Exploiting Random Chemistry for Digital Security

Abstract

Randomness is such an important resource of our digital information era that new random number generators are invented continuously. Typically, these are electronic devices. Recently, researchers of the Universities of Glasgow and Edinburgh devised a novel method using stochastic chemical reactions. Randomness is such an important resource of our digital information era that new random number generators are invented continuously. Typically, these are electronic devices. Recently, researchers of the Universities of Glasgow and Edinburgh devised a novel method using stochastic chemical reactions. In our current era of digital information, randomness is an invaluable resource. Why is this so? Behind the scenes (to the user), the quasi-totality of digital communications uses cryptographic protocols to secure the exchange of information between communicating parties. Cryptographic protocols require random numbers to generate the keys for encryption and decryption. For this reason, random numbers generators (RNGs) can be found inside computers, mobile phones, bank tokens, etc. In the vast majority of these devices and platforms, RNGs are subroutines called within the larger piece of software that implement the cryptographic layer of the communication protocol.1Ferguson N. Schneier B. Kohno T. Cryptography Engineering: Design Principles and Practical Applications. Wiley Publishing, Inc., Indianapolis2010Google Scholar Although this might seem a convenient and fast way to generate the keys, such algorithms only produce artificial randomness or, as it is commonly called, pseudo-randomness. Simply put, if it is programmed/coded, it is predictable. If a pseudo-random number generator (PRNG) algorithm is used to simulate a sequence of dice rolls, the dice will always produce that same sequence of outcomes. The signature of randomness is unpredictability, and this can be achieved using physical random processes only. The lack of unpredictability of PRNGs is an obvious and well-known security issue, which is typically mitigated by initializing the PRNG algorithms with dynamic environmental inputs as a numerical starting point (or seed), for example, from the sensors of the mobile phones, movement of the mouse, or even the time and date.2Wikipedia contributorsRandom number generator attack.https://en.wikipedia.org/w/index.php?title=Random_number_generator_attack&oldid=930888162Google Scholar Using a simple seed such as date, a number generated on February 2nd, 2019 will differ from that on April 4th, 2020. But if you know the date of generation, you can easily guess (predict) the number! Although PRNGs continue to be widely employed, it is widely known that there is an ever increasing and cross-disciplinary need to develop devices that actually employ physical processes to generate unpredictable random numbers. The typical working scheme of a physical random number generator, also known as true random number generator (TRNG), features two main blocks. The first block is the entropy core, i.e., the physical system which evolves in an unpredictable way. The second block is the measurement system, which samples the entropy core and converts its random state into digital numbers. As an example of a well-known TRNG architecture, we can think to generators based on thermal or so-called Johnson noise. The entropy core is a resistor, and the measurement system is a comparator that measures the voltage at its terminals. Due to the thermal agitation of the charge carriers inside a conductor, the voltage fluctuates randomly. The comparator then outputs a digital signal, low (0) or high (1), if the instantaneous fluctuation is measured to be above or below a given reference threshold.3Horowitz P. Hill W. The Art of Electronics.Third Edition. Cambridge University Press, New York2015Google Scholar Electronic noise is a preferred process for research and patents, since it is easier to develop miniaturized TRNG by exploiting such “microscopic” quantum or chaotic phenomena. In a recent Matter article,4Lee E.C. Parrilla-Guitierrez J.M. Henson A. Brechin E.K. Cronin L. A Crystallization Robot for Generating True Random Numbers Based on Stochastic Chemical Processes.Matter. 2020; 2: 649-657Abstract Full Text Full Text PDF Scopus (11) Google Scholar random numbers are instead cleverly produced by exploiting macroscopic phenomena of chemical origin. In this case, the entropy core of the generator is compounds formed by chemical substances that undergo processes of crystallization and clustering. While chemistry can tell us that by making the input reagent X to react with the input reagent Y, we will produce crystals of XY, what chemistry cannot predict is the “macro-state” of such a compound, i.e., the number of crystals, their shape, and their appearance. This uncertainty about the actual end product of a chemical reaction is what Cronin and colleagues have exploited to build their TRNG. In the previous example, numbers are extracted from the random voltage fluctuations. The generator introduced here instead extracts numbers from the morphological features of the resulting compounds. As a matter of fact, the generator measurement block is an imaging system that collects and analyzes images of the crystals formed at the end of the reaction. A camera takes a picture of the vial from the top, and then a computer analyzes the digital image by means of feature detection algorithms. Since the final macro-state of the reaction is unpredictable, such is the number of pixels spanned by each crystal, their position, their orientation, and their color. These randomly varying quantities are then extracted from each crystal and converted into random bits, which are then concatenated to the bits obtained from the other crystals. To the bit strings, algorithmic post-processing is applied, a usual procedure employed to enhance the statistical uniformity of the strings. Thus, the post-processed bits are analyzed with statistical tests for randomness assessment and passed with high success ratio. This is an interesting result, and it is consistent with other TRNGs in the literature. The approach adopted by the authors indeed reminds of other methods to distil randomness out of images with complex features. As an example, random patterns can be generated by propagating a laser beam through a fixed or moving volume of scattering centers and then project on a screen the output intensity field.5Marron J. Martino A.J. Morris G.M. Generation of random arrays using clipped laser speckle.Appl. Opt. 1986; 25: 26Crossref Scopus (23) Google Scholar, 6Horstmeyer R. Chen R.Y. Judkewitz B. Yang C. Markov speckle for efficient random bit generation.Opt. Express. 2012; 20: 26394-26410Crossref Scopus (8) Google Scholar, 7Marangon D.G. Vallone G. Villoresi P. Random bits, true and unbiased, from atmospheric turbulence.Sci. Rep. 2014; 4: 5490Crossref Scopus (24) Google Scholar A so-called speckle pattern is then obtained, with randomly distributed spots of bright and dark intensity due to constructive and destructive interference caused by the random phase shift experienced by the field along the propagation path. Thus, from frames of the complex pattern, random numbers can be obtained by means of an image analysis similar to that employed by the authors. With the exception of the famous “Lavarand,”8Wikipedia contributorsLavarand.https://en.wikipedia.org/w/index.php?title=Lavarand&oldid=917654657Google Scholar these kinds of image-based TRNGs typically did not evolve into something more than a proof of principle. However, the authors of this paper took steps toward realizing a working prototype by manufacturing a robot, which fills the array of vials with chemical reagents and takes the pictures of the compounds in an automatic way. This is also relevant because with an automatized system to capture the images, the entropy core is sampled with fixed experimental conditions. With a camera acquiring pictures of each vial at the same coordinates, angle, illumination, etc., possible artifacts are easily detectable, the method becomes reproducible, and the average amount of bits generated by each reaction can be directly compared. There are at least two points that might be worth investigating before using this TRNG for cryptographic application. The first point to investigate would be a theoretical model of the chemical processes. This is important to understand which factors affect the reactions and consequently the randomness, e.g., temperature, reagent concentration. In a cryptographic framework, it is indeed essential to monitor and control those side channels that could exploited by an adversary to steer the reactions and possibly predict the output numbers. The second point is the bottleneck relative to the time required by a reaction to be completed, which is 150 min. This represents a clear limitation for cryptographic applications that continuously demand fresh randomness. In this respect, suggesting to periodically acquire pictures of the vials during the formation of the compounds would not be a recommendable solution without a theoretical model of the process. In fact, one should demonstrate that it is not possible to predict the evolution of the compounds from a picture taken earlier. A theoretical model would then help also to understand if a minimum timescale exists after the compound has “lost memory” of its previous states. Obviously, the biggest challenge to face before this TRNG could be an appealing alternative to a PRNG is scaling down its dimensions. To achieve this, the authors present an ambitious idea. While we wait as they implement it, a useful service the robot could provide is distributing its chemical random numbers online to be used for simulations (and maybe even for predictive chemistry simulations!). A Crystallization Robot for Generating True Random Numbers Based on Stochastic Chemical ProcessesLee et al.MatterFebruary 10, 2020In BriefCrystallization of compounds can be used as an entropy pool in the generation of random numbers, which have applications in data encryption, and to investigate stochastic chemical processes. Automation of chemical reactions and crystal detection has enabled the collection and processing of the required large amounts of data on crystal growth and formation to allow production of such numbers and investigations. Full-Text PDF Open Access