this format with these other formats they follow some different rules when it comes to encoding.

What this means is that depending on the implementation of the wallet software the recipient uses you may not be able to validate the signature they produce without some manipulation of the text.

As for what this format looks like, its fairly straight forward:

[email protected] - my voice is my passphrase authenticate me

The core differences with this format (as specified in this thread and the PGP rule-set are:

  • No “empty-line” delineator between the headers and message;
  • Beginning and end whitespace / newlines ignored excluded when verifying the signature;
  • Length of rows are not limited to 80 characters;
  • No concept of header values (like versions).

The reason I point this out is that since there really isn’t a standard for this signature format and the format diverges from what has been used historically you may still encounter interoperability issues when validating messages between clients that have not been tested with each other.

With that said when you have managed to successfully verify a message like this you know that whoever produced the message owns the key associated with the wallet associated with it.

To address the risk of a message substitution the sender would need to communicate a challenge out of band to the recipient. For example you may notice in my message above I included “my voice is my passphrase authenticate me”. My inclusion of this message (presumably exchanged out of band) helps assure the sender that it was me who signed the message.

To make this process a little easier Andrew Yanovsky and I put together a simple site that can validate both formats, it’s all client side so you can save the files locally and run without the dependency on the website if you like.

NOTE: It is worth noting that this workflow does not accommodate P2SH and multi-signature wallets both of which will see increased use as time progresses.

Do a micro-transaction

The simplest way to verify an address is to simply send a small amount of money to that address and verify out of band with the recipient that they confirm seeing it in their balance. This is what most online payment services but again this requires the keys to be accessible to the sender they can perform the transaction.

There are a few things to keep in mind if you go this way, specifically:

  1. Don’t send less that .0001 BTC because the transaction may get “stuck” and not be processed.
  2. Be sure to include some transaction fee even if tiny so it doesn’t stay unprocessed for too long.

Once the transaction has been sent and you use a tool like to see that the transaction has been confirmed you can verify out of band with the address owner again that they see the funds as well.

This approach unlike the wallet signing key approach can also work with multi-signature and P2SH wallets which will be in use increasingly as clients better support these techniques.

Verify the wallet address two times via out of band channel

If they keys are offline (in cold storage) the only viable option is to carefully validate each character of the address via an out of bound secure channel, I would personally not rely on this approach for large sums but if both parties are careful it can work. By doing the check twice you reduce the chance of human error but mistakes can happen and in this case they can not be undone so use this approach with caution.

None of these solutions are perfect and moving forward I expect we will see services like and exchanges with authenticated account profiles will become the way that we solve these problems but in the mean time you can reasonably manage the transaction workflow via these two mechanism.

Certificate Path Building in PKIjs

Now that its possible to decode and verify the signature on X.509 certificates within the browser the natural question to ask is what can I do with that?

Well first off to build an interesting application you will need to have the ability to validate that a certificate is trusted the first step in doing that is building the certificate path associated with the certificate.

The defacto standard for path building libraries is the NIST PKITS tests our goal is to create a library that will be able to pass the sane tests from this suite (some are odd for sure).

This is a pretty high bar and will take some time. At the time of writing this blog post we pass 1-33 of this test suite in with flying colors these tests cover all of the basic certificate validation rules. We also think the library will pass all Policy Constraints and Name Constraints but more testing is needed to confirm.

So how does building a chain look like today with this library?

var certs = new Array();

// Load cert to be validated, its intermediates and root
for(var i = 0; i < cert_buffers.length; i++)
    var asn1 = org.pkijs.fromBER(cert_buffers[i]);
    certs.push(new org.pkijs.simpl.CERT({ schema: asn1.result }));

var crls = new Array();

// Load any CRLs we have
for(var i = 0; i < crl_buffers.length; i++)
    var asn1 = org.pkijs.fromBER(crl_buffers[i]);
    crls.push(new org.pkijs.simpl.CRL({ schema: asn1.result }));

var cert_chain_simpl = new org.pkijs.simpl.CERT_CHAIN({
    certs: certs,
    crls: crls

        alert("Good result");
        alert("Error: " + error);

The current incarnation of the API expects that the bag of certificates that is passed in will include all intermediates as well as all trust anchors. We will be changing this in a future release so that trust anchors are passed in another bag.

This will help ensure that the certificate inputs to be validated don’t contain anything that might accidentally result in the certificate being treated as valid when it should not be. With that said as it is currently structured we can begin developing automated testing which is great.

Note: Updated the post to indicate the goal is to pass the sane PKITS tests, some of which are not and some are not possible to pass in a web environment.