A neat data sharing technique

2 June 2017

We needed to share some data with another company, and this related to Credit card transactions. But we did not want to share the actual card numbers (PANs), what to do. What we came up with is quite neat, and can probably be used by others.

The external company collects the card numbers they want information on, they encrypt these with RSA with a key they generate and do not share. They send these encrypted numbers to us, we further encrypt them with our own RSA key and jumble the order of the entries, then send them back. So now they have a set of PANs double encrypted.

We perform an extract of the relevant transactions and encrypt the PANs with our RSA key, and send these as well. Now the recipients of these can encrypt these with their key and because RSA is a commutative function, can match up the two sets to see if the PANs they sent to us were used in the extracted transactions.

We have added a daily salt to these encryptions so that correlations can’t be used to work out which encrypted PANs map to the original PANs, and we bulk up the transactions so that individual transactions cannot be identified.

A friend of mine wrote up the proof of this:

Let Nv = our Key Modulus

Let E = our Operand (a number greater than 1, which is carefully chosen…)

So our Public Key, Vp = the couplet (Nv, E)

[Ignoring the Private Key as it’s not important…]


Let Ng = their Key modulus

F = their operand

So their public key Gp = couplet (Ng, F)

Let X = a PAN

Let Encrypt(K,M) be the RSA encryption algorithm of encrypting message M using key K

To encrypt the PAN using our Public Key: Cv = Encrypt( Vp, X)

This is actually Cv = X^E mod Nv

Then encrypt again using their Public Key: Cvg = Encrypt(Gp, Cv)

This is actually Cvg = Cv^F mod Ng

Similarly, encrypt the PAN using their Public Key: Cg = Encrypt (Gp, X)

This is actually Cg = X^F mod Ng

Then encrypt again using our Public Key: Cgv = Encrypt(Vp, Cg)

This is actually Cgv = Cg^E mod Nv

So Cvg = CvF mod Ng

= (X^E mod Nv)F mod Ng

= X^EF mod Nv mod Ng

= X^FE mod Ng mod Nv

= (X^F mod Ng)^E mod Ny

= Cg^E mod Ny

= Cgv

1. it doesn’t matter whether we encrypt with our key first or second, we get the same answer.

2. this means that the RSA algorithm is a commutative encryption algorithm

And so if we produce a value Cvg and they produce value Cgv and the values are the same, we can deduce that both organisations encrypted the same PAN, and nobody is actually sharing any PANs in the process.

End to End encryption under attack

30 March 2017
Amber Rudd

UK Home Secretary, Amber Rudd

During the consultations on the #SnoopersCharter or the Investigatory Powers Bill we were assured that there were no plans to break end to end encryption. And now with the most minor of incidents, of a single misguided individual, killing fewer that an average day of road traffic in the UK, that is being called a terrorist attack, we should give up all our privacy.

Thursday 30 March 2017

Dear Alok Sharma,

You wrote to me on 17 November 2015 (ref: CRM12097) in respect to my concerns over the Snoopers Charter aka Investigatory Powers Bill (now an Act).

In that letter you assured me that: “However the Government does not advocate or require the provision of a back-door or support arbitrarily weakening the security of internet applications and services in such a way. Such tools threaten the integrity of the internet itself.”


The comments by the Home Secretary, Amber Rudd, directly contradict that position. She is calling for messaging applications to be provisioned with back-door access.

I and other security professionals keep telling you it is not possible to safely provide back door access to encryption systems.


This extraordinary level of access must require extraordinary evidence that it is necessary. At the moment there is no evidence that access to this data would have any material effect on the outcome of the recent criminal attack in Westminster, nor any other situation.

Calling a misguided individual, a Terrorist only inflates the situation and causes fear. Lets keep things in perspective.

Yours sincerely,

Stuart Ward

Adventures in IPv6

18 February 2017

Because I use a home ISP that supports IPv6 and has done for quite a few years, I have been using IPv6 for some time. But recently a problem meant that I was losing IPv6 connectivity. IPv4 was working fine so only a minor hiccup. But the process of investigating this I learnt quite a bit about IPv6 and I thought I would document this here. It might help someone else.


IPv6 addresses are 128 bits long and are written down in a standard notation, this looks like:

It is rare to have a completely populated address, so the notation allows for shortening the bits of the address that are zero. So the address ::1 is all zero except for the last bit. Also the CDIR format of showing the number of significant bits is often used. When used in some commands the interface to send on is specified with a %eth0 suffix

There are several types of IP address, and they can be recognised by the most significant part of the address. These are the ones I came across.

::1/128 this is the loopback address same as
fe80::/64 anything staring with fe80 is a link local address. A bit like 10.x.x.x or 192.168.x.x and can only be used on a single link
ff0X::   These are multicast addresses. the most useful ones are
ff02::1 All nodes in the link-local
ff02::2 All routers in the link-local


IPv6 has been designed for auto-configuration, so an endpoint should not have to have anything set in order to use a network. Everything is automatic.

The link local address is automatically calculated from the MAC address of the interface, it should be there for any interface that is connected regardless of the network supporting IPv6 or not. You can display the IP addresses with ifconfig (or ipconfig on windows) or ip -6 address show

Neighbor discovery protocol allows the discovery the link local addresses of locally connected interfaces. We do this with a ping or as it is more formally known via Internet Control Message Protocol version 6 (ICMPv6) to a multicast address.

$ ping6 -c2 ff02::1%eth0

$ ping6 -c2 ff02::1%eth0
PING ff02::1%eth0(ff02::1%eth0) 56 data bytes
64 bytes from fe80::3e98:c0ee:51ae:b461%eth0: icmp_seq=1 ttl=64 time=0.072 ms
64 bytes from fe80::1e74:dff:fe2c:b897%eth0: icmp_seq=1 ttl=64 time=2.53 ms (DUP!)
64 bytes from fe80::3e98:c0ee:51ae:b461%eth0: icmp_seq=2 ttl=64 time=0.059 ms

--- ff02::1%eth0 ping statistics ---
2 packets transmitted, 2 received, +1 duplicates, 0% packet loss, time 1001ms
rtt min/avg/max/mdev = 0.059/0.887/2.532/1.163 ms

This should give you a response form the link local of that interface, and anything else that has an IPv6 interface on that network segment. But

$ ping6 -c2 ff02::2%eth0
PING ff02::2%eth0(ff02::2%eth0) 56 data bytes
64 bytes from fe80::1e74:dff:fe2c:b897%eth0: icmp_seq=1 ttl=64 time=2.58 ms
64 bytes from fe80::1e74:dff:fe2c:b897%eth0: icmp_seq=2 ttl=64 time=0.946 ms

--- ff02::2%eth0 ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 1001ms
rtt min/avg/max/mdev = 0.946/1.764/2.582/0.818 ms

will only give responses from routers on the segment. So now we have local addresses sorted, we need to get a routeable IPv6 address. There are a number of ways this can happen, most commonly this uses Stateless Address Autoconfiguration (SLAAC).

The steps in getting a Global scope IPv6 address then are first to find the router. This is either from the response to a Router Solicitation (RS) ICMPv6 message, or from just listening as a router will periodically send out a Router Advertisement (RA).

Wireshark Capture

Wireshark capture of IPv6 address auto-configuration

Lets step through the auto-configuration process. The first step is setting the Link Local address, this is configured from the MAC address, but there still could be conflicts. in packet 83 we send a neighbour solicitation out for the address we want to use. If nobody responds then we go ahead and use that address, packet 97.

The next important bit is the Router Solicitation, and Router Advertisement. Routers will send out a Router Advertisement periodically to the multicast address ff02::1 periodically, but can be prompted by Router Solicitation, packets 118 & 119. The Router Advertisement is displayed in the packet analysis window, and we can see the Prefix information as well as DNS servers from the router.

Next is packet 141 where we send out a Neighbor Solicitation for 2001:8b0:1679:ea38:cf58:def3:b993:1412 to check that nobody else is using this address. If nobody replies then we go and use this address.

Password Managers

7 February 2017

Lastpass Screenshot

I am constantly surprised that ordinary people don’t use password managers. I would expect most security professionals to use them, but even there I find many do not use a password manager.

So what is a Password manager? Basically a database that stores usernames and passwords for you. The data is encrypted with a master password so you do have to remember that one password. When you visit a site or start an application that needs a password the password manager fills in the credentials for you.

Why is this better than what I do at the moment? If you don’t use a password manager then you must be doing one of the following:

1. Use the same password, or a small set of passwords on many sites.

This is a bad idea, mostly because if one of those sites is compromised then you will need to change your password on all the sites you have used that password on. Can you remember all of those sites? How long will it take you to do that?

2. Write passwords down.

Actually this is not too bad, as long as you look after the password book. You can do some things to make sure that if the book is stolen then it doesn’t immediately compromise all your passwords. But if you lose that book how do you go about changing your passwords?

3. Use an algorithm to generate the password for each site.

It could be paper based or something you can remember and do in your head or a combination. Usually you use the domain name of the site to work out your password. The problem here is if that site is compromised then you have to change your password, lots of these and you have a long list of exceptions, or alternative methods for passwords. It will soon become unmanageable.

Software to the rescue.

So the answer is to use a password safe software. There are a number of systems available. I recommend both lastpass and keepass. Lastpass is internet based, and implemented through a browser plugin whereas keepass is an application you run locally on your machine.

Both allow you to store usernames, passwords and the URL of the login page. Both have a master password to encrypt the password store, and only decrypt the password in memory on the local machine.

Keepass has a local database, but this can be synced with other machines with a Dropbox, GDrive, OneDrive, or even sftp. Because the database is only decrypted in memory this is safe. Keepass is open source and there are clients for all desktop operating systems, and some mobile as well. There is a huge range of extensions to extend the basic functionality.

I personally use lastpass, but I also regularly export my keystore and import it into keepass so I have a backup.

So why is this good.

1. you have a different password on every site. OK if you don’t have this when you start you can progress towards this. Because you only have to remember one password, there is no effort in having a different password for every site.

2. You use long, randomly generated passwords. These systems will generate a new password for you, so you may as well make it long and complex as you don’t have to remember it. And that makes you much more secure. So when you set up a new account or change an existing password, generate it randomly and a make it 16 characters long (if the site accepts this).

3. Your password manager checks the domain you are visiting and will only enter the amazon password into the page at amazon.com not amason.com amaz0n.com amazom.com arnazon.com

4. Use your password manager as your bookmarks, if you need to visit your bank, select it in the password manager and it will go there and log you on.

5. Use you password manager to store password recovery information. Because you are using a password manager you don’t need to be able to recover a forgotten password, but some sites insist on this. Never answer the security questions with the correct information, if they want your mothers maiden name put something random in there, otherwise it may be possible to have your account taken over using the password recovery process.

6. You can use this to store and auto fill other sensitive information like Name, address, credit card numbers etc. This avoids storing cards on a website, from where it may be compromised. And because it is automated just as fast as having the website store the data.

So if you have read this far you should be totally convinced and ready to start using a password manager now. Well done.

System Rescue CD & Spinrite

7 January 2017

I am a great fan of System Rescue CD. It is a great distribution for fixing computers. It is basically a gentoo installation customised with a whole lot of tools for fixing systems. Disk partitioning, wiping, password resetting, etc. As well some tools that are separate bootable images like memtest and freedos. I also have and like Spinrite which is not a free tool, so can’t be included in the distribution, but I wanted to find a way of adding this to a USB drive so I had all these tools on the same USB stick.

First I used the included scripts to create the USB stick as a System Repair CD image.

I found that the text file in syslinux/syslinux.cfg that had the menu system in. Looking through this there is a section:

MENU TITLE A) Run system tools from floppy disk image...

I added the following to this menu item and copied my spinrite.img file to the bootdisk directory.

LABEL SpinRite
MENU LABEL SpinRite: Analyse and recover disk problems
kernel memdisk
append initrd=/bootdisk/spinrite.img floppy raw

Works a treat.

Old Computer technology forgotten

3 December 2016

Have we already forgotten how to use some of the computer technology from just a few years ago. Perhaps they need better (and older) tech consultants on programs. I was watching the latest episode of Timeless S01E08 Space Race. The plot takes us to the Apollo 11 moon landing and that a modern day virus is inserted into a 60’s era mainframe computer to disrupt the communications stranding Neil Armstrong and Buzz Aldrin on the moon.

A virus attacks flaws in the operating system or other privileged programs on a system. I suppose that we now have more knowledge of these problems, but how these would relate to operating systems of the 1960s is just not credible. So let us take the leap of faith and believe that a modern virus could affect computers of this era, and just look at the mistakes in dealing with this technology for a moment.

We enter into what looks like a IBM system/360 computer room. Firstly they insert a papertape program into the machine. This presumably is some form of bootstrap program, though that is not clear from the dialogue.

Loading the papertape machine

They have obviously just got a random bit of papertape and used that. In reality there would be a substantial leader of blank tape at the beginning. This would normally have either a typed label, or more likely, a handwritten description of the contents. The reader would normally have a sprocket wheel to thread the tape onto and a clip over tab to hold and guide the tape through.
I can remember it being cool to punch holes in the leader tape that made up dot-matrix like characters. This was a matter of setting up the bit image of the characters and adding this to the start of the punch sequence. That makes it sound much easer than it was.

example of fanfold papertape with printed label and punched characters

I mostly worked on Digital Equipment and Data General machines that used fanfold papertape, but there was also the teletype machines, that used loose tape. You would typically use a pair of plastic bins to hold the in and out portions of the tape. But this was quite untidy and the tape could easily be torn if it tangled up in the bins.

Next there is a sequence of loading a reel of ½” magnetic tape. In the background there are racks of tapes with the plastic surrounds (frequently called a “tape seal belt” because its purpose was to prevent humidity and dust on the media). There was a later innovation of these that usually matched with a automatic loading system. The tape, with the ring was loaded onto the drive, and the drive mechanism would open the seal, vacuum the tape through the machine onto the fixed takeup reel. The point is that you did not remove the seal belt before loading the tape. In the episode they remove the ring before taking the tape over to the machine. The tape drives do not appear to have the auto-loading mechanism, so they did this bit right.

Tape reels have an opaque side and a transparent side. The opaque side is the back of the reel and you would always load the tape with the transparent side outwards. As well there is a plastic ring that can be inserted into the back of the reel. This was a write protect ring. The drive would only be able to write to the tape if this ring was inserted. The ring was detected with a probe on the tape drive. so this ring must be on the back side of the reel when loaded into the drive. Also the tape on the reel was always loaded clockwise, inserting a reel back-to-front would not work at all.

Tape loaded back to front

With the write protect ring visable

They then press the top left button to supposedly load the tape. I suspect that this is the rewind button so it makes the drive spin as though it is doing something. I am surprised that they did not end up with lots of tape everywhere as the drive would try to unwind the rather than rewind. Perhaps they only had empty reels of tape, so they had to put them on backwards so you couldn’t see that there was no tape on them?

Financial Information

20 May 2016
A braclet that shocks you if you overspend

A bracelet that shocks you if you overspend

This story is typical of the difficulty people have in managing their finances well. The real problem is that banks and card companies have little interest in providing detailed analysis of financial information to their customers. As well as the divided nature of those institutions in that any one will only see a portion of the situation.

Firstly Banks, so my bank has all of my direct debits, and standing orders, and they know pretty well when and how much those are. They could project that forward and show me a projection of what my balance will be at the end of the month, next pay day or similar.

Card providers generally only show cleared transactions, sometimes there is a special display for pending transactions. This may be an important distinction for the Bank, but not to the consumer. As soon as I have made the transaction I want to see that in a statement.

What I want is at the point of making the decision, to have the information to make that decision. Can I afford to buy this thing this month or should I wait till next month. In the simplest case.

A more detailed analysis would be for the longer term. Like am I maintaining enough savings to cope with a not having a job for a couple of months, perhaps there are rumours of a takeover at my employment and I want to increase this target. Give me feedback on how many months before I can achieve that at my current spending / saving rate.

Personally I do this myself, and have done so for many years. I use GNUCash to record all my transactions. This allows simple reconciliation with statements. Then by ensuring that standing orders and direct debits are entered in a month before they are due I have a good projection of my future balance.