I was first introduced to Briar (Briarproject.org) at Pierogarnia, a nice bar with Polish food in Wedding, Berlin. Briar is basically a robust messaging app using device-to-device communication, avoiding a central server, and capable of being used as some sort of mesh network to communicate off-internet. While using the Internet, Briar can sync via the Tor network which defends users against traffic analysis, interception and surveillance.
(Arguably), the most lucid article that exists about Briar is by Torsten Grote, responsible for transportr, an android app which I have installed on F-Droid, alongside many others. Torsten has also influenced German policy since he is responsible for the addition of Free Software to the agenda of German's Pirate Party, after his memorandum about a free democratically controlled technological infrastructure was accepted.
Here is his article entitled "Briar – Next Step of The Crypto Messenger Evolution".
Before introducing Briar, he navigates from the old days when we were lucky if we had transport encryption on our stone-edge messengers such as, ICQ to more recent apps providing end-to-end encryption (obviously whatsapp's recent security update by Open wishpers comes to mind) and discusses different types of centralised and decentralised communication architecture on the Internet, including peer-to-peer (as in Briar). Among others, he discusses Open Wishpers's app called Signal, which I use on my phone (despite it drags your battery quite ferociously!)
One of the highlights of the article, and something that I have been thinking for a while, is that when discussing personal privacy, is no longer so much about protecting the interception of information in a one-off communication call but about protecting your metadata. Targeted single interceptions are not so valuable for traffic analysis although it may reveal some specifics. It does not provide the adversaries with a full picture of who you are, what are your patterns of behaviour and with whom do you relate to on your daily routines. It is the protection of your metadata what really matters. The adversary can make use of such valuable overlay of data emerging from your persona for any kind of purposes. A very annoying one is targeted advertising (this is at the core of social networks and browser searches) but of course, the aforementioned article goes a step further when pointing to the youtube video where a well-known organisation admit that they "kill people based on metadata". However, to contextualise the video comment more, it also makes the point on the fact that such powerful algorythms are clearly not applied to "domestic databases". I understand that the latter refers to metadata collected from normal citizens via social networks and google-like searches. That is good to know...! Also it suggests where are the legal limits and authorisations for the organisation in terms of what can be done with what is called "the Ocean of Data". So, normal citizens can rest asured.

Back in 2008 I wanted to investigate how metadata can also be powerful for sound retrieval. Therefore, I constructed a retrieval system based on soundfiles attached metadata which proved to be extremely powerful for the articulation of sound. I used such local retrieval metadata system in the context of an interacive game-audio work called "Ho" a sonic expedition to vietnam.

Here is a bit of an explanation of how the metadata is used to retrieve a collection of sounds from MaxMSP using punakea to tag clusters of files.

Retrieving Audio using custom metadata in maxmsp via Blender from Wunderbar Lab Berlin on Vimeo.

This was later called from a number of game-engine events (mostly collisions) taking place in Blender's game-engine (BGE) via OSC (Open Sound Control) protocol. Here is a 2009 test with the integration of both using Python in the BGE. https://vimeo.com/16473899

Now my final question: What if my soundfiles had become aware of the the fact that they are used, behave according to and become organised, due to their metadata? What if they could prevent me (the composer, the adversary) from adding such meta-information to they aural entities? Well, who knows, but I might have to get back to the era of crotchets and quavers to make music!

So, do you fancy building Briar from the source-code? If not, be patient until its public release!

Speech recognition has become a highly-popular child of computational linguistics. Its algorithms make use of acoustic and language models (e.g. Markov models, Neural Networks among others). Speech recognition current and portential applications are endless (mobile devices, in-car automated systems, in health / disability problem solving environments etc). However, it came across an interesting application (a proper attack) which called my attention. It can be used to tell us what an old dot-matrix printer machine is actually writing. This is basically what Backes, Durmuth, Gerling, Pinkal and Sporleder discuss on their article. “Acoustic Side-Channel Attacks on Printers”. http://bit.ly/1UzN5k3. The paper surprisingly demonstrates that there are still thousands of institutions (hospital, banks etc) still using this wonderful noisy printers, mostly because they last (new whole new printer's obsolescence model is arguably bordering environmental crime!) and because modern ones are not compatible with bank-like older computer clusters. Their method for acoustic analysis is based around the 24 needles of the printer head. The more the needles striking the paper at a given time, the louder the emitted sound becomes, providing some room por statistical comparison with prerecorded dot-matrix printer sound phrasing. They use a 3-step method starting with sound feature extraction (FFTs for frequencies within the human ear); followed by grammatical thinking (word-based rather than individual letters) by looking at decay time and the understanding of the blurring effect of the many needles, significantly different rom the analysis of computer keyboard strokes. Finally, they look at statistical frequency methods aforementioned based on Hidden Markov Models. Their trained models for the attack consist of recordings of dot-matrix printed text is used as the database for the recognition phase. They used MATLAB's Signal Processing Toolbox and I am not sure if I got lost here but I understand they must be similar to those used by a neural network machine learning algorithm. I have used the latter for my new musical composition called "B is for Bird", where I extract features from bird calls and songs which I map into an analogue synthesizer live (paper due on June 8th...). Another cool thing about the dot-matrix printer implementation is that they propose a few alternative countermeasures for the attack, including masking the environment with white noise... I would rather not work in that enviroment!
Well, a lot of people would think 'who the hell wants to do an attack o a dot-matrix printer in a hospital?' You will be surprise to know that, as mentioned in the article, several european countries require by law the use of dot-matrix printers for printing prescriptions of narcotic substances!

I came across this interesting paper (and video presentation which I can no longer find!) about decryption permission based on confirmation of a specific geographical location as a proof of identity, as valid as a public key or your biometric fingerprints. The authors are Nishanth Chandran, Vipul Goyal, Ryan Moriarty, Rafail Ostrovsky. The methods employed by Position-based Cryptography can also by applied to guarantee the geographical position of a message being sent (Position-based Authentication) and to allow access to specific resource only if the client is at a verified geographical position (Position-based Access Control). The main method to observe position in classical modes of communication was the Time of response between the "Prover" and the "Verifier". Phoenicians already used the Time method to calculate distance: Phoenicians ‘gauged the distance to a headland concealed by fog by making a loud noise such as ringing bells and listening for the echoes’ (Kahar, 1999). More modern methods use triangulation (e.g. in GPS positioning) and for fixing error of calculations they implement a factor to incorporate Einstein's theory about the fact that Space is curved. I was not aware of the latter but a pure mathematician friend told me about that when I saw him at one of theses Sonic Code Sessions (SCS) at SPEKTRUM, an art science community in between Neukölln and Kreuzberg in Berlin. The paper presentation discusses the Vanilla Model (some assumptions made) and the possibility of position-based key exchange which results impossible. Then it proposes an alternate model, the Bounded Retrieval Model (BRM),where position-based cryptography can be secured as well as key exchange and position cheating detected.

It seem like interest in cyber security is growing also in the city of Manchester in UK. According to a communications from senior colleagues at Manchester University, there have been conversations with the Chief Executive of NCC group, to explore the potential to lead Manchester University student graduates and researchers to this area!.

Here a presentation entitled: "Taking NCC Group into cyber arms race" by Rob Cotton, Chief Executive of NCC Group at Business North-West 2012

I ordered this book by Jessica Fridrich for the University library: "Steganography in Digital Media: Principles, Algorithms, and Applications". It has taken me two months to read it and luckily found time mostly on a continental flight! It focuses on visual steganography and steganalysis and gets into quite a lot of depth into digital image formats and theory. I quite liked the Summary section on each chapter as a recap of what was discussed. LSB embedding is treated as the easiest forms of hiding data in the bit sequence, something which I have experimented on Matlab before. Moving from naive methods of steganography, discusses steganographic security assuming certain constrains about the cover source and provides a definition for security based on the Kullback-Leibler divergence (some sort of probabilistic distribution gap) and also spread-spectrum methods. The following block discusses practical applications of how to diminish the divergences of those distributions found on the cover and the stego image and avoid detection. Then moves to steganalysis. The key thing I learnt from this book, (beyond the fact that I realised that I have forgotten almost everything on statistics, econometrics and Mathematics for Financial Operations -MOF- I studied at University of Valencia) is that this area has become a serious discipline of academic study and of vital important for security and distribution of digital media across the internet. The maths and statistical methods presented have nothing to do with alchemy and imagination to hide the secret. They involve pure scientific methods.

A friend let me Simon Singh's book entitled "The Code Book - The Secret History of codes and code-breaking". It is very easy to read and very entertaining. It encapsulates many of the key historical cyphering events which had called my attention when reading and learning about cryptology, reading papers and watching films. For instance, it navigates from Vigenère cipher to the intercepted letters which led Mary Queens of Scots to execution. Of course, it did not miss the cracking of the Enigma machine(s), and I learnt about PGP encryption which I had already installed on my Mail account. Finally it introduced quantum computing, which apparently may already be a reality somewhere.

Our research group was very honoured and delighted to accept and invitation to showcase NOVARS's Research as part of the Digital Industry Day, hosted by the University of Manchester (UoM) at The Landing, in MediaCityUK on Sept 30th, 2015. The main purpose of the event was to facilitate the dialogue, transfer of knowledge, research and ideas between industry leaders, academic researchers, start-up companies and think-tanks. I showcased a couple of research posters on game-audio and more recent ideas on audio cryptography and steganography in the Augmented City, recently supported by SALC.
The day included the participation of keynote speakers Neil McArthur – Group Innovation Director, Talk Talk, and Vikas Shah - Chair, Future Everything; CEO, Swisscot Group and a welcome speech by Prof Luke Georghiou, Vice President for Research and Innovation, The University of Manchester. Among others, it also featured talks related to funding opportunities, BBC R&D and UoM Research; e.g. Matt Brown - Innovate UK, Katie Gallagher - Manchester Digital, Laura Harrison & Robin Cramp - BBC Connected Studio, Lee Frater – Growth Hub and our colleague Abigail Gilmore, alongside John Knell - Knell Intelligence. The session was closed by Andrew James, assistance associate Dean for Business engagement and the visit of Prof. Keith Brown, Vice-President & Dean Faculty of Humanities to the final Network event.

Click on the poster to scale up

Today I have been on a MATLAB and Simulink 1-day workshop at the Renold building, Manchester, UK. It is entitled "Boosting your research impact" and to be honest, I am a bit fed up to see the word "impact" as I see it everywhere in academic environments. Please relax, this is just a workshop!. The session started with Data Analytics using Matlab and a few examples employing logic indexing. This part included analytical access to Big Data, calculating a mean in 4.5 seconds on a collection of merged excel marksheets with 1 million records. I see applications on this. The second part was on Image Processing, Computer Vision and Machine Learning with Matlab and this seems like a vast area of work. I learnt interesting things on image acquisition using colour thresholder and image analyser using optimisation processes. The section about object and tracking presented a few examples of sign post recognition and city objects with a camera in real time. This seems similar to AR technology and quite useful for the identification and data overlaying of urban assets. The final hour was about Simulink and stateflow for system modelling and Prototyping . Overall I got a broader idea of what Matlab and Simulink can do about what can be done beyond Signal Processing, which is what I am mostly interesting and I also learnt about Matlab network licenses and Toolboxes for Universities.

Returning to the idea of visualising a scene or city asset while a message is being decrypted, I created this test in blender. I can control the speed of grow and the number of letters which populate the scene etc.

For the last couple of weeks I've been looking at how to run my own server to host a website, remote access it and point a domain to the server IP address, open ports on a Superhub, and create a few services such as VPN ,website host, wiki and created private and public Certificates. I also learnt how to create a free SSL certificate Issued by StartCom Class 1 Primary Intermediate Server CA. I got my first one issued to http://server.procedural-audio.org. Still to solve a few mismatches but basically it all works. The algorithm signature is SHA-256 with RSA Encryption.
For the server computer I got a new macmini 2.6 GHz Intel Core i5 with 8 GB RAM 1600 MHz DDR3 and purchased OSX server software for Yosemite. For the hub side, I experimented with opening a number of ports for the different services, two for hosting (http, https) and thee for VPN. I also installed a couple of airport express devices (old and new version for communicating audio within the network (via airplay).

I also installed PGP on my own email account using PGPfreeware 7.0.3 and tested it with another email account. It was interesting to learnt about the Phil Zimmermann case and some legal aspects of PGP (it seems like outside US is fine).
Not sure how this relates to audio steganography but I though it would be interesting to learn about this and use the server for both hosting material and remote control but also also for creative purposes so that I do not have to rely on an external server or I can run my own parallel backups.

My neurons are now starting to make connections between my previous collaborative work on augmented aurality (as seen in the four editions of locativeaudio.org), augmented reality, steganography and cryptography.
In a nutshell, augmented aurality tools and locative audio, provide sound artists, composers and aggregators with the ability to 'implant' sound (and media) in site-specific city environments in several forms. With the use of a geolocative-aware device such as, a smartphone or tablet and a pair of headphones, individuals can 'explore' a real space which has been "augmented" with a virtual overlay. The overlay usually consists of sounds, aural instructions for navigation purposes but can also be textual or visual (augmented aurality). Composers often collaborate with historians, archaeologists, writers, etc., to build contents informed by the history of the targeted space (past, present or even future) providing a new reality and understanding of it.
My prior experience in facilitating projects to "augment the city" has progressively become more and more complex, pushing the boundaries of classic geolocated soundwalks and use the city as an extension of the concert hall. I have also pushed the idea to create multi-city locative audio walks which can be experienced and remixed from a single gallery space with the push of a button. When I wrote '[5]', I created a game-engine environment where real walkers' geolocation coordinates were passed to the 3D space.
But how to implement Steganography or cryptography in the context of the augmented city?
A typical approach would be to implement stenography is some sort of geocatching project, to add a few hurdles before the final "treasure" is found. Far from my thinking.
An alternative approach would be to hide some information to solve some sort of game puzzle deciphering or unveiling hidden information. Away from my thoughts.
I am thinking of something much simpler; to leave thoughts in carefully selected places which can only be retrieved by those we want. Of course I need to unfold an strategy and different case-scenarios for this idea, clearly inspired by the work of urban anthropologists and the citizen urge. So, yes. The city can afford a virtual overlay to become your new secret diary and you might be able to give the key of the lock to those you desire, even if they still have not been born or the place no longer exits in the future.
Pictured above: hidden speaker in a flower by Sam Salem

From scary faces and ghostly hands to cats, self-portraits and female lips. Musical artists have been hiding secret images in their music for quite a while. One just need a spectral analyser to "see" hidden figures and bitmaps in "[Equation]" and "Windowlicker" by Aphex Twin, and in works by Disasterpeace (FEZ Soundtrack), sippenaken, Venetian Snares or DJ Sonix.
When I first came across the idea, I did not find it that interesting because of the sound quality generated from a low resolution 120x120 bitmap was timbrally quite poor. But then I thought that perhaps there could be a way to improve it, if I were to understand the process a bit more (from the code itself).
After looking at some Matlab code by a couple of students in Mathematical Methods for Measurement Technology, at Tampere University of Technology, Finland (Katarzyna Zarnowiec and Frederico Contente), I looked at their source, which is a short programme written in Pearl by by Evan Salazar. Evan is a Computer Science graduate from New Mexico State University. The program basically uses a spectrogram tool to reveal embeded images. To encode the images, Evan devised a sinewave frequency oscillator for the Y Axis of the sonogram, Time parameter for the X and amplitude levels to control the pixel colour intensity. Then, he created a small audio library to write an audio file with enough sound quality (44.1 kHz, 16 bit, mono) onto disk, which of course is already defined in Matlab [y,Fs]. Then he found a piece of code to read a Bitmap, in this case a spatially mapped array of bits, which could indicate the colour depth (1,2, 8,...64 bits per pixel), which is somehow similar to audio bit depth (per sample, not pixel) and that was it. Some tweaking had to be done to avoid clipping and to match amplitude scaling to match the colour intensity (adding all colour channels to a given pixel and scaled it to the maximum amplitude). In the example video below, I have a simple demo of sound created from a couple of pictures; a frog and then the 'das boot" DVD cover (squared). The latter seems to have more potential for revealing something with musical content so, I tried EQing and trimming audio sections but also a number of options in Matlab to increase the sample rate and the number of pixels per second being transformed.
Sonic results are somehow promising and quite honest as methods for the investigation but sonically they may not go on their own in a final composition. They could be used to reveal images as well (in the interactive media context), since it seems like the EQ does not affect much the quality of the image; the trim does though.

I came across these two papers (*) on Acoustic cryptanalysis by Daniel Genkin, Adi Shamir and Eran Tromer about careful listening (broad frequency-range) to Computer noises and power drops which may embed sonic patterns revealing encrypted keys.
This research from Technion and Tel Aviv University won the Black Hat 2014 Pwnie Award for Most Innovative Research. The first paper describes how attacks can be perpetrated on RSA implementations (the second expands to ElGamal implementations) by listening to high-pitched noise to identify unique sound patterns found in RSA keys. It focuses on how these attacks can extract full 4096-bit RSA decryption keys from a laptop.
Listening to laptop's acoustic leakage or printed code to find "software bugs" is not new. Our ear is so subtle that it may be often faster or more efficient to scan a piece of C++ code including thousand of lines through "hearing" than to read it entirely. Prior to that the mapping of some sort of code to harmonic frequency is needed though so that the ear could detect the odds.
This is providing me ideas for extraction of usable sounds, not for decrypting purposes but for the context of the piece and investigation. I remember a performance at N.K. Berlin years ago, where a female artist was already live amplifying noisy hard drives with contact mics forcing very old laptop computers to shutdown, start etc. It was very well done and also sonically quite engaging to magnify those innerworlds in slow cranky storage units.
Too bad N.K. project just closed this summer. I wonder if this had something to do with their neighbours' complains (concerts were arguably too loud). Sound can be the key.

(*) Papers::
- Daniel Genkin, Adi Shamir and Eran Tromer, "RSA Key Extraction via Low-Bandwidth Acoustic Cryptanalysis". Online publication at Technion and Tel Aviv University. Here.
- Daniel Genkin, Lev Pachmanov, Itamar Pipman and Eran Tromer, 2015. "Stealing Keys from PCs using a Radio: Cheap Electromagnetic Attacks on Windowed Exponentiation". Here.

In 1917, British cryptographers Nigel Grey and William Montgomery decrypted the famous Arthur Zimmermann telegram, which would invite the Mexican Government to join the Germans, should the United States had entered World War I. The deal would provide Mexico with three former States (in US soil) in return.
The telegram's route was:
[US Embassy, Berlin] >> [Diplomatic cable to London (via Copenhagen) >> [Transatlantic cable to Washington - via Porthcurno in England] >> [To German ambassador in Mexico, Heinrich von Eckardt] >> [Mexican President Venustiano Carranza]
The British wanted to use the content of the telegraph to force the US to enter the WW I but Telegram was intercepted between Copenhagen and London eavesdropping a US cable at "Room 40". Therefore they could not tell without revealing the US Government that their cable communication were eavesdropped (so 'Cablegates' were older than we think!). But also without revealing the Germans that the Brits had successfully deciphered their codes (as a result of the Russians finding in 3 code books and encryption keys at the wrecked cruiser SMS Magdeburg).
Project Strategy: There are two aspects here that can creatively being implemented in the context of an interactive cryptographic game-engine work. One is the location of the real interception, which could be compromised (so that there are different intercepton zones or alternative channels of communication, some of which could be compromised) and the second is how to deal with the content of the information, once decrypted (reveal or not reveal?)
The British solved the former dilemma via "Mr. H", some sort of James Bond in Mexico (in reality Thomas Hohler), who provided a local copy of the telegram, so that it could be justified that the interception happened at destination and that the information was just plain text, not coded.

Pictured above: Zimmermann Telegram, 1917 From: Decimal File, 1910-1929, 862.20212/82A (1910-1929);

Chirp software and its concept behind (discussed previously), clearly demonstrate that short-distance Digital Communication between devices (e.g. smartphones) outside the Internet is possible and accessible. Pulling this thread a bit more, I came to know Locally-encrypted Mesh Networks, which have been out there for a while. I started to think about how an interactive media project using audio steganography could exist without the need for participants to be plugged to the internet, and what pros and cons it would have in terms of encryption of information and speed and reliability of the transmission. Needless to say that the idea of de-centralised communication holds some degree of inherited aesthetics in itself. In the field of aural encryption, geo-located small modular nets of nodes can provide some extra layer of complexity to the code-breaker. The need to locate near-by nodes for local transmission can become an additional playful aspect to the project and to visualise the network can also be fun. Mesh Networks could potentially bring this interactive project to remote areas and villages, where there is no internet access.
Encrypted Mesh Networks are often discussed in the context of catastrophic events (e.g. earthquakes), when every electronic means of communication, including the internet, are down. With open source software such as Serval for Android, smartphones become small radio transmitters (as nodes), which can pass on an encrypted message to the next node. They act as wireless routers using wifi 802.11 standards, but locally-configured routers can strengthen the signal too. Advantages are many; nodes are dynamically self-organised, self-routed and self-configured, looking for the most efficient path to pass the message. They are also self-healing when one of the nodes switches off. The more the number of nodes, the faster the communication. Most importantly, carried messages are encrypted so, they may be intercepted but still would need to be deciphered. Projects somehow inspired by the non-centralised system currency Bitcoin such as Serval, Commotion, projectmeshnet, Darknet, okturtles, twister and bitmessages are already rowing into this direction. As expected, out-of-range "handshakes", can't easily get away from the strong influence and control of smartphone manufacturers and wireless carriers over Governments, which at the same time see their security efforts potentially being breached. Beyond the creation of new lawsuites and the use of wifi patents to enforce some degree of control over Mesh Networks, it seems more reasonable that the code-breaker will test the security of the Mesh by infecting it with deciphering nodes (up to 8 miles away can be intercepted), this Man-in-the-Middle Attack could redirect the off-internet traffic to the internet. Once on the internet, all safe.

Audio Steganography is used to embed (hide) information in an apparently normal digital audio file. After testing a couple of audio steganography methods such as Phase Coding and Spectrum Spreading (including matlab code with working examples), I came across an interesting online reference discussing the typology of Audio Steganography is some detail. Basically these are the systems being discussed, which provide a very clear typomorphology for Audio Steganography methods:
- LSB Coding:
      - Simple method using the last significant bit only
      - LSB coding using the random interval method
- Parity Coding
- Phase Coding
- Spread Spectrum:
      - The direct-sequence
      - The frequency-hopping schemes and
- Echo Hiding
I trust this is a massive area of work which I will be investigating in depth in the months to come. Not only in isolation but also combined with other cryptographic techniques.
Many audio principles needed for working and understanding audio steganography are in the Curtis Roads Computer Music tutorials! So, I do not feel lost in areas of digitising sound (PCM-Sampling, Quantising and Coding).

• After figuring out some Python and Blender tricks, I have created a live input system which incorporates some random (as real) sound design for typing keys. The .blend file has four python scripts which talk to each other via passing property values through messages. The Python Scripts are:
• - The live Input python code: it reads 26 letters in real-time with live sound from an old Mercedes typewriter, (which I got form Neukölln). It reads a global dictionary (key, value) fetching the ascii values first from a blender all-Key sensor and then it matches each ascii value with its alphabetic pair in the global dictionary. The 26-key alphabetic string is stored in a local owned property, which is then passed onto the Brutal-forced script (the cipher-decipher method) via a copying the string on Property actuator (which receives the string via a unique message). The Brutal-force script makes the string uppercase and it stores in a local property which now matches the 'message', (which before this implementation it was given beforehand.) This means that I can now get a ciphered message from anyone, enter it live in the BGE and then brutal-force its meaning. Or I can simply use this for entering the 'plaintext' and then send one of the key-encripted strings to someone and then the key separately. Of course, the algorithm is the easiest to break but this is just about creating the live input for any algorithm in the future.
  - The BruteForce script: provides all the 26 different possibilites to cipher the string, including the plain text. A Blender adaptation from Al Sweigart's book.
  - A sound script using OpenAL in Python which reads up to 54 sounds in random with slight variations in pitch and dynamics. The computer keyboard sounds like the Mercedes typewriter. Sure you thought mercedes sounded somehow different...
  - A wheel rotator script to move the inner wheel to any of the 26 positions. It is too fast to see it for each key. So, I will be looking into it so see how can I do some 'Max Payne' tricks here... or bullet-time effects (as described in the game engine jargon)
  
• Get the Blender Caesar Cipher Live Input example here.
  
• 
  
• I bet you thought German Mercedes sounded somehow different!
  
• 
  

• Here is another Python implementation in the Blender Game Engine. The weather is quite awful so, I keep doing some work. This is a primitive rotary code using the Caesar cipher which can cipher the alphabet using keys from 0 to 25. To avoid too long 3D modelling I just created a version with 9 only keys. Albert Myer used this system in the American Civil War back in 1865! The code and module are a Blender -python adatation from Al Sweigart's excellent book entitled "Hacking Secret Ciphers with Python". This is potentially a python-bible for this research.!
  The idea of going back to simple forms of cryptography with strong visual feedback was suggested my composer colleague Iain. It seems important to communcate to the audience what is going on. I have only illustrated the encoding key methodology but there will be more to come. The final blender file includes sound design for the key rotator, recorded from an old German hydraulic taximeter.... yes Mauer Park chunk! BTW, this flee market was closed last weekend and I wonder why. This is very odd!.
  
• Get the full Blender Caesar Cipher example here.
  

  http://inventwithpython.com/hacking

• Dwayne warns about regulations to export cryptography software so, I need to check that out (although this is not what I am doing).
  
• 
  

• In this post I am demonstrating the use of the DES Encryption Algorithm (56-bit key, not terribly safe), to cipher and decipher a plaintext in real-time within the Blender Game Engine (via a python script and 3 assets). It was fun. To do so, I reinstalled Anaconda on OS X so that I could run Python 3.4 as the primary Python installation (before was 2.7), so that it will match the Python version in Blender 2.73. Then, I installed PyCrypto, a wonderful Python Cryptography Toolkit by Dwayne Litzenberger which includes modules written by himself and others.
  
• The game engine test in the video below uses a variation of Dwayne's example algorithm (DES) and his plain text, which I extended to 40 char (to be multiple of 8). I created 3 Assets in Blender. One is the Courier (the FPS red chap) who carries the plaintext and who needs to encrypt it before it is intercepted by the code-breaker (a blue chap called Rejewski - let us leave Eve, Alice and Bob alone for once!-). The Courier also needs to safely deliver the message to the 3rd character (the Queen) who will decipher it back to the source. Rejewski, the codebreaker, has a seek AI algorithm so, follows the Courier everywhere. When the codebreaker is less than 2 Blender Units away from the main character, a 10 second count down starts, which is the fictional time it would take to steel the non-ciphered message (a robbery!). To avoid so, the Courier must encode it by carrying the message to the bombe-like rotator and this will DES-cipher the message with the Python code and de-activate the countdown and the code-breaker future attacks. After chipering, the message will appear as encrypted in the form of a bytesArray ( '\n' reformated).
  • When the chiphered message is received by the desired recipient (distance with the Courier become less than 2 Blender Units), the python code executes the deciphering, allowing the Queen to read the plaintext. Yes, simple idea but it is my first proof of concept and it seems to work. I have started other tests and related ideas (Input Text etc)
• Note: Python itself within Blender can't find the PyCrypto libraries because it is a stand alone version of Python therefore, in the python script I had to point to the library in Anaconda via the path sys.path.append("/..."). Get the full Blender DES example here.
  
• 
  
• 
  

• The entry today is rather technical!...Jumping randomly from method to method of audio scrambling when reading papers and testing Matlab code is like opening a can of warms. I was building my own typography of methods and Residual Intelligibility (well the opposite), until I found a paper by A.Srinivasan, P.Arul Selvan(*) with an extremely detailed taxonomy of analogue (mostly) and digital methods, which I have resumed in a mind map for hanging on the wall. If Residual Intelligibility is low, we are dealing with a strong algorithm, of course but for the purpose of this investigation and its creative outcomes, any method is potentially welcome, including analogue ones. In fact, when encoding delay is too high, the algorithm may have potentially less applications in real-time. Both two-dimensional techniques and transform domain techniques have low Residual Intelligibility overall and therefore, they highly secure.
  See a MIND-MAP version of Residual Intelligibility in Audio scrambling HERE
• 
  (*) A.Srinivasan, P.Arul Selvan: “A Review of Analog Audio Scrambling Methods for Residual Intelligibility”. Innovative Systems Design and Engineering www.iiste.org ISSN 2222-1727 ISSN 2222-2871 (Online) Vol 3, No 7, 2012
  
  

• Williams Poole refers to the trial of Mary Stuart, (the Queen of Scotts) in his article about Jacob Cole(*). Mary was sentenced to death after being caught in a letter-trap set by spymaster Sir Francis Walsingham, principal secretary to Queen Elizabeth I of England. Poole says "Walsingham broke the code between Anthony Babington (1561-86) and Mary, then forged an encoded addition to a letter, causing Mary to reveal the identities of the Babington Plot conspirators". Mary Stuart thought her ciphered letter system was secure. This article also reveals what most frequent Cryptanalysis techniques for text messages were employed in the late XVIth Century. For instance:
  - Frequency Analysis: via matching up most commonly occurring cipher-text symbols with plain-text (English) letters. A more sophisticated technique was Viète's analysis of the relative frequencies of triads of ciphered symbols.
  - Polyalphabeticity: using orbicular techniques; e.g. using rotating rings and code-wheels.
  - Looking at consonant strings proper to given languages (aside English), use of palindromes, etc.
  - Switching the ciphered alphabet for the first of last few words of each line.
  An of course, it was strictly necessary to have a good knowledge of current affairs.
  Lesson learnt today: Deciphering put an end to Catholic Restoration in England!
  (*) Poole, William. 2011. A late sicteenth-Century Cryptographic al tratise: JACOBUS COLIUS'S 'TRACTATUS DE FICTIS CHARACTERIBUS' (1584–86). Journal of the Warburg and Courtauld Institutes, Vol. 74 (2011), pp. 213-239.
  
  

I know... more Mauer park chunk... but I have it for years now. I bought it when people could still find something interesting in that flee market, regardless of the price... ! This GDR/DDR line phone had a sticker on in which reads "do not discuss classified information".. This telephone is subject of monitoring at all times....
Are communications secure...? Were radio-telephone communications secure during WW II? I had a look at Bell Labs 'A-3' Speech scramble, which used Voice Inversion scrambling, a not too secure method which preserves the rhythm of the conversation. I trust it uses a carrier frequency to modulate the original frequency and applies a low pass filter, producing low frequencies to be heard as high frequencies and viceversa. A similar process happens if we replace the notes of a well known melody using up and down octave changes and therefore preserving the original pitch class (not the octave). It gets quite hard to recognise the tune for the human brain. A more advanced version of the A-3 was using Split Band Voice Inversion (VSB), using a 5-band system, and according to Christos's detailed article, only 6 combinations out of 3.840 were trully unintelligible (bands could be up-shifted, in the original position or inverted). It did not take the enemy long to de-scramble the A-3 system.
I also learnt that although I trully admire the work undertaken at Bell Labs, they never had a female president!

I have this door (top left) at the entrance of the house for 9 years. It wasn't charming but that was my door. Most neighbours in the building seem to have a similar one. In an old DIY 1960er book from the GDR found at the Mauer park said that people used to cover old doors with a flat panel for the hall to look more modern. With the use of a screwdriver I promptly realised that there was an old door (top right, probably from the 1920s) underneath my door, which now I am restoring. This 'shared secret' was revealed to me by pure chance though but what would be the best method for sharing a secret?; what about the optimum way for distributing a secret amongst participants allocated with partial information?. What alternative combinations one may choose to reveal a secret or secrets?; how can one combine them and how do they 'exist' (eg. as sounds) if they are on their own, or if they couple with the wrong option? Ultimately, why we humans share secrets while most species arguably don't?
message = 'daed era meht fo owt fi ,terces a peek nac eerhT.'

I have spent quite a few days fully focused on audio scrambling. I thought of how would it be to go to a restaurant, order scrambled eggs and then try to reverse two minutes of frantic egg whipping until I could get the original yolk and white intact. That may be hard! This is the 'one-way function' that Diffie and Hellman were looking for to solve the key distribution problem. Based on 'one-way functions', I created a new section in the bibliography for speech scrambling and speech encryption, which are the forms of audio scrambling that I am more likely to use for the interactive piece. I started by reading some many contemporary techniques for audio scrambling / de-scrambling and tested some of the ideas using Matlab such as, Dan Ellis's| 'Time-domain scrambling of audio signals' or Brhanemedhn Tegegne's Channel Coding using hamming codes. Both are excellent tools but as I was trying to understand the detail, I though I was lacking of a full perspective about where speech scramble initiated and then I began to rewind the over-exciting tape of speech scramble history. I mostly went back to speech security systems implemented in the Second World War or right after. Starting with the A-3 voice scrambler developed at Bell Labs by AT&T, which was successfully deciphered by the Germans (more details to come) and then, the SIGSALY system using random noise for scrambling. I also investigated Alan Turing's Speech System 'Delilah', a research which started towards the end of the war. I did not want to jump to Turing's work at Bletchley Park, which now is widely known, because I will reference it in the near future. I saw the enigma movie this week too... Reading Alan Turing's from 6 June 1944 brought me memories from Queen's University Music Department, where I learnt how to operate the Alan Turing machine in Dr Alan Marsden's class. Alan (Marsden) was teaching psychoacoustics for PGTs. That was back in 1997 when we were using a turing machine online simulator run on NeXT computers in the music lab in the basement. Although I have not been able to find the exact online simulator we used at the time, there are quite a few good others now. Why music labs were always built in obscure basements??? I don't understand. Well, let us focus on 'trapdoors' for now.

The RSA algorithm is made out of the initial letters of the surnames of its authors Ron Rivest, Adi Shamir and Leonard Adleman back in 1977. It was a an asymmetric public-private key cryptosystem break-through, which published a public key based on the two large prime numbers which must be kept secret. The matlab code below is a modular set of functions, where the user is prompted to introduce the value of 'p' & 'q' (two prime numbers) to generate the value of the Public & Private keys and then is asked to embed a secret message. This code here is by Shaun Gomez first published at http://bit.ly/1Mcht1O (digital iVision Labs) and I have redistributed retaining the requirements of the copyright license.
Note: Do not confuse the RSA algorythm with Santa Rosa Airport, Argentina, (IATA code RSA)

Tested this Matlab code Algorithm to encrypt an Audio file posted on an blog entry by Farheen Bibi (2013) for Crunch modo, which I slightly adapted to work for Matlab 2014b. It provides user input at ciphering and de-ciphering staged stage with the possibility to include a degree of audio 'distortion' from the original from 1 to 6.
Matlab code I used (slightly adapted from the original to run in Matlab 2014b) is here.
To run it including the audio samples, try this one here.
Screencast below: Note that first I use a standard noise audio file to ciphering the original sound file "years ago...". Then I try with a sound I recorded using a 1963 VCS3 synthesizer (which belong to Lancaster Music department, UK). The original file is successfully cyphered in both cases but sonically speaking the latter example is more useful in creative context (as it will be in the locativemedia experience I am planning as a result of this investigation).

Tested how phase cancellation using the two matlab files (left1.wav and righ1.wav) behave in Blender in real time as compared to using two earplugs on one ear (one fixed and the other moving in parallel facing the ear). Try blender example here

Reading Desmedt's paper on Audio and Optical Cryptography (1998) has been extremely revealing. He and his colleagues Shuang Hou, and Jean-Jacques Quisquater proposed two cryptographic schemes using high quality audio and optical cryptography which do not require computation to decrypt the ciphered text, although computers are required for the encryption. It starts with a demonstration of two aural decryption methods used using simple wave forms (referred as harmonic sound), one based on Destructive Interference of soundwaves resulting in phase cancellation (a concept we composers are very familiar with). The second is based on sound localisation which can produce a similar phenomena, I presume once the wave is perfectly out of phase by 180 degrees via source displacement (well moving the speaker the 1/2 distance of the wave-length, which depends on the frequency of the tone and the speed of sound). The plaintext message is introduced alongside the harmonic sound as a binary string. I tried a similar experiment with the following matlab code from Murtaza, RMIT University which I have modified (sinewave seen above on the right) and the cancellation is clear if we listen to it by using left and right earplugs nearby the same ear and move one of them slightly (the low frequencies disappear). In the most advanced case, they introduce a musical example. The paper examples are also tested using high quality audio (music). I was able to locate the broken audio links in the paper but the audio files are no longer there.

Finally, the paper includes a parallel scheme using optical cryptography, which employs a Mach-Zehnder interferometer, a laser beam and some lenses for decryption without any computational aid for decryption. What the paper really questions and reveals is the possibility to go 'analogue' (to rely on physical or human phenomena) in deciphering text in a era of digital computation. Sound certainly can prove that the answer is yes.

I've been looking for a new domain to host this research diary. I was searching for domains around the word 'cryptography' but believe me, it is a very popular term for an internet domain. All taken. Even remote derivates such as, 'steganography' were are also taken, not to mention 'cipher' et al. Since this research is primarily around the concept of audio and cryptography, I checked 'cryptography' with the new 2015 domain extension .audio and it was available. From now on, this research diary may stay as http://cryptography.audio which is think it is pretty accurate. Since this research also embarks on a cross of disciplines, I will also keep it under cross-disciplinary.com as before.

French Mathematician François Viète (1540-1603) explored in depth the connections between Cryptanalysis and Algebra, advocating for the 'infallible rule' for decryption of ciphered information. Earlier than him, Machiavelli stated in 1521 that couriers bearing a ciphered dispatch should not know the nature of its content, to avoid codebreaking of the message by the enemies. Europe was setting the grounds to enter into an era of Deciphering, Occultism and industrialised Cryptanalysis. It was the era of "Cabinet Noirs" in France (black chambers), which intercepted letters from suspicious senders without breaking the flow of the postal services (this rings a bell with recent phone-hacking scandals by news corporations and governments).
Although this context is trully interesting, my aim behind these cryptographic readings is to find connections between cryptology, music and creative practice and therefore, I am obliged to refer to Gottfried Wilhelm Leibniz (coinventor of the differential calculus) and his understanding of cryptanalysis. Leibniz said(1) that cryptoanalysis shows the way towards a new art of making hypothesis, the art of inferring (Ars Conjectandi), because it operates "pure and abstracted from the subject matter". This also clearly resonates with the 1950's Schaefferian acousmatic principle of reduced listening (écoute réduite) in his Traité des objets musicaux and Solfège de l’objet sonore.
(1) Pesic, Peter. 1997. Secrets, Symbols, and Systems: Parallels between Cryptanalysis and Algebra, 1580-1700. Source: Isis, Vol. 88, No. 4 (Dec., 1997), pp. 674-692. Published by: The University of Chicago Press on behalf of The History of Science Society.

Maya written language is based on syllabic glyphs / hieroglyphics similar to Japanese kanji (chinese) and kana writing or even Egyptian hieroglyphs. I made two tests in Blender using Mayan glyphs and Chinese characters from an old Confucian manuscript I found in S. Korea years ago.

Last 1st of July 2015 I put a Research Support Fund Proposal to SALC, School of Arts Histories and Cultures at University of Manchester in support for this project on Aural Cryptography; encrypting the augmented city. It is a small amount (£992) but it will help me to get the ball rolling via research assistance in areas that go beyond aural encryption. For instance, to accelerate research on visual encryption and to port it to Unreal/Unity by improving the Physics Engine Implementation and by dealing with the deployment to Android - / iOS of the interactive encrypted system. Fingers crossed!

Update 3 Sept 2015: The application was successful!

After presenting Putney for game-audio and a VCS3 at xCOAx 2015 in Glasgow (performing with Mark Pilkington and sharing a VCS3 ARVR installation with Architect Alena Mesarosova), I visited Dr Stefan Bilbao in Edinburgh in my way to examining a PhD in Belfast. I discussed some ideas with Stefan about how to scramble audio and reverse it in full based on my not-so-good tests using fft~/ifft in Max. He wrote some quick code in Matlab which I have slightly modified (open it here, including the plot). This code basically adds some random garbage to the original soundfile read as 'y' [y SR] = audioread('yearsago.wav'), were SR is the sampling rate and 'y' is represented as values plotted between -1.0 and 1.0 across a wavetable. The random array r = rand(ys,1).*sin(2*pi*4000*[1:ys]'/SR) creates a new file 'y1' when it is multiplied element by element with the original soundfile plotted values, which is by the way normalised to 1 (max) and positive only (abs). Then, a new second audiofile is created 'y2' as a result of substracting 'y1' to the sound source 'y' and therefore it is easy to recover 'y' by adding 'y1' and 'y2'. This is a non-realtime process but a good starting point for me to start checking values and hearing sound qualities of the transforms.
Here is the aural result:
Below the original soundfile [y, sample rate 44.1kHz]

Below [y1, sample rate 44.1kHz]


Below [y2, sample rate 44.1kHz]


Below [y3, sample rate 44.1kHz]
The reversed process (y1+y2)

I got MatlabR2014b installed on my laptop via the University of Manchester network license (so need VPN to run it when I am away). Thanks to Andy for the install.

Thanks to Andy Davison I learnt about Chirp. An excellent idea to share files, links and data among smartphones using chirpy /blippy sound phrases. The concept is very intelligent and they seem to have come with a golden idea after crowdfunding 3/4 of a million pound. I have tested the app and signed for the SDK. I can see immediate applications of this technology alongside Aural scrambling.
Don't miss it. Get chip from here and try Chirp composer to decode images into sound for transferring to people near you with their smartphone mics on!

My first attempt to transform a voice sample which could not be recognised while its signal process could be reversed, was by using the Fast Fourier Transform (FFT) in MaxMSP. This algorithm basically converts time domain into frequency domain applying the FFT onto a mono audiofile called snowmono1.wav and it creates two outputs: real and imaginary which process can be reversed. The imaginary output is clearly not identifiable from the source (fftsnow.wav included in case you can't run it) but it is sonically unusable (aurally not engaging enough for compositional purposes). However, the scrambled sound preserves the rhythmic / pulse aspects of the original soundfile.
The reversed process to recover the original file uses the object ifft~ which is the inverse of the Fast Fourier transform, to undo the transformation.
This works but I can't create a parallelism with the overlay technique as learnt in the visual encrypting post before.
Download the MaxMSP fft~/ifft example from here. You need Max runtime from cycling74

I have investigated Intrasonics technology for embedding inaudible codes in audio and recovering it acoustically. The commercial outcome in iself is fun but not interesting for this research. It uses an app to virtually return a tennis serve from world’s fastest server, Sam Groth (263kph). It is interesting how audio can be used to trigger codes for precise time-syncronisation and how inaudible watermarks are becoming a key technology for tracing music online (for whichever purposes!)

I did some duckduck online searches on the topics above and compiled a few interesting and publicly available articles and publications which can be found here (as a reference list). I will be updating the list as the investigation progresses.
UK in particular has been supporting cybersecurity research underpinning for two decades and it is clear a matter of national priority. Here a list of UK Future priorities (source image above):
• Cyberwar: countering the financial and social damage.
• Global threats, cyber war, ethics, regulation, policy and legality: understanding the complexity and countering the threats.
• Human factors and usable security: understanding human behaviour as a route to improving the security of systems.
• Risk identification, reduction, mitigation and management: looking at emerging uses of the Internet and the risks associated with them.
• Secure management and use of data: looking at better ways of storing and sharing data as well as considering ethical and legal issues.
• Making systems more resilient: investigating ways to protect infrastructure against malicious attacks.
• Understanding and monitoring systems and networks: understanding system behaviour so abnormal activity can be identified

I was very sorry to miss Victor Madeira's recent talk at Bletchley Park, Milton Keynes, where he talked about his book Britannia and the Bear - The Anglo- Russian Intelligence Wars, 1917-1929 and cryptology.

I have also investigated simple forms of visual cryptography using a python script by Robert Donovan from the LessonStudio. This classic VC script takes one image file and splits it into two encrypted images. The original image is only revealed when the two encrypted images are overlaid (e.g. using two acetate sheets). To test this idea in a game engine, I have used a couple of planes assets in Blender Game Engine, mapped the two encrypted images and applied transparency to them. I discovered that when they overlap it seems to work and reveal the phrase. I used Winston Churchill's famous quote recently posted, "It is a riddle, wrapped in a mystery, inside an enigma", which can hardly be read above.
Download Donovan's Python script from here
To run the example on OSX I have installed Anaconda and I am running ipython, ipython-notebook or ipython-qtconsole from the terminal.
Download the Blender (.blend) example using the two Churchill encrypted images overlaid from here

Since the start of 2015 I've been following Prof. Christof Paar's online lecturers on applied Cryptography, Ciphering and Encryption and got quite hooked to it. Many things went over my head but I think I have a good understanding of Applied Cryptography. This has helped me enormously to envision concepts and frameworks to start working on aural cryptography and how could I apply it to creative practice.
A comprehensive video of historical modern cryptography can be also found here

This project on "Aural Scramble" focuses on a relatively unexplored form of data cryptography which uses Sonic Encryption, as a form to hide Aural messages in the context of the Augmented City. This technology will allow citizens to implant encrypted audiovisual information in the City, which can be retrieved by other citizens using GPS-aware devices (e.g. smartphones).
The project examines how aural and visual encryption can play an important role to determine the pace, distribution and grammar of such encoded audiovisual signifiers, within the wider context of digital humanities. Critical engagement with technology will provide strong evidences of how social network data might be implemented in the Internet of Things, raising a number of new questions about creativity but also about security and privacy.