Journal of Signal and Information Processing, 2011, 2, 1-10 1 doi:10.4236/jsip.2011.21001 Published Online February 2011 (http://www.SciRP.org/journal/jsip) An Efficient Noise Generator for Validation of Channels Equalizers Nihar Panda1, Siba P. Panigrahi1, Sasmita Kumari Padhy2 1Electrical Engineering, Konark Institute of Science & Technology, Bhubaneswar, Orissa, India; 2ITER, SOA University, Bhu- baneswar, India Email: [email protected] Received December 10th, 2010; revised January 11th, 2011; accepted February 18th, 2011 ABSTRACT This paper develops an efficient pseudo-random number generator for validation of digital communication channels and secure disc. Drives. Simulation results validates the effectiveness of the random number generator. Keywords: Digital Communication, Channel Equalization 1. Introduction encryption keys, and so forth, that is, loss of security. Physical entropy sources are used to initialize genera- Digital data while transmission over communication chan- tors for cryptographic random number at every power up, nels becomes corrupted because of Inter Symbol Inter- and at special requests, like at reinitializing the firmware, ference (ISI), Co-Channel Interference (CCI), multi- or before generating long used cryptographic keys. Seed- path-fading etc. All the parameters that make a data cor- ing with physical values that can not be predicted makes rupted are known as noise. Extraction of transmitted data a cryptographic random number generator to supply mitigating this noise is known as equalization. Details pseudorandom sequences, with negligible probability of about the algorithms used in the channel equalization repetition. Correspondingly generated secure random se- filters can be found in [1,2]. However in order to validate quences this way needs no secure protected storage for an equalizer, the same need to be tested in presence of a keys or for the internal state of the generator, therefore it noise or random number generator. reduces costs and improves security. The noise sources and levels have been extensively In next sections, we discuss how an available digital studied in [3,4], their effects on the signal in the read signal with random components, the coefficients of the channel have also been investigated in [5-8]. The result- adaptive channel filter, is used in seeding a cryptographic ing inherent randomness in the channel filter coefficients random number generator in self-encrypting disk drives. has been proposed for use for random number generators The available physical entropy estimation is discussed, in [9], but the included randomness extraction algorithm resulting in an efficient seeding process. These will pro- is very inefficient. vide confidence in the generated random numbers for Generators for Cryptographic random number are em- their users, and tools for developers of embedded random ployed in many systems, like in self-encrypting disk number generators in testing and evaluation of designs. drives, such as the Seagate Momentus Full Disk Encryp- tion (FDE) drives. The random numbers so generated can 2. System Overview be used for encryption keys, facilitating secure commu- 2.1. The Architecture nication (via nonces), performing self-tests, and so forth. Previous data of the random number generator are diffi- The read and write transducers are embeded on the head cult to store securely, because an attacker could read, and which is separated from the rotating disk by an air bear- in some point in the future restore earlier states (together ing that keeps the read/write transducers at a distance of with any possible local authentication tags) with the help about 10 nm from the disk surface [10]. The head is em- of specialized tools (spin stand), and so force the genera- bedded on an arm, which is connected to an actuator. In tion of the same random sequence as earlier. This may a usual 3.5" disk drives this arm of 5 cm long and prone cause repeated nonces, of which recurring use of the same to mechanical vibrations, affected by air damping while Copyright © 2011 SciRes. JSIP 2 An Efficient Noise Generator for Validation of Channels Equalizers the drive is in operation. The vibration in vertical direc- 3. Entropy Sources in Rotating Disc Drives tion affects the amplitude of the read signal, while the radial vibration affects the noise pattern from the granu- There are many random physical processes, noise sources lar structure of the magnetic particles and cross talk from in disk drives. Cost constraints compel using electronic neighbor tracks, because of the small spacing between signals, which are available in digital form in standard tracks (in the range of 10-100 nm). unmodified disk drives, and which contain strong ran- To have the head to be on track, servo patterns are dom components. At the time of booting, or at a special written on the disk. These servo patterns are arranged in request they provide the entropy sources to seed an radial spokes, which are traversed by the head about 200 SW-based generator for cryptographic random number of times per revolution (at a rotational speed of 5400 rpm, self-encrypting disk drives, ensuring the uniqueness of 18000 times per second). After the head covers these the generated (pseudo) random sequences with very high servo patterns, a controller used to evaluate the read sig- probability. nal and corrects the radial position accordingly. It also In disk drives currently available in the market several makes the channel equalizer filter for optimum signal such sources are used. Combinations of their data give a shaping. The tracking correction is based on the present better quality; the speed of the random number genera- radial position, velocity, and acceleration of the head. tion, and the safety against potential attacks influencing These values are random, strongly affected by turbulent the entropy sources. air damping and mechanical vibrations. This is still to be 3.1. Timing Variations explored with a useful model of the disk drive physics. In [3] some mathematical formulations are presented, but still In the disk drive literature there are internal high-speed lacks a reasonably accurate picture of disk drive internals. counters available. Least significant bits of these disk drives are sufficiently random when sampled during the 2.2. Entropy Requirements disk boot up process, or in general, after actions involv- In this paper, we show that disk drives can provide physi- ing a lot of mechanical activities of timing uncertainties, cal randomness for seeding generators for cryptographic such as at spin-up and rotation of the motor and platters, random number, but they are targets to specific attacks, and at arm movements in seek operations. These random exploiting their use and special characteristics, leading to bits can be collected into an entropy pool, and consumed specific entropy requirements of the disk. The general- on requirement. The entropy of the timing data can be ized “birthday bound” speaks that after taking 2n/2 sam- found in [11]. ples there is a 50% chance of a uniformly distributed Such random number generators have been presented, n-bit random variable to attain the same value more than the slow [3], implemented externally in the host com- once. In a data center an virus could observe thousands puter, where synchronous communication masks off of disk drives rebooting thousands of times, so 107 ≈ 223 most of the original timing variations. samples from different random number sequence are 3.2. Tracking Error easily taken. When a network shares these results, one could build a database from over 232 initial sets of values Another type of randomness source was investigated in of the random number generator, to search for a collision. [12], with the tracking error of the magnetic read head It gives rise to a requirement of at least 64-bit entropy of trying to remain in the middle of the path of the recorded the seed. Of course, a 50% chance of a successful attack data. Consecutive samples are strongly correlated, which is too high. A commonly accepted allowable collision limits the entropy that can be used. Results in this paper probability is 10−8 (half of the chance of hitting the in the newest generation of disk drives showed much less jackpot in a 5-out-of-90), which adds 27 bits to the en- achievable speed or entropy/s than claimed in [12], but tropy requirements for the seed, so for unlikely repeated the position error of the read/write head certainly repre- sequences the entropy of the seed has to be more than 90 sents another alternative source of randomness bits. To serve for HW differences, environment changes, 3.3. Channel Filter-Coefficients and so forth, at least 128-bit entropy is desired for the seed of a cryptographic random number generator. The drive firmware can access the coefficients of an adap- The smallest AES cipher needs 128-bit fully random tive channel-equalizer, via a diagnostic interface between encryption keys, also posing the requirement of at least the main control ASIC and the channel signal processor, 128-bit seed entropy. (High entropy public keys and which also does the coding/decoding of the head signal longer symmetric keys must be generated with several [13]. Resistor values of an analog filter represented by calls to a reseeded generator for cryptographic random the coefficients, continuously tuned by the control number). mechanism of the read/write channel chip to make the Copyright © 2011 SciRes. JSIP An Efficient Noise Generator for Validation of Channels