Home

View/Open - Oregon State University

1. 0x00000000 0x00000000 0x00010000 A 0x00000000 0x00000000 0x00000000 B 0xe24bdalc 0x878e451c 0x3b2a0000 modulus 0xe29bdac8 0x7c83251c 0xb24d0000 ECC point addition in F 2 n e Running the above test in F 2 m e From table 7 the average time to execute 1869 8 26 1 microsecs ECC point doubling e Elliptic curve point doubling of Repetitions Time Repetition microseconds Error ECC 625 1891 196 addition 625 2158 196 F p 500 2264 196 625 2424 196 625 2211 196 TABLE 6 WindView timing for ECC addition in F p of Repetitions Time Repetition microseconds Error ECC 500 1998 196 addition 625 1891 196 F 2n 500 1731 196 625 1811 196 625 1918 196 TABLE 7 WindView timing for ECC addition in F 2n e Routine used mpc180EccDoubleQ Data Used B 0xec8a7a0b 0x05a00a4c 0x6b880000 X1 0x20f9d2a6 0x9fb3d9cb 0x7ee70000 Y1 0x35c900a5 0x678c976d 0x789f0000 Zl 0x00000000 0x00000000 0x00010000 A 0x00000000 0x00000000 0x00000000 MODULUS 0xec8a7a0b 0x05a00a4c 0x6b8b0000 49 50 of Repetitions Time Repetition microseconds Error ECC 625 1518 196 point doubling 625 1651 196 F p 500 1898 196 625 1971 196 625 1838 196 TABLE 8 WindView timing for ECC point doubling in F p Similar to the tests above ellip
2. Random Number Generator RNG Module 50 4 4 Message Digest Hardware Accelerator MDHA Module 52 5 Discussions 1 os e o c i e 56 56 512 Conclusions ER Eeer 57 Did ege Works ae eae e AES 58 BIBLIO GRAPHY Lente HH vk doch oc ERE ERE E 59 LIST OF FIGURES Figure Page 1 Architecture of the MPC180 security co processor 5 2 Real time operating systems sen 11 3 The real time operating system kernel 12 4 The host side of the Tornado environment 16 5 The target side of the Tornado environment 20 6 Tornado s Target HGerver 20 7 Tornado s Deet due eres RE e eres Rs 23 8 Tornado s built in simulator esses 25 9 Tornado s project management facility 20 10 Tornado s WindView facht 28 11 Hyperterminal startup screen 31 12 VxWorks ready screen 33 13 Tornado startup screen emn 37 14 Tornado Application Creation Wizard 38 15 Tornado Application Creation Wizard Step 2 39 16 WindView timing graph for BC 42 LIST OF TABLES 1 WindView timing for ROCHA 41 2 WindView timing for RSA exponentiation 44 3 WindView timing f
3. e MDEU Message Digest Execution Unit which handles the SHA 1 and MD 4 5 computations e AFEU Arc Four Execution Unit which handles the RC4 computations and e RNG Random Number Generator which generates 32 bit random numbers These EUs are connected to the FIFOs and the controller via multiplexers The EBI mentioned above is responsible for selecting the necessary execution unit Once the appropriate EU has been selected it receives the input data from the input FIFO processes the data and returns the output to the output FIFO This output data is then written back to a specific address in the host PC s memory map A description of the individual execution units is given below e Public Key Hardware Accelerator PKHA Module The PKEU supports elliptic curve cryptography and Rivest Shamir and Adleman 5 algorithms Elliptic curve is supported in both F 2 m polynomial basis and F p modes 7 The mpc180Pkha module performs advanced mathematical functions to sup port RSA and ECC public key cryptographic algorithms The PKHA module comprises of three sub modules The names of the sub modules are mpc180PkhaUtil mpc180PkhaEcc mpc180PkhaRsa The first of the sub modules the mpc180PkhaUtil executes the following func tions Load and read large number registers Load and read the modulus size Clear the register memory and perform supporting calculations The second sub module the mp
4. of security for information that needs to be protected This thesis deals with how embedded systems are being used to provide cryptographic security to network data The focus of this thesis is the MPC180 which is a network security co processor that is designed to execute computationally intense algorithms generally associated with cryptography 1 1 2 Embedded systems in today s networks An embedded system is a combination of hardware and software designed to per form a specific function 1 Embedded systems are becoming the new frontier for secure computing These systems are now being used to provide security for cellular and wireless data on demand video and internet traffic to name a few areas 2 The real time nature of embedded system processing presents great advantages to the na ture of applications mentioned above Other advantages that embedded systems offer 2 over traditional systems are the secure storage of secret information such as private keys and higher speeds of execution 1 2 The Motorola MPC180 security co processor Data security is a critical concept in today s world Cryptography is essential to securing data and communications and is extensively used in electronic commerce wireless communication and banking The rapid miniaturization of processing power has led to increasingly smaller devices that perform custom duties and in our case cryptography This thesis deals with such a device called
5. 1024 bit RSA exponentiation in 33 milliseconds 14 SIDSA s RSA co processor performs 1024 bit RSA exponentiation in 60 milliseconds 15 The underlying point to be noted here is that the MPC180 does not claim to be faster than software implementations It is a dedicated solution to encrypting network traffic and its purpose is to shoulder the complete responsibility of performing cryptography while the host processor performs its other duties while not being bogged down by the load of intensive cryptographic computations A hardware implementation also provides a critical advantage from the security viewpoint The hardware implementation is tamper proof and ensures that sensitive information like secret keys are not compromised Software implemen tations on the other hand are susceptible to reverse engineering and other common attacks 5 2 Conclusions This work presented the benchmarking of Motorola s MPC180 security co processor This processor is intended for use with Motorola s POWERQUICC family of com munication processors and it was tested by running code for various cryptographic algorithms and operations like RSA 5 16 stream ciphers RC4 message digests random number generation and operations over Elliptic curves 17 This security co processor has shown that it performs intensive computations with good timing while offloading this burden from the host processor The MPC180 is thus a good solution to provide dedica
6. Security Co Processor by Uday Banerjee A THESIS submitted to Oregon State University in partial fulfillment of the requirements for the degree of Master of Science Presented December 4 2003 Commencement June 2004 Master of Science thesis of Uday Banerjee presented on December 4 2003 APPROVED Major Professor representing Electrical amp Computer Engineering Director of the School of Electrical Engineering and Computer Science Dean of the Graduate School I understand that my thesis will become part of the permanent collection of Oregon State University libraries My signature below authorizes release of my thesis to any reader upon request Uday Banerjee Author ACKNOWLEDGMENT I would like to thank Dr Ko for giving me the opportunity to work on this interesting project Dr Kog s discussions directions and guidance helped me greatly in accomplishing this work I thank Dr Tenca who made sure that this project received the necessary tools and software I also thank Dr Traylor for his support I greatly appreciate the help and discussion provided by my friends at the ISL Gokay Saldamli Lo ai Tawalbeh and Dana Zottola I would also like to thank the community at comp os vxworks for all their help Finally I am grateful for all the support my family has given me during this project TABLE OF CONTENTS Introduction 2 neiacadasecudacaerdedadaddadaduiddebrisdddas 1 LI Motivation
7. de scribed above and the VxWorks environment that is running on the board This interface comprises of three main components namely the Target Agent the Target Server and the Tornado Registry Each of these are explained in detail below e Target Agent All Tornado tools are represented by the target agent on the tar get The target agent is a compact implementation of the core services necessary to respond to requests from the Tornado tools The agent responds to requests transmitted by the target server and replies with the results These requests include memory transactions notification services for breakpoints and other target events virtual I O support and task control The agent synthesizes two modes of target control task mode addressing the target at application level and system mode system wide control including Interrupt Service Routine de bugging The agent can execute in either mode and switches between them on demand The agent is independent of the run time operating system inter facing with run time services indirectly so that it can take advantage of kernel features when they are present but without requiring them The agents driver interface is also independent of the run time avoiding the VxWorks I O sys tem Drivers for the agent are raw drivers that can operate in either a polling or an interrupt driven mode A polling driver is required to support system level 19 breakpoints Run time independence means
8. from Windview Tornado s run time analysis software The run time scenario is depicted clearly and shows the introduction of specific Event points These event points are instrumental in accurately measuring the execution time of the AFHA unit Note that the task labeled t10 represents the execution of the RC4 algorithm by the MPC180 Using the event logging control library wvEvent is used to capture the exact timing of the encryption and decryption sequences The figure shows how the user defined events show up during run time analysis The encryption cycle is composed of two parts namely setting the key and actually processing the data encryption or decryption That is why the times are represented as a sum of two parts The timing obtained from the MPC180 running the RC4 algorithm is shown in table 1 42 0 06 0 87 INTG Z IMT User Betined Control Ewart lExcTask iLogTask hell tTelnetd wWdbTask to iNetTask DARIBullidMgr idle FIGURE 16 WindView timing graph for RC4 Encrypting multiple sets of data with the AFHA It may be noted that the sequence of commands used to encrypt the data is 1 mpc180AfhaSetKey 2 mpc180AfhaProcess Thus if it is desired to use the same key to continuously encrypt successive blocks of data using the same key mpc180AfhaSetKey need be used only once and mpc180AfhaProcess can be called as many times as is required The AFHA also supports context switchin
9. host and the target A useful feature of the target server is that it caches target information like memory addresses so that in case a tool requests a memory read from the target there is no need for the target server to even perform a request for target memory access When the target server runs its name is linked with the name of the host machine and the name of the target For example if the name of the host machine is izmir and that of the target is zpc860 the Target Agent FIGURE 5 The target side of the Tornado environment 7 wxsimzSuma Lop Console totevr vasin ouaa Fri Wor 14 10 50 46 2003 I Wind Hiver Systens Target Server HTVim35 version Censmectimg te target agent et Attachimg C inter sce Shutdoen Attaching pascoli ONE Smee fer SINET CPU Lanz Lo zuzena FIGURE 6 Tornado s Target Server 20 21 target server would be named zpc860 izmir Figure 6 shows a screenshot of a target server in action The target server keeps a record of the symbol table for the target executable on the host computer This symbol table is built by the target server from a file known as the core files and symbols which can be selected from the drop down menu called Target Server Configuration From several users experiences as reported in 10 synchronization problems were rectified by specifying the vxWorks file in the core file and symbols option instead of using the vxWorks rom fil
10. of a registry that is running on the host at any given time but it is recommended that only one instance of the registry be run at a time to allow for unhindered access to the target by the tools 8 Figure 7 shows a Tornado Registry in operation 2 2 2 Tornado Tools The Tornado Integrated Development Environment IDE is a comprehensive suite of software and tools used for real time application development The salient features of the IDE are as follows 8 e Integrated source code editor Tornado ships with an integrated source code editor that performs standard functions like C C syntax color highlighting The editor tracks execution of the code and is integrated with the debugger and the project management facility e Integrated C C compilers Tornado includes a standard C C compiler 3 e CrossWind CrossWind is a Graphic User Interface GUI based debugger based on the GNU source level debugger CrossWind can be used to perform tasks like debugging on the target attach to existing tasks view application code and use debugging features at both the application and system levels 3Most commonly GNU s gcc compiler 8 9 25 E VxWorks Simulator for Windows PRSE VxWorks Copyright 1984 1338 Wind River Systems Inc CPU VYVeSim for Windows VxWorkzg 5 4 2 BSP version 1 1 1 Creation date Jan 29 2001 WDB Read FIGURE 8 Tornado s built in simulator WindSh WindSh is a C based command s
11. real time scheduling intertask communication and mutual exclusion This kernel is linked to the file systems device drivers I O systems and the network stack More information about the various facilities offered by VxWorks are described below The subsequent sections will deal with how VxWorks is used in real time development 2 1 2 VxWorks RTOS Facilities Real time systems as mentioned above are designed to work in conjunction with standard operating systems The following material will provide a brief overview of the salient facilities and tools provided by VxWorks 7 13 Facilities e Input Output System VxWorks comprises a speedy stable and flexible I O system It includes several drivers that encapsulate key input output mecha nisms like keyboard driver RAM Disk driver display driver etc e High Performance Real time kernel facilities VxWorks kernel called Wind is a multi tasking kernel and its key features are real time scheduling mu tual exclusion semaphores and intertask communication These features are what distinguish the real time operating system from the non real time operat ing system The kernel is very fast and engages in preemptive priority based scheduling of its tasks This kernel belongs to what we would call a micro kernel architecture for it is small and highly configurable e Fast file systems VxWorks employs custom tailored file systems that are de signed for real time oper
12. the MPC180 security co processor developed by Motorola Inc T his network security co processor is designed to perform fast on the fly encryption decryption of data thus adding security to data that is to be sent over a network The following is a brief summary of the features of the MPC180 security co processor 3 1 2 1 Features The MPC180 is based on Motorola s MPC8xx or MPC826x family of POW ERQUICC communicator processors The MPC180 can handle intense cryptographic applications and offloads the computational burden associated with them from the host processor The MPC180 is designed to support all the algorithms associated with IPSEC IKE and SSL TLS This processor is capable of providing elliptic curve cryptography support for WLTS WAP in wireless applications The MPC180 has the following features 3 e PKEU Public Key Execution Unit RSA and Diffie Helman Signature time of 32ms 1024 bit 10 IKE handshakes per second Programmable field size of 80 2048 bits Elliptic Curve cryptography in either FO or F p 155 bit signature time of 11ms 30 IKE hand shakes per second Programmable field size of 55 511 bits DEU DES Execution Unit 3DES HMAC SHA 1 performance upto 15 Mbps ECB and CBC modes for both DES 3DES Support for 2 key and 3 key 3DES AFEU Arc Four Execution Unit RC4 up to 20 Mbps Programmable key size from 40 to 128 bits MAU Message Authentication Unit SH
13. 104 5 1 500 103 5 1 TABLE 5 WindView timing for RSA modular multiplication We then have tmu 1 F MS 10 MS 27 1 50 a 109 64 10 64 27 96 microseconds The windview timing from table 5 shows a timing of 103 7 microseconds which is reasonably close to the expected theoretical value Please refer to table 14 for an explanation of the variables used in the formulae for the performance of the MPC180 4 2 20 ECC Module The Elliptic Curve Cryptography module performs elliptic curve point multiplica tion elliptic curve point addition elliptic curve point doubling ECC modular addi tion ECC modular subtraction ECC modular multiplication and the Montgomery modular multiplication in F p The above capabilities are tested and the results are shown below 48 ECC point addition in F p e Elliptic curve point addition R P X Q where P X1 Y1 Z1 Q X2 Y2 Z2 and Z X3 Y3 Z3 in F p e Routine used mpc180EccAddPtoQ e From table 6 the average time taken to execute 2189 6 26 196 microsecs The above function has been tested by the function timex from the WindView timing library in order to get an average time Data X1 0x6472bc69 0xd14ceTeb 0x574d0000 Y1 0xdd10558d 0x61ad63b7 0xd00f0000 Zl 0x00000000 0x00000000 0x00010000 X2 0x707c9fdf Oxdfd914f1 Oxe2af0000 Y2 0xa064d790 0xf575c335 0x722b0000 Z2
14. 23 iei Tornado Registry Iren egay Meise wtxregd wteregd Suma Fri How 14 11 27 44 2003 He Vind River Syetens Tornado Registry HI Wini Ternado Registry initializing Succeedad fies Tornado Registry restoring data base from fi bal Tornado Pegistry Server Installing SCC FIGURE 7 Tornado s Registry 1 n server name specifies an alternate name for the target server In this thesis the server name was mpc860 2 V This activates the Verbose mode where the target server displays information and error messages on the designated output 3 B backend name Specifies the name of the back end used to commu nicate with the target agent In this case the back end used was wdbrpc 4 R root Specifies the root directory of the host file system 5 RW Grants read write privileges to the target processes over certain host files 24 e Tornado Registry The registry is a service that maintains a database of target servers boards and RPC port numbers in the cross development system The registry facilitates the Tornado tools to connect to the target server The reg istry keeps track of each individual target server by its name which is in the form of a unique identifier e g wtxregd izmir Just like the target server the registry need not run on the host computer but it should have network access to the host in order for it to function properly There can be more than one instance
15. 311111141 Dewslnpmenk Eyma 13111181113231311113311111 11111131111 332111121111 Trka varela 3 4 111112211111111211111111 KECRHEL WIND wermibn 1111113111111111111111 copyright Wind Riwer Zywismes imo A9n Henn CPH Hoternla mprll cpo r 1a Meoetaerula Power H BO Procummor Hemery Tire a3cODOUOOO EXE wnexben i f Copypr n ht 2001 Hotu roha Copyright 1994 1997 wind nivar Systimes Ln Wt feste FIGURE 12 VxWorks ready screen This message is indicative of the fact that the board is completely booted and ready A screen shot of the board in ready state is shown in figure 12 3 4 Embedded system programming with VrWorks and Tornado Embedded systems require extremely efficient programming in order to adhere to the timing and performance constraints imposed upon them C and its derivatives C EC have always been extremely popular when it came to developing embedded applications The reason for this is that C gives developers a great degree of control over the hardware being used 12 The fact that C is a platform independent architecture also helps The MPC180 s software was written using C C owing to the fact that the Tornado VxWorks system uses C C The following section deals with how a 34 VxWorks program is written compiled built and downloaded onto the target for execution The demonstrations provided by WindRiver as part of Tornado are extremely useful in learning and understan
16. 34 rss cw cnet as Shee eee a ieee Rud 1 1141 Cryptography kocc oroen ex e e een ERROR RD 1 1 1 2 Embedded systems in today s networks 1 1 2 The Motorola MPC180 security 00 11068330 2 ded Beatures sisiunsisrke m Rp ober a dde 2 1 2 2 Architecture of the MPC180 4 The development environment and tools VxWorks and Tornado 10 27 MSEWOEPkSi eme e eter he d tace e utet e rau 10 2 1 1 Introduction to VxWorks 10 2 1 2 VxWorks RTOS Facilities 12 2 1 8 Why VxWorkst cei eee rU PERRO 15 252 4 TIOBIBGOn speret sese bre erre te ost 15 2 2 1 Introduction to Tornado 15 2 2 2 Tornado Tools e aut e eddie 24 Cross Development with Tornado and the ZPC860 29 3 1 What exactly does cross development mean 29 3 2 Basic operation of the system A user perspective 30 3 9 2Broceduresssciu bui Ee bee au RR E 30 3 4 Embedded system programming with VxWorks and Tornado 33 Results and Timing Analysis 40 4 1 ARC Four Hardware Accelerator AFHA Module 40 4 2 Public Key Hardware Accelerator PKHA Module 43 TABLE OF CONTENTS Continued 4 2 1 RSA Module eee ete ee 44 4 2 2 ECC Module orent 47 4 3
17. A 1 with 160 bit message digest MD5 performance 128 bit message digest HMAC with either algorithm RNG Random Number Generator Supplies upto 160 bit strings at 5Mbps data rate Software and Development Support VxWorks drivers Sample drivers SBC8260C development board and board support package from WindRiver e Packaged in a low cost 100 pin LOFP package e 1 8V VDD 3 3V I O e Low Power Dissipation at 0 6W e Input and Output buffer 4 Kbits each e Glueless interface to MPC8xx system or MPC826x local bus 50 MHz and 60 MHz e DMA hardware handshaking signals for use with the MPC826x 1 2 2 Architecture of the MPC180 This section outlines the architectural features of the MPC180 security co processor 3 For the purpose of this study the MPC180 is interfaced to the ZPC860 processor card which has the following features 4 e 32 MB SDRAM SODIMM e Five board to board connectors e 2 MB Flash memory e Built in RS 232 serial port e Built in ethernet port The MPC180 itself is meant to integrate into any system that supports Motorola s integrated communication processors 3 The MPC180 works such that it resides in the host processor s memory map Any application that wishes to use a cryptographic routine simply needs to write to a specific memory location and read the result back from another location Figure 1 shows the functional block diagram of the MPC180 s internal architecture 3 DMA Re
18. AN ABSTRACT OF THE THESIS OF Uday Banerjee for the degree of Master of Science in Electrical amp Computer Engineering presented on December 4 2003 Title Hardware Encryption Using the MPC180 Security Co Processor Abstract approved Cetin Kaya Ko Hardware cryptography offers a reliable way to provide security for data that is sent over networks Security processors have taken the stage in the movement for secure computing Some of the more prominent security processors belong to the S1 family created by Motorola The primary goal of this thesis is to document and test the performance of one of these processors the MPC180 by executing efficient code for standard cryptographic algorithms The MPC180 is a network security co processor which is capable of performing on the fly encryption of all network data thus adding an extra layer of security This processor is capable of performing inten sive cryptographic computations while offloading the burden of these computations from the host processor thus freeing up valuable resources on the host to perform other functions A real time operating system VxWorks and its associated tools Tornado were used to accurately examine and time the operation of the security processor The results are compared to other cryptographic hardware and software implementations Copyright by Uday Banerjee December 4 2003 All rights reserved Hardware Encryption Using the MPC180
19. PkhaRsa is used to perform modular arithmetic like the RSA algorithm DSA algorithm and Diffie Hellman Again the computations are classified into two levels The first level high level performs modular exponen tiation X Ymod N The second level of computation involves modular addition subtraction and multiplication It may be noted that the minimum number of digits that the RSA and ECC routines can work with are 128 and 32 respectively 44 of Repetitions Time Repetition milliseconds Error 1 160 5 196 RSA 1 159 5 196 Exponentiation 1 161 5 196 1024 bit 1 156 5 196 1 159 5 196 TABLE 2 WindView timing for RSA exponentiation 4 2 1 RSA Module The basic functions of modular arithmetic are covered in this section Modular exponentiation addition subtraction and Montgomery multiplication are performed for 1024 bit numbers RSA modular exponentiation e RSA Exponentiation X Y mod NI e Routine used mpc180RsaExpA e From the results in table 2 the average time 159 5 milliseconds It may be noted that RSA exponentiation is a slow process when compared to other modular operations RSA exponentiations involve repeated squaring and multiplica tion and in our case we are using a 1024 bit exponent This exponentiation would alone require 1024 squaring operations and approximately 512 multiplications Thus the large timing required to perform RSA exponentiation RSA modu
20. This method has been used in this thesis to time certain functions during run time A screenshot of a Wind View output is shown in figure 10 The subsequent chapter deals with cross development on the VxWorks Tornado system 28 Erem EN Defien Ze wietere Dua jm aw zixiw sj mix v E gt gt Tal Gal DER RE H INTE Wve Tonk ugtask itil feline PAbTaik n IMetT Panew idle 4 L eral Rice Sr pg Bari FIGURE 10 Tornado s WindView facility 29 3 Cross Development with Tornado and the ZPC860 3 1 What exactly does cross development mean Cross Development in the context of real time application development involves two connected systems which are completely different platforms but co operate to facilitate real time application development Real time application development re quires features from both traditional real time operating systems such as accuracy real time operation etc as well as features from non real time operating systems such as UNIX and Windows such as processing power memory capacity etc It makes sense to utilize the features of both the types of operating systems in order to successfully develop real time applications Thus the RTOS would handle important real time tasks while the regular operating system would handle tasks such as program development For instance consider the systems involved in this thesis The
21. ation The file system formats used are compatible with DOS RT 11 SCSI CD ROM and raw disk file system Overall these fast file systems are critical to the operations of a RTOS where timing is of the essence e C support VxWorks ships with a variant of GNU s GCC C C compiler There are other compilers available like the Diab compiler which is Wind River s own compiler e Target Resident Tools complete set of tools used for development are present on the host machine thus conserving target memory and resources but there is also a target resident shell a symbol table and a module loader unloader that can be configured into the target OS as required e Evaluation Tools and Utility libraries Performance evaluation tools that in clude utilities to display processor utilization percentages by a particular task and an execution timer are present VxWorks includes a comprehensive set 14 of utilities like timers interrupt handling memory allocation and ANSI C libraries Network Facilities VxWorks network facilities are compatible with standard internet protocols Thus setting up communications between the host and the target using regular equipment like routers is an extremely simple process Board Support Packages A board support package is the board specific part of VxWorks It contains initialization code for the hardware as well as device drivers for included components on the board like serial pa
22. c180PkhaEcc is used to perform elliptic curve cryptography Elliptic curve cryptography is supported in both F 2 n and F p The ECC module allows for two levels of operations The first is high level and executes elliptic curve point multiplication using either projective or affine co ordinates The second level allows for the execution of elliptic curve additions and point doubles T hese applications that run these operations are required to work with projective co ordinates with all the inputs and outputs being in the Montgomery residue system The third sub module the mpc180PkhaRsa is used to perform modular arith metic like the RSA algorithm DSA algorithm and Diffie Hellman Again the computations are classified into two levels The first level high level performs modular exponentiation X Y mod N and the second level of computation in volves modular addition subtraction and multiplication It may be noted that the minimum number of digits that the RSA and ECC routines can work with are 128 and 32 respectively e DES Hardware Accelerator Module The DEU is used to perform bulk data encryption and decryption in accordance with the Data Encryption Standard algorithm The DEU also computes Triple DES The MPC180E supports two key K1 K3 or three key 3DES The DEU operates by permuting 64 bit data blocks with a shared 56 bit session key and an initialization vector The MPC180E supports two modes of Initialization Vector ope
23. d for application development are an integral part of this thesis Real time operation is a very important part of this thesis and this is why VxWorks one of the popular real time operating systems RTOS and its associated Integrated Development Environment IDE Tornado have been used The reason why a real time operating system is used is because traditional operating systems like Windows and UNIX are ill suited to handle any real time applications On the other hand operating systems like VxWorks built especially for handling real time processes are poor performers when it comes to non real time application development VxWorks can operate in tandem with either UNIX or Windows so that each type of operating system RTOS and non RTOS can do what it does best This means that VxWorks would handle the time critical aspects while the host operating system would handle the program development and non real time aspect of the application VxWorks is very flexible in that it lets a user customize it to include the features that the application requires For example VxWorks allows the user to configure it such that features like networking can be added to speed up the development cycle and the same can be excluded in the post production phase Thus we see that a real time operating system has significant advantages over traditional 11 RS 252 MWetwork FIGURE 2 Real time operating systems operating systems when it comes to working
24. d to a Mew or Exising W olkspace C Tomado Mangel profWorkspaceTh wep zl lt gt lt lt FIGURE 14 Tornado Application Creation Wizard 39 Schu a teoichan lor budding Fus Project This toolchain can be set explcilly of can be copied ftom an esatting Downlaadahie Project Wiould you like Ip base pour popact orc DER Itten c MEE PENTILIMe SIMA Tara FIGURE 15 Tornado Application Creation Wizard Step 2 40 4 Results and Timing Analysis This chapter discusses the results obtained during the course of this thesis Presented below are the analyses and timing of the following execution units of the MPC180 security co processor They are e Arc Four Hardware Accelerator AFHA Module e Public Key Hardware Accelerator PKHA Module e Random Number Generator RNG Module e Message Digest Hardware Accelerator MDHA Module Timing is obtained from the real time analysis tools supplied by WindRiver s Tor nado The tool is called Windview described in Section 2 2 2 and its role is to collect and analyze run time data Windview can be configured in a variety of ways and I have chosen to use task state transition as the parameter This allows us to pin point when exactly the mpc180 s various routines execute For routines that take microscopic amounts of time Tornado provides a timing library that contains routines for timing the execution of programs individual functions and groups of functi
25. de Brian W Kernighan and Dennis Ritchie The C Programming Language Pren tice Hall PTR second edition March 1988 Crypto 5 1 benchmarks http www eskimo com weidai benchmarks High performance rsa co processor www sci worx com Sidsa rsa2048a rsa co processor www sidsa com K Koc High Speed RSA Implementation RSA Laboratories second edition November 1994 N Koblitz Elliptic curve cryptosystems Mathematics of computation vol 48 no 177 pp 203 209 January 1987 18 nShield A hardware security module from nCipher http www ncipher com nshield index html 60
26. ding cross development Developing code for VxWorks Firstly all C code must include the VxWorks header file vxWorks h in order for Vx Works dependent code to be able to function The code is written in standard C C format and can take advantage of the functions that are supported by the hardware To do this one must install the necessary device drivers for that particular hardware and include the necessary header files Once the code has been written and added to the main project using the Project Facility in Tornado the code is compiled by the GNU s GCC compiler This project is then built by Tornado that generates the corresponding project_name out files which are to be downloaded to the target After these files have been downloaded to the target the Tornado Shell is used to call functions and routines that are present in the code To summarize e All necessary device drivers and Board Support Packages for the target must be installed initially e All VxWorks based C C code must make sure to include the vxWorks h header file e After starting Tornado the user is presented with the option to create a project in a new existing workspace as shown in figure 13 e After clicking on the New tab the user will be presented with the option of either creating a Bootable VxWorks image or Downloadable application modules for VxWorks for our purposes we must select Downloadable appli cation modules f
27. e It is also suggested here that the option synchronize host target symbol table is selected in the target server configuration The target server could also operate on a different computer from which the Tornado tools reside The tools need to have network access to the remote target server by some means in order to allow them to communicate with the target The target server needs to be configured with the following parameters Jarget Server Name This is what the target server will be called Upon choosing the option Add description to menu this name will be available for selection in the Tools Target Server menu Target Server Properties The following are listed under this option 1 Authorizations Option used to restrict users access to the target server This option allows one to specify the location of an autho rization file which contains the names of authorized users This file can be found in WIND_BASE wind userlock 2 Back End Allows the user to specify which back end is to be used to connect to the target The back end which needs to be used in the case of hardware targets is wdbrpc as specified in 8 For a complete One can refer to 10 for arguably one the most comprehensive frequently asked questions list compiled on VxWorks Tornado 22 list of options one can refer to Tornado s Help menu Manuals Contents Tornado Reference Tornado Tools
28. e one 32 bit random number is approximately 50 26 microseconds 100 block 32 bit random number generation e Routine used to generate 100 32 bit rng blocks is mpc180RandLongs e From table 11 the time required to generate one 32 bit random number is approximately 5606 microseconds 52 of Repetitions Time Repetition microseconds Error 100 175 5235 196 32 bit 175 5246 196 RNG s 175 5617 196 175 5469 196 175 6283 196 TABLE 11 WindView timing for 100 blocks of 32 bit RNG generation Sample 32 bit random numbers generated are shown below 0x31f4c0 value 117361797 0xf901337b 0x31f4c0 value 1402180736 Oxac6c6b80 Ox31f4c0 value 850937036 0x32b844cc Ox31f4c0 value 1952494407 0x7460b347 WindShell displays the output in the above format giving us a random number value in both decimal form 117361797 as well as hexadecimal form 0xf901337b 4 4 Message Digest Hardware Accelerator MDHA Module The mpc180Mdha message digest hardware accelerator module is capable of per forming hash algorithms like the SHA 1 and MD5 It is also capable of generating an HMAC which can be built upon an existing message digest calculation The basic routine used in these calculations is mpcl80Hash The arguments of this function are algorithm input buffer length of message output buffer The algorithm parameter specifies which algorithm this function must use to gene
29. e workspace window 2 Select Toolchain name which in this case is PPC860gnu 3 Right click it and select Build Project name out This will build all the files in the project and check for dependencies in the code If there are any A toolchain is a set of cross development tools that are used to build applications for a specific target processor 11 Please refer to 8 for information on how to configure Vx Works in the Project Manager 36 errors present in the code the errors must be rectified using the integrated source code editor 4 Once the build executes flawlessly the application needs to be downloaded to the target by choosing Download Project name out 5 The shell can be used to call the name of the application usually the name of the function inside the code by typing the function name at the prompt The target responds to the function call by producing the output on the shell window The above steps demonstrate how to build a simple application for use with the VxWorks Tornado system The next chapter deals with the testing and running of code on the MPC180 in order to perform fast encryption in hardware Tornade f Show this window on statup FIGURE 13 Tornado startup screen 37 38 Create downloadable application modules for VxWorks step 1 Hame Projeci19 BE la E Tomada poa reais zl Figpect desenplien opona lc Workspace Cr de bai fe Ad
30. e860 x Works e host name This is to help identify the host machine on the network for example Izmir e inet on ethernet This is the IP address of the board and it must be set properly with a valid value before the board tries to boot e host inet This is the IP address of the host machine The board tries to connect to this IP address over the network before loading the VxWorks file specified in the file parameter e gateway inet This is the IP address of the gateway computer and need be used only if the target and the host are not on the same subnet If the host and target are not on the same subnet it is recommended that the user first set the subnet mask of the board in the inet on ethernet parameter as xxx Xxx xxx xxx Oxfffffe00 which represents the IP address followed by the sample subnet of 255 255 254 0 4 Once the desired parameters have been set the command is used to boot the board with the current configuration At this time the board searches for the VxWorks image on the host machine as specified in the file parameter Provided all the network settings are correct the board then downloads the VxWorks image onto its memory and loads up the VxWorks real time operating system The system displays a message that says something like WDB Ready 33 e gearen Hieperlermenaat i que pa CH bane resp III 111 1111111 1311111311113 2111111331111331 11111228111113111111
31. eration of the system A user perspective This section describes the steps involved in cross development with the Vx Works Tornado system As explained above cross development involves two systems that work together in a co operative fashion In this project the host system was a personal computer running Windows 2000 and the target board was the ZPC860 board containing the MPC180 processor The two systems were connected to each other via a serial cable as well as over the network via an ethernet cable The board comes with a set of programs called the Board Support Package BSP The BSP is essentially the hardware specific part of VxWorks code The BSP per forms board specific functions like hardware initialization memory mapping and in terrupt handling 11 The BSP acts like an interface between the VxWorks kernel and the host based tools The final procedure used for cross development is detailed in the following section 3 9 Procedure 1 Turn on the FTP server supplied with Tornado and give access to the user with a username and password This username and password will be used later 31 r vxworie HyperTerminal Fis d Wew Cel Trasie Map Dae oS dm e wvxW orks Boot vxWorks Boot VxWorks Boot p boot devices cpm unit number 0 processor number 0 host name Suma file name wattorks inet on ethernet e 192 2G8 2 15 host inet h 192 168 2 15 user ui hi ftp pa
32. for tgtsvr exe 3 Core Files and Symbols This option allows the user to specify the location of the core file The core file is used by the target server to compute the symbol table for the target executable This option also allows you to synchronize the host and target symbol tables This will resolve any synchronization problems that may arise when the target server tries to connect to the target An example of a Core files and symbols error is The target system is not initialized yet Can t proceed 10 4 Memory Cache Size This option lets the user specify the cache size This can either be set to the default value 1 MB or a user defined value 5 Target Server File System This option allows the user to enable the use of the Target Server File System 6 Console and Redirection This option allows the output to be redi rected to the console 7 Logging This allows the creation of a log file 8 Miscellaneous This option allows the use of portmapper in order to ensure compatibility with Tornado 1 0 Target Name IP Address Specifies the target name or IP Address of the target T he target server uses this value to connect to the target over the network The target server can also be executed from the command line us ing tgtsvr exe The command used in this thesis was Command prompt tgtsvr exe 192 168 2 15 n 860 V B wdbrpc R c RW The options are briefly explained below
33. g where current encryption is halted and another stream is encrypted with a different key The original encryption is then resumed The context switching depends on the context keyword in the arguments This allows for switching between two different encryptions using different values for this variable 43 4 9 Public Key Hardware Accelerator PKHA Module The mpc180Pkha module performs advanced mathematical functions to support RSA and ECC public key cryptographic algorithms The PKHA module comprises of three sub modules The names of the sub modules are 1 mpc180PkhaUtil 2 mpc180PkhaEcc 3 mpcl80PkhaRsa The first of the sub modules the mpc180PkhaUtil executes the following functions 1 Load and read large number registers 2 Load and read the modulus size 3 Clear the register memory and perform supporting calculations The second sub module the mpc180PkhaEcc is used to perform elliptic curve cryp tography Elliptic Curve cryptography is supported in both F 2 n and F p The ECC module allows for two levels of operations The first is high level and executes elliptic curve point multiplication using either projective or affine co ordinates The second level allows for the execution of elliptic curve additions and point doubles These applications that run these operations are required to work with projective co ordinates with all the inputs and outputs being in the Montgomery residue system The third sub module the mpc180
34. hell that runs on the host The shell also supports C and includes a Tcl Tool command language interpreter The shell is very much like any UNIX shell in that it can be used to navigate call program routines and functions examine and modify memory The shell maintains a history of commands used and this can be accessed by pressing the Esc key and the J or K keys The shell can also be used to redirect input and output Browser The Browser is a visual monitoring system used to monitor the state of a target Some of its displays include Active task summaries CPU usage Memory usage and allocation Summary of modules linked to the run time system Semaphores watchdog timers and message queues 26 II Workspace Workspac E BE RB Workspace 2 inkSyme e 2 priConfig c E saline z ushpplna c sx 23 Erena Dependencses vi Deech Modules defaut FIGURE 9 Tornado s project management facility e VzrSim VxSim is a simulator program used to simulate the behavior of the target It is essentially VxWorks ported to the host system in order to simulate a target s behavior A screenshot of the simulator is shown in figure 8 e Project management facility Tornado comes with a complete project manage ment facility which one can use to customize the project structure and content The project facility gives the user a graphical method of working with VxWorks and the bui
35. hese advantages the MPC180 s board support package was written for use with VxWorks A detailed section on the procedures involved in actually running VxWorks on a board is presented in Chapter 3 2 0 Tornado Tornado is a development environment designed for use with VxWorks One can call it a suite of software and tools that are designed to interface with VxWorks and provide an efficient platform for cross development 2 2 1 Introduction to Tornado Tornado from WindRiver Systems is a development environment for VxWorks While VxWorks runs on the development board Tornado and it s tools run on the host computer communicating with VxWorks and facilitating exchange and analysis of data Tornado comprises of three main components They are 1 VxWorks 2 Tools for application development 16 Host Side i L D i Shell 8 i larret i Editor Debugger Server Project Browser Target Avent WindView Simulator m FIGURE 4 The host side of the Tornado environment 3 IDE Integrated Development Environment The Tornado Host Target interface serves as the connection between the host and target environments This is explained in detail later in this section The Tornado IDE is a comprehensive suite of software and tools used for real time application development The salient features of the IDE are as follows 8 e Integrated source code edi
36. host system was a personal computer pc running Windows TM 2000 and the target board was a PowerPC These boards were connected via a serial interface as well as through the network During application development the Tornado tools like WindShell and WindView would run on the host while VxWorks system tasks would run on the target This setup provides the development of the application with many benefits Some of these advantages are detailed below e Powerful resources memory processing power of the host system can be uti lized e Scaling and customization of the application are easily possible e Testing and debugging are taken care of by Tornado s tools running on the host Tornado ensures that there is minimal intrusion upon the target during application development by segregating the development facilities on the host Tornado also keeps 30 the run time facilities as independent as possible from the development environment thus ensuring as little a difference as possible between the target system during actual development and the target system after development The following section discusses the cross development experience in relation to this thesis from a user s perspective It seeks to explain the steps involved in going through the actual cross development cycle There are many details that need to be attended to during this process see Section 2 5 of 8 for a comprehensive treatise on this topic 3 2 Basic op
37. lar addition e 1024 bit RSA modular addition B A B mod N RSA Modular addition 1024 bit of Repetitions Time Repetition microseconds Error 625 4 8 196 625 5 1 196 625 5 1 1 625 4 8 1 625 4 8 1 TABLE 3 WindView timing for RSA modular addition e Routine used mpc180RsaModularAdd e From the results in table 3 the average time 4 92 microseconds 45 46 of Repetitions Time Repetition microseconds Error RSA 625 4 8 1 Modular 625 5 1 1 subtraction 500 4 9 196 1024 bit 625 5 1 196 625 5 1 196 TABLE 4 WindView timing for RSA modular subtraction RSA modular subtraction e 1024 bit RSA subtraction e Routine used mpc180RsaModularSubtract e From table 4 the average time taken to execute was 5 microseconds RSA modular multiplication e 1024 bit Montgomery multiplication e Routine used mpc180RsaModularMultiply e From table 5 the average time taken to execute was 103 7 33 196 microsecs Cross referencing the timing obtained above with the formula for running time of the MPC180 during multiplication Eu 1 F MS 10 MS 27 For our purposes F Frequency of operation 50MHz MS No of 16 bit blocks in the modulus 64 47 of Repetitions Time Repetition microseconds Error RSA 400 104 5 1 Modular 500 101 5 1 multiplication 500 104 5 1 1024 bit 500
38. ld environment There are three tabs in the project facility as shown in figure 9 1 Files This window allows the user to add delete files access the code of a program that is included in the project calculate dependencies in the code build the VxWorks images and projects and download the object files on to the target 2 VxWorks This tab shows the user all the components that have been included in the project like operating system components C C com ponents and network components In this tab users can include and ex 27 clude components and can resolve any potential conflicts with the project facility s help Users can also build VxWorks from this tab 3 Builds This tab shows the user the specifications of the build the project is using Double clicking on the build shown will bring up a dialog box that shows specifications about the ToolChain BSP Board Support Package C C compiler rules macros assembler linker link order and rules Everything but the ToolChain and the BSP settings can be modified in this dialog box e WindView WindView is a Tornado tool that provides the function of a run time analysis tool WindView monitors the target system and logs critical informa tion like timing events user defined or otherwise and task state transitions User defined events set with the wvEvent function can be referenced in the corresponding Wind View output in order to accurately determine timing
39. lgorithm This EU is also capable of generating a HMAC Hashed Message Authentication Code built upon the aforementioned hash functions The basic routine used in these calculations is mpc180Hash The arguments of this function are algorithm input buffer length of message output buffer The algorithm parameter specifies which algorithm this function must use to generate the hash and possible values are SHA 1 and MD5 The output buffer must be able to hold 16 bytes for MD5 and 20 bytes for SHA 1 e The Random Number Generator RNG is designed to produce 32 bit random numbers The random numbers are tested for compliance with the Federal Information Processing Standards FIPS 140 guidelines for randomness and non determinism The RNG is a valuable resource to the MPC180 because of the fact that it is highly desirable to have a physically private random number generator in order to keep secret keys confidential These keys can even be hidden from higher level application code 10 2 The development environment and tools VxWorks and Tornado 2 1 VxWorks 2 1 1 Introduction to VxWorks Real time operating systems are developed to perform a specialized set of tasks and have a strict set of requirements for their operating systems 6 VxWorks is one of today s most widely implemented real time operating systems VxWorks can make scheduling guarantees that normal operating systems cannot VxWorks and Tornado a set of tools use
40. ons with an acceptable percentage of error Both the above methods were used to obtain timing 4 1 ARC Four Hardware Accelerator AFHA Module The AFHA module performs encryption and decryption compatible with RSA s RCA stream cipher where a byte of plaintext is encrypted with a key to produce a byte of ciphertext The AFHA module supports key ranges from 40 bits to 128 bits The mpc180Afha module uses two routines to perform encryption decryption e mpcl80AfhaSetKey This is used to set the key used for the encryp tion decryption RC4 Key length bytes Data length bytes Time microseconds Encryption 5 54 736 255 13 1520 1033 952 Decryption 5 54 664 02 13 1520 1092 427 Al TABLE 1 WindView timing for RC4 e mpcl80AfhaProcess This routine is used to perform the RC4 encryp tion decryption The mpc180Afha module is very versatile in that it is able to easily perform encryp tion context switching This means that the module is capable of halting a current execution and allowing the encryption of another message with a new key This feature is very useful for applications where a subsequent encryption has a higher priority and needs to be processed immediately However the time overhead involved in context switching is very large due to the large amounts of data that must be transferred to and from the mpc180 in order to process the context switch Fig ure 16 shows an output
41. or RSA modular addition 45 4 WindView timing for RSA modular subtraction 46 5 WindView timing for RSA modular multiplication 47 6 WindView timing for ECC addition in F p 40 7 WindView timing for ECC addition in F 2n 40 8 WindView timing for ECC point doubling in F p 50 9 WindView timing for ECC point doubling in F 2 n 51 10 WindView timing for 32 bit RNG generation 51 11 WindView timing for 100 blocks of 32 bit RNG generation 52 12 WindView timing for SHA 1 53 13 WindView timing for MIR 54 14 Run Time Formulae s 55 HARDWARE ENCRYPTION USING THE MPC180 SECURITY CO PROCESSOR 1 Introduction 1 1 Motivation 1 1 1 Cryptography The security and integrity of data that is sent over today s networks is of paramount importance Applications and services that use networks to send and receive data range from banks to stock market data to military applications and to common e mail If proper precautions are not taken every bit of data that is sent over the internet is susceptible to malicious activity The art and science of cryptography has been around for many hundreds of years and is being used today to provide a layer
42. or VxWorks and click OK e This brings up the application creation wizard Here the user is asked to name the project specify the location where the project is to be stored describe the 35 project and specify which workspace this project should be added to This is shown in figure 14 e The next screen figure 15 will ask the user to specify a Toolchain that should be used to build the application In our case we choose PPC860gnu As the name suggests the PPC860 refers to the target architecture while the onu part refers to the compiler used e The next screen simply asks the user to confirm the choices he she has made Upon clicking OK a workspace opens up where the user can add files and compile the project etc project workspace window is shown in figure 9 e The C C source files are added to the project with the Project Management Facility using the context menu and selecting Add Files e Once VxWorks has been configured properly on the system the target server needs to be started by going to Tools Target server name Please note that the target server can be configured using the configure option in the above menu e Launch the Tornado Shell by clicking on the i button in the Tornado tool bar e The following steps are used to build the application code in order for it to be downloaded onto the target for execution 1 Choose the Builds tab in th
43. quest Controller OUTPUT AK bi Da ripa Lok FIGURE 1 Architecture of the MPC180 security co processor External B DMA Request The MPC180 interfaces to Motorola s MPC8xx system bus or the MPC8260 s local bus For the purposes of this thesis the MPC180 was interfaced to the ZPC860 s system bus through the External Bus Interface EBI which performs the following functions e Translates the processor core bus timing into a simple read write interface for the execution units e Decodes the memory address in order to select the appropriate execution unit The EBI contains two First In First Out FIFOs registers one input and one output of size 4096 bits each which are used to maximize throughput and make it easier on the host processor s data management responsibilities The MPC180 manages data transfer between the FIFOs and the execution units EU with no intervention necessary on the host processor s part The MPC180 operates in one of two modes e Open address mode where the host processor can write directly to the MPC180 address map 6 e FIFO mode where the MPC180 accepts data transfers upto 4 kilobits to and from the FIFOs There are five execution units on the MPC180 e PKEU Public Key Execution Unit which handles the RSA and ECC compu tations e DEU DES Execution Unit which handles DES and 3DES computations in both CBC and ECB modes
44. rallel and ethernet ports SCSI and IDE controllers etc VxWorks supports a large number of boards and thus provides an easy development process Virtual Memory VxWorks includes virtual memory management for boards that require this feature Shared Memory Objects In the case of a multi processing environment Vx works has features that enable it to share resources like memory and semaphores between different processors The VxWorks Simulator The VxWorks simulator VxSim simulates a VxWorks target in order to test the software Tornado s tools are designed to work closely with VxSim in order to analyze and test the software effectively Utility libraries The following are the utility libraries provided with VxWorks 1 Interrupt handling 2 Watchdog timers 3 Message logging 4 Memory allocation 5 String formatting and scanning 6 ANSI C libraries 15 2 1 8 Why VxWorks There are other real time operating systems available from various companies like QNX LynxOS Windows CE Integrity and ThreadX But it so happens that while certain companies products like Green Hills RTOS products have been gaining a lot of popularity VxWorks probably is still the most widely used real time operating system RTOS in the embedded industry VxWorks has built a good reputation for itself owing to its flexibility compatibility and scalability Furthermore WindRiver Systems provides excellent technical support Due to t
45. rate the hash and possible values are SHA 1 and MD5 The output buffer must be able to hold 16 bytes for MD5 and 20 bytes for SHA 1 SHA 1 of Repetitions Time Repetition microseconds Error 400 3331 1 400 2998 1 400 3748 1 400 2831 1 400 2581 1 TABLE 12 WindView timing for SHA 1 SHA 1 hashing e Routine used to perform SHA 1 hashing is mpcl80Hash MDHA_SHA1 53 e Average time to execute SHA 1 for a data length of 241 bytes is 3097 8 41 196 microsecs MD5 hashing e Routine used to perform MD5 hashing is mpc180Hash MDHA MD5 e Average time to execute MD5 for a data length of 131 bytes is 2781 41 1 microsecs MD5 of Repetitions Time Repetition microseconds Error 500 2531 196 400 3414 1 400 2706 1 400 2456 1 400 2789 1 TABLE 13 WindView timing for MD5 54 The timing formulae for the PKHA embedded routines provided in the user s man ual are shown in table 14 3 The following definitions apply to the symbols in the table F operating frequency MS number of 16 bit blocks in the modulus Ne number of bits in the exponent or multiplier k avg average run time assumes half of the bits in the exponent are 1 5 wcs worst case run time bes best case run time 55 Run Time Formula Operation Symbol multPtoQ
46. ration namely CBC or Cipher Block Chaining mode and ECB or Electronic Code Book mode e Arc Four Hardware Accelerator AFHA Module The AFHA module performs symmetric encryption and decryption compatible with RSA s RC4 stream cipher where a byte of plaintext is encrypted with a key to produce a byte of ciphertext The AFHA module supports key ranges from 40 bits to 128 bits The mpc180Afha module uses two routines to perform encryption decryption mpcl80AfhaSetKey This is used to set the key used for the encryp tion decryption mpcl80AfhaProcess This routine is used to perform the RC4 encryp tion decryption The mpc180Afha module is very versatile it is able to easily perform encryption context switching This means that the module is capable of halting a current execution and allowing the encryption of another message with a new key This feature is very useful for applications where a subsequent encryption has a higher priority and needs to be processed immediately e Message Digest Hardware Accelerator MDHA Module The mpc180Mdha message digest hardware accelerator module is capable of performing hash algorithms like the SHA 1 and MD5 It is also capable of generating an HMAC which can be built upon an existing message digest cal culation This unit takes 16 32 bit words as input runs them through the specific algorithm and produces a hashed message of 160 bits for SHA 1 or 128 bits for the MD4 MD65 a
47. ssword pw gt boardi flags f oOx targat name tn esddaniela vxworkg Boot FIGURE 11 Hyperterminal startup screen by the board to log into the host computer and download the vxWorks image that it needs to boot from the host The board is connected to the host computer via two interfaces e serial port connection and e An ethernet connection using a router or any standard networking device Upon powering up the target the user must start Hyperterminal a version of Wind River s hyperterminal is provided with Tornado and connect to the target using either COM1 or COM2 The terminal window is shown in figure 11 The user is given an 8 second time interval to interrupt the board from loading the VxWorks image from the location specified in the configuration Since the board comes with a pre set configuration it must be changed to reflect current 32 parameters To boot up the board uses pre defined settings These settings can be displayed in the hyperterminal window by typing p at the VxWorks boot prompt One can change these settings by typing in c at the prompt The common parameters are e file This setting is very important because the board searches for the VxWorks file on the host system at the location specified by this parameter If the file is not present there the board will display an error and fail to boot for example C Tornado farget onfig rhpc
48. ted network security 58 5 9 Future work Future directions of work with the MPC180 security co processor could aim to wards application development turning it into a cryptographic box so to speak 7 This system would function as a very secure server peripheral providing dedicated cryptographic support for all network traffic Similar to nCipher s nShield products 18 10 11 12 13 14 15 16 1T 59 BIBLIOGRAPHY Michael Barr Programming Embedded Systems in C and C O Reilly first edition January 1999 Jorge Guajardo Thomas Wollinger and Christof Paar Cryptography in em bedded systems An overview 2003 Motorola Semiconductors Inc MPC180LMB Security processor user s manual rev 1 3 2002 Micetek International Inc mpc860 processor card user s manual October 2000 L Adleman R L Rivest and A Shamir A method for obtaining digital signature and public key cryptosystems Comm of the ACM vol 21 no 2 pp 120 126 February 1978 S R Ball Embedded Microprocessor Systems Real World Design Newnes third edition November 2002 WindRiver Systems VxWorks 5 4 Programmers Guide First edition March 1999 WindRiver Systems Tornado 2 2 Users Guide Windows Edition August 2002 John Gordon The vxworks cook book 2003 J A Borkhuis VxWorks Tornado II FAQ 2003 WindRiver Systems Tornado 2 0 Getting Started Gui
49. that the target agent can execute before the kernel is running This feature is valuable for the early stages of porting VxWorks to a new target platform A key function of the agent is to service the requests of the host resident object module loader If the agent is linked into the run time and stored in ROM the target server automatically initializes the symbol table from the host resident image of the target run time system as it starts From this point on all downloads are incremental in nature greatly reducing download time The agent itself is scalable you can choose what features to include or exclude This permits the creation of final production configurations that still allow field testing even when very little memory can be dedicated to activities beyond the applications purpose Target Server The target server is a component of Tornado that runs on the host computer and its function is to connect Tornado s tools to the target agent The target agent is a very condensed collection of services which run on the target while providing information to Tornado s tools There is one target server that runs on the host computer for every target that the host is attached to The tools that run on the host request information from the target via the target server T he target server communicates with the target agent and requests from it information that was solicited by the tools The target server also manages the connection between the
50. tic point doubling was carried out and the timing was obtained using the timex routine ECC point doubling in F p e Point doubling in F p e From table 8 the average time to execute 1381 26 1 microsecs ECC point doubling in F 2 n e Point Doubling in F 2 m e From table 9 the average time to execute 1382 2 26 196 microsecs For elliptic curve point multiplication the routine used is mpc180EccMultPtoQ 4 9 Random Number Generator RNG Module The mpc180Rng module performs random number generation and provides ap plications with 32 bit numbers The mpcl80Rng has two interfaces wherein one interface generates a single 32 bit random number mpcl80Rand the other gen erates a block of random numbers of a specified length mpcl80RandLongs The timing includes testing for randomness ECC point doubling F 2n of Repetitions Time Repetition microseconds Error 625 2211 196 625 1624 1 625 1678 1 625 1784 1 625 1864 1 TABLE 9 WindView timing for ECC point doubling in F 2 n 32 bit RNG of Repetitions Time Repetition microseconds Error 175 5236 1 175 5331 1 175 5236 1 175 4283 1 175 5045 1 TABLE 10 WindView timing for 32 bit RNG generation 32 bit random number generation e Routine used to generate one 32 bit rng is mpcl80rand 51 e From table 10 the time required to generat
51. tmurfp avg Ne tapifp 0 5 Ne tadafp 8 tutti 6 MS move FpAddPtoQ taddfp 16 tutti 4 tada 5 tsub 19 MS move FpDoubleQ tablfp 10 to utt1 11 tada 2 tsup 10 MS move multPtoQ tmul f2m avg Ne L n r 0 5 Ne tadaf2m 8 tmult1 6 MS move F2maddPtoQ taddf2m 20 tmult1 7 tada 15 MS move F2mdoubleQ tablf2m 10 tmuit1 4 tada 9 MS move expA tezp avg 1 5 Ne t uiti tmutt1 wes modularmultiply tmuiti wes tmulti bes 1 F IMS 10 MS 27 1 F MS 9 MS 22 modularmultiply2 tmult2 wes tmylt2 bes 2 tmault1 wes 2 tu uiii bcs modularadd tadaa wes taqq bes 1 F 4 MS 11 1 F 3 M S 6 modularsub tsub wes tut bes 1 F 3 MS 11 1 F 2 M S 6 r2 tra tbd clearmemory 1 F A MS 5 TABLE 14 Run Time Formulae 56 5 Discussion 5 1 Analysis This chapter discusses the results gathered during the course of this thesis The MPC180 was tested for timing and performance by running code for today s currently used cryptographic algorithms The following sections will present the conclusions and discuss future directions of research in embedded cryptography As mentioned before Tornado was used to write code and time various routines supported by the MPC180 The BSP Board Support Package supplied with the MPC180 was written for use with Tornado
52. tor e Integrated C C compilers Most commonly GNU s gcc compiler 9 17 e Cross Wind a GUI based debugger e WindSh a C based command shell e Browser which is a visual monitoring system e VzSim a simulator program used to simulate the behavior of the target e Project management facility e WindView a timing and analysis tool software and logic analyzer An in depth explanation of these tools is given in section 2 2 2 Figures 4 and 5 adapted from 8 show the Tornado Development Environment s structure Figure 4 shows the host side of the environment while figure 5 shows the target side of the environment Tornado s File Structure Tornado Root Directory docs Online documentation host Location of the host resident tools SETUP Tornado setup program share Protocol definitions shared by host and target software target config all System configuration information bsp name files related to a particular board h Header files lib Processor independent libraries and modules src C source code files 18 config Source for VxWorks supplied modules demo WindRiver supplied demos drv Device drivers usr User changeable code wind Files used to customize Tornado tools Tornado s file structure on the host computer is shown above in order to give the reader a general idea of how Tornado s files are organized The Tornado Host Target interface is the connection between Tornado s tools
53. version 2 02 in mind At the lab however the computers were equipped with Tornado 2 2 This initially presented us with many problems and compatibility issues such as malfunctioning target servers and project creation errors So I had to go back to using version 2 02 which took care of many of the problems After a few modifications to the configuration I managed to get the board to boot To ensure proper migration when and where possible WindRiver Systems provides a Migration Guide that is a useful reference to have The results obtained in the previous chapter are indicative of the performance of the MPC180 security processor when tested on a Pentium processor running at 1 7 Ghz Upon comparison with current software benchmarks 13 the performance of the MPC180 appears to be slower But it must be taken into account that today s standard benchmarks for cryptographic libraries in C called Crypto4 4 5 1 13 were tested after various optimizations for speed and Pentium 4 code generation and were tested on a Pentium 4 2 1 GHz machine running Windows XP D The timing of the Crypto 5 1 13 libraries for 1024 bit RSA encryption is 0 18 milliseconds while the MPC180 takes around 160 milliseconds for a 1024 bit RSA exponentia tion There are other hardware solutions for cryptography being developed As an These libraries were compiled with Microsoft VC NET 57 example SCI Worx s high performance RSA co processor performs a
54. with sensitive timing requirements The basic requirements and capabilities of VxWorks are summarized below e Real time operating systems should always respond to an event in a guaranteed amount of time which is in the order of microseconds or nanoseconds This is the essence of real time e Real time operating systems are capable of working with a minimal set of re sources like memory e real time operating system is rugged and will be resilient to situations that would otherwise cause a non real time operating system to crash Figure 2 shows how a real time operating system fits into the scheme of things with the development tools running on a host computer and VxWorks running on the development board The real time operating system will run on the hardware target while the develop ment tools run on the host computer The network functions as the interconnection between the host and the target It provides facilities like file transfers between the host and the target and has the ability to boot the board over the network The 12 I O System Multitasking Network File Sy siem Kernel stack Device Drivers FIGURE 3 The real time operating system kernel development tools communicate with the real time operating system and gather in formation that can be analyzed Figure 3 shows the functional structure of a real time operating system It comprises of a multitasking kernel which performs functions like

View/Open - Oregon State University

Contents

Download Pdf Manuals

Related Search

Related Contents