DOCSLIB.ORG

Autolock: Why Cache Attacks on ARM Are Harder Than You Think

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Autolock: Why Cache Attacks on ARM Are Harder Than You Think AutoLock: Why Cache Attacks on ARM Are Harder Than You Think Marc GreenW, Leandro Rodrigues-LimaF, Andreas ZanklF, Gorka IrazoquiW, Johann HeyszlF, and Thomas EisenbarthW August 18th 2017 – USENIX Security Symposium The Big Picture Cache Attacks AutoLock ————————- ————————- Transition of attacks: −! Dedicated countermeasures desktop & server ! mobile are still necessary for protection Vulnerabililty and risk assessment −! Eviction-based attacks are are important, but challenging harder than previously believed Limited understanding of −! Undocumented performance commercial microarchitectures feature of inclusive caches on ARM AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 2 Cache Basics AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 3 Cache Basics Hierarchy AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 4 Cache Basics Hierarchy AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 5 Cache Basics Hierarchy AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 6 Cache Basics Hierarchy AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 7 Cache Basics Hierarchy AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 8 Cache Basics Hierarchy AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 9 Cache Basics Inclusiveness AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 10 Cache Basics Inclusiveness AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 11 Cache Basics Inclusiveness AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 12 Cache Basics Inclusiveness AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 13 Cache Basics Inclusiveness AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 14 Cache Basics Inclusiveness AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 15 Cache Basics Inclusiveness AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 16 Cache Basics Inclusiveness AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 17 Cache Basics Inclusiveness AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 18 Cache Attacks AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 19 Cache Attacks Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 20 Cache Attacks Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 21 Cache Attacks Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 22 Cache Attacks Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 23 Cache Attacks Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 24 Cache Attacks Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 25 Cache Attacks Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 26 Cache Attacks Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 27 Cache Attacks Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 28 Cache Attacks Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 29 Cache Attacks Cross-core Eviction 1.0 without eviction with eviction 0.8 0.6 0.4 Normalized frequency 0.2 0.0 400 600 800 1000 1200 1400 Clock cycles Qualcomm Krait 450 AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 30 Cache Attacks Literature Method Task Reference Evict + Time Eviction [OST06] Prime + Probe Eviction [OST06] Evict + Reload Eviction [GSM15] Evict + Prefetch Eviction [GMF+16] Flush + Reload Flush [YF14] Flush + Prefetch Flush [GMF+16] Flush + Flush Flush [GMWM16] AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 31 Cache Attacks ARM Processors Flush on ARM Eviction on ARM Easy, fast, and robust Complex, slow, and error-prone Only from ARMv8 onwards All ARM architectures: v6, v7, v8 No guaranteed userspace access Userspace privileges are sufficient Limited number of devices Larger number of devices AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 32 AutoLock AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 33 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 34 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 35 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 36 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 37 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 38 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 39 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 40 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 41 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 42 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 43 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 44 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 45 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 46 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 47 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 48 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 49 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 50 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 51 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 52 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 53 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 54 AutoLock Cross-core Eviction AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 55 AutoLock Definition A patented and undocumented performance feature of inclusive cache levels that transparently prevents the eviction of cache lines, if they are contained in higher cache levels. Automatic + Lockdown = “AutoLock” AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 56 Implications of AutoLock AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al. j USENIX Security 2017 j 57 Implications Impact in Theory Method Task Same-core Cross-core Evict + Time Eviction 37 Prime + Probe Eviction 37 Evict + Reload Eviction 37 Evict + Prefetch Eviction 37 Flush + * Flush 3 3 AutoLock: Why Cache Attacks on ARM Are Harder Than You Think j Green et al.
Load more

Recommended publications
  • NVIDIA Tegra 4 Family CPU Architecture 4-PLUS-1 Quad Core
    Whitepaper NVIDIA Tegra 4 Family CPU Architecture 4-PLUS-1 Quad core 1 Table of Contents ...................................................................................................................................................................... 1 Introduction .............................................................................................................................................. 3 NVIDIA Tegra 4 Family of Mobile Processors ............................................................................................ 3 Benchmarking CPU Performance .............................................................................................................. 4 Tegra 4 Family CPUs Architected for High Performance and Power Efficiency ......................................... 6 Wider Issue Execution Units for Higher Throughput ............................................................................ 6 Better Memory Level Parallelism from a Larger Instruction Window for Out-of-Order Execution ...... 7 Fast Load-To-Use Logic allows larger L1 Data Cache ............................................................................. 8 Enhanced branch prediction for higher efficiency .............................................................................. 10 Advanced Prefetcher for higher MLP and lower latency .................................................................... 10 Large Unified L2 Cache .......................................................................................................................
    [Show full text]
  • Hardware-Assisted Rootkits: Abusing Performance Counters on the ARM and X86 Architectures
    Hardware-Assisted Rootkits: Abusing Performance Counters on the ARM and x86 Architectures Matt Spisak Endgame, Inc. mspisak@endgame.com Abstract the OS. With KPP in place, attackers are often forced to move malicious code to less privileged user-mode, to ele- In this paper, a novel hardware-assisted rootkit is intro- vate privileges enabling a hypervisor or TrustZone based duced, which leverages the performance monitoring unit rootkit, or to become more creative in their approach to (PMU) of a CPU. By configuring hardware performance achieving a kernel mode rootkit. counters to count specific architectural events, this re- Early advances in rootkit design focused on low-level search effort proves it is possible to transparently trap hooks to system calls and interrupts within the kernel. system calls and other interrupts driven entirely by the With the introduction of hardware virtualization exten- PMU. This offers an attacker the opportunity to redirect sions, hypervisor based rootkits became a popular area control flow to malicious code without requiring modifi- of study allowing malicious code to run underneath a cations to a kernel image. guest operating system [4, 5]. Another class of OS ag- The approach is demonstrated as a kernel-mode nostic rootkits also emerged that run in System Manage- rootkit on both the ARM and Intel x86-64 architectures ment Mode (SMM) on x86 [6] or within ARM Trust- that is capable of intercepting system calls while evad- Zone [7]. The latter two categories, which leverage vir- ing current kernel patch protection implementations such tualization extensions, SMM on x86, and security exten- as PatchGuard.
    [Show full text]
  • A Tour of the ARM Architecture and Its Linux Support
    Linux Conf Australia 2017 A tour of the ARM architecture and its Linux support Thomas Petazzoni Bootlin thomas.petazzoni@bootlin.com - Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 1/1 I Since 2012: Linux kernel contributor, adding support for Marvell ARM processors I Core contributor to the Buildroot project, an embedded Linux build system I From Toulouse, France Thomas Petazzoni I Thomas Petazzoni I CTO and Embedded Linux engineer at Bootlin I Embedded Linux expertise I Development, consulting and training I Strong open-source focus I Linux kernel contributors, ARM SoC support, kernel maintainers embedded Linux and kernel engineering - Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 2/1 I Core contributor to the Buildroot project, an embedded Linux build system I From Toulouse, France Thomas Petazzoni I Thomas Petazzoni I CTO and Embedded Linux engineer at Bootlin I Embedded Linux expertise I Development, consulting and training I Strong open-source focus I Linux kernel contributors, ARM SoC support, kernel maintainers I Since 2012: Linux kernel contributor, adding support for Marvell ARM processors - Kernel, drivers and embedded Linux - Development, consulting, training and support - https://bootlin.com 2/1 I From Toulouse, France Thomas Petazzoni I Thomas Petazzoni I CTO and Embedded Linux engineer at Bootlin I Embedded Linux expertise I Development, consulting and training I Strong open-source focus I Linux kernel contributors,
    [Show full text]
  • Modem Product Portfolio
    Blogger Workshop Benchmark Summary: MDP Tablet Featuring APQ8064 Quad-Core © 2012 QUALCOMM Incorporated. All rights reserved. 1 Executive Overview . On July 24th, 2012, 30 bloggers (15 International, 15 U.S) descended into San Francisco, CA to take part in a blogger’s workshop where they spent unsupervised time benchmarking and testing our latest tablet MDP equipped with the APQ8064, otherwise known as the Snapdragon S4 Pro. The tone of coverage overall was extremely positive. Bloggers who attended the workshop raved about the performance of the quad-core devices and the potential benefits for developers. The graphics garnered particular acclaim, especially in comparison to competitor’s quad-core devices. The articles from the workshop attendees spurred a second wave of global coverage, especially in China, that took note of their reviews as well as the availability of the MDP. Below is a compilation of benchmark data pulled from various third party sources, including Anandtech, TheVerge, SlashGear, and Android Community. This benchmark data reflects performance across multiple vectors in the Snapdragon S4 Pro, including CPU, GPU and browsing. In the same deck, commercial devices featuring our dual core Snapdragon S4 processor has been updated to reflect data from the Samsung Galaxy S3 (MSM8960). © 2012 QUALCOMM Incorporated. All rights reserved. 2 Snapdragon S4 Pro APQ8064 MDP HEADLINE: “Qualcomm Snapdragon S4 Pro quad-core developer tablet shows off stunning benchmark performance” “It's fast, handily eclipsing the Samsung Galaxy SIII’s Exynos 4 Quad, the HTC One X’s Tegra 3, and Qualcomm's own Snapdragon S4 on a spate of synthetic benchmarks.” – Nate Ralph, TheVerge Source: TheVerge by Nate Ralph 7/24/2012 http://www.theverge.com/2012/7/24/3185016/qualcomm-snapdragon-s4-pro-APQ8064 © 2012 QUALCOMM Incorporated.
    [Show full text]
  • ARM NEON Assembly Optimization Dae-Hwan Kim Department of Computer and Information, Suwon Science College, 288 Seja-Ro, Jeongnam-Myun, Hwaseong-Si, Gyeonggi-Do, Rep
    Journal of Multidisciplinary Engineering Science and Technology (JMEST) ISSN: 2458-9403 Vol. 3 Issue 4, April - 2016 ARM NEON Assembly Optimization Dae-Hwan Kim Department of Computer and Information, Suwon Science College, 288 Seja-ro, Jeongnam-myun, Hwaseong-si, Gyeonggi-do, Rep. of Korea kimdh@ssc.ac.kr Abstract—ARM is one of the most widely used The ARMv8 architecture [3, 4] introduces a 64-bit 32-bit processors, which is embedded in architecture, named AArch64, and a new A64 smartphones, tablets, and various electronic instruction set to the existing instruction set to support devices. The ARM NEON is the SIMD engine inside the 64-bit operation and the virtual addressing. ARM core which accelerates multimedia and In this paper, various assembly software signal processing algorithms. NEON is widely optimization techniques are proposed for the ARM incorporated in the recent ARM processors for NEON architecture. Practical example code is given to smartphones and tablets. In this paper, various the optimization technique, whose performance is assembly level software optimizations are analyzed, compared to the original code. provided such as instruction scheduling, instruction selection, and loop unrolling for the The rest of this paper is organized as follows. NEON architecture. The proposed techniques are Section II shows the overview of the assembly expected to be applied directly in the software optimizations, and Section III presents each technique development for the ARM NEON processors. in detail with an example. Conclusions are presented in Section IV. Keywords—ARM; NEON; embedded processor; software optimization; assembly II. ASSEMBLY OPTIMIZATION OVERVIEW NEON is the SIMD (Single Instruction Multiple Data) I.
    [Show full text]
  • 1 in the United States District Court for the Northern
    Case: 1:18-cv-06255 Document #: 1 Filed: 09/13/18 Page 1 of 19 PageID #:1 IN THE UNITED STATES DISTRICT COURT FOR THE NORTHERN DISTRICT OF ILLINOIS EASTERN DIVISION COMPLEX MEMORY, LLC, Plaintiff Civil Action No.: 1:18-cv-6255 v. PATENT CASE MOTOROLA MOBILITY LLC Jury Trial Demanded Defendant COMPLAINT FOR PATENT INFRINGEMENT Plaintiff Complex Memory, LLC (“Complex Memory”), by way of this Complaint against Defendant Motorola Mobility LLC (“Motorola” or “Defendant”), alleges as follows: PARTIES 1. Plaintiff Complex Memory is a limited liability company organized and existing under the laws of the State of Texas, having its principal place of business at 17330 Preston Road, Suite 200D, Dallas, Texas 75252. 2. On information and belief, Defendant Motorola is a Delaware limited liability company headquartered at 222 W. Merchandise Mart Plaza, Chicago, IL 60654. JURISDICTION AND VENUE 3. This is an action under the patent laws of the United States, 35 U.S.C. §§ 1, et seq., for infringement by Motorola of claims of U.S. Patent Nos. 5,890,195; 5,963,481; 6,658,576; 6,968,469; and 7,730,330 (“the Patents-in-Suit”). 4. This Court has subject matter jurisdiction pursuant to 28 U.S.C. §§ 1331 and 1338(a). 5. Motorola is subject to personal jurisdiction of this Court because, inter alia, on 1 Case: 1:18-cv-06255 Document #: 1 Filed: 09/13/18 Page 2 of 19 PageID #:2 information and belief, (i) Motorola is registered to transact business in the State of Illinois; (ii) Motorola conducts business in the State of Illinois and maintains a facility and employees within the State of Illinois; and (iii) Motorola has committed and continues to commit acts of patent infringement in the State of Illinois, including by making, using, offering to sell, and/or selling accused products and services in the State of Illinois, and/or importing accused products and services into the State of Illinois.
    [Show full text]
  • Cache Attacks on ARM
    Moritz Lipp Cache Attacks on ARM Master thesis Graz, University Of Technology NOTICE If you want to cite this work based on last-level cache attacks on ARM, please cite the following research paper instead: •L ipp, Moritz ; Gruss, Daniel ; Spreitzer, Raphael ; Maurice, Clementine´ ;Mangard, Stefan: ARMageddon: Cache Attacks on Mobile De- vices. In: 25th USENIX Security Symposium (USENIX Security 16). Austin, TX : USENIX Association, August 2016. – ISBN 978–1–931971– 32–4, 549–564 If you want to cite this work in respect to rowhammer on ARM-based devices, please cite the following publication instead: •V een, Victor van d. ; Fratantonio, Yanick ; Lindorfer, Martina ; Gruss, Daniel ; Maurice, Clementine´ ; Vigna, Giovanni ; Bos, Her- bert ; Razavi, Kaveh ; Giuffrida, Christiano: Drammer : Determinis- tic Rowhammer Attacks on Commodity Mobile Platforms. In: ACM Conference on Computer and Communications Security – CCS, 2016 Contents 1 Introduction 1 1.1 Motivation . .3 1.2 Key Challenges and Results . .4 1.3 Contributions . .5 1.4 Test devices . .6 1.5 Outline . .7 2 Background 9 2.1 CPU caches . .9 2.2 Cache coherence . 18 2.3 Shared memory . 29 2.4 Cache Attacks . 31 2.5 DRAM . 39 3 Attack primitives 43 3.1 Timing Measurements . 43 3.2 Cache Eviction . 47 3.3 Defeating the Cache-Organization . 53 4 Attack Case Studies 57 4.1 High Performance Covert-Channels . 58 4.2 Spying on User input . 62 4.3 Attacks on Cryptographic Algorithms . 69 4.4 Rowhammer on ARM . 75 5 Countermeasures 79 6 Conclusion 81 i List of tables . 83 List of figures . 86 List of acronyms .
    [Show full text]
  • Huawei's Mobile Processors
    Huawei's mobile processors Dezső Sima Vers. 2.1 November 2018 Sima Dezső, 2018 Huawei's mobile processors • 1. Overview • 2. Examples: The Kirin 950 • 2.1 The Kirin 950 • 2.2 The Kirin 960 • 2.3 The Kirin 970 • 2.4 The Kirin 980 • 3. References 1. Overview 1. Overview (1) Huawei [1] • Huawei Technologies Co. Ltd. is a Chinese multinational networking and telecommunications equipment and services company headquartered in Shenzhen China. • It was founded in 1987 and became the world's largest telecom equipment manufacturer. • The name Huawei (华为) means Chinese achievement. • Recently it has about 180 000 employees. Figure: Huawei's logo [1] Figure: Huawei's headquarter in Shenzen [1] 1. Overview (2) HiSilicon [2] • HiSilicon, a global fabless semiconductor and IC design company, headquartered in Shenzhen, China. • It was founded in 2004 as a subsidiary of Huawei. • It has offices and research centers worldwide, the number of the employees is about 7000. • HiSilicon purchases licenses mainly from ARM and designs in the first line application processors for mobiles. It filed over 5,000 patents. 1. Overview (3) 1. Overview Worldwide smartphone shipments by vendors Q4/2009 to Q4/2016 (million units) [3] 1. Overview (4) Worldwide market share of application processors in 2015 used in smartphones (based on revenue) [4] Market Processor lines Vendor Cores ISA share (examples) Qualcomm Snapdragon Qualcomm designed Krait cores ARMv7 42 % (USA) 200-800 ARM Cortex A line ARMv7/v8 Apple Apple A7-A9 Apple designed Cyclone core 21 % ARMv8 (USA) Apple A10 2xbig./4x LITTLE cores 8xARM Cortex A53 ARMv7 MediaTek Helio x10 (ARM big.LITTLE) 19 % (Taiwan) Helio X20 2xARM Cortex A72/8x A53 ARMv8 (ARM big.LITTLE) Samsung Exynos ARM Cortex A line ARMv7 (S.
    [Show full text]
  • Sig Nature Redacted
    Platform Ecosystem: ARM's answer to Intel's dominance by Tanmoy Sen B.E. Electronics and Communication Engineering (2001) Delhi College of Engineering, University of Delhi Submitted to the System Design and Management Program in Partial Fulfillment of the Requirements for the Degree of Master of Science in Engineering and Management at the Massachusetts Institute of Technology Jan 2014 [FC- -Zo"t'J C 2014 Tanmoy Sen All rights reserved The author hereby grants to MIT permission to reproduce and to distribute publicly paper and electronic copies of this thesis document in whole or in part in any medium now known or hereafter created. __Signature redacted Signature of Author Tanmoy Sen System Design and Management Program Jan 17th, 2014 Certified by Signature redacted- 'I V Michael Cusumano Thesis Supervisor SMR Distinguished Pr essor of Management & Engineering Systems MIT S Schoo of Unaement ___Sig nature Accepted by redacted -Patrick Hale Director MASSACHUSETTS INSTITUTE System Design and Management Program OF TECH NOLOGY JUN 2 3 2016 LIBRARIES ARCHIVES SDM Thesis Platform Ecosystem: ARM's Answer to Intel's Dominance Abstract The personal computing industry has witnessed significant changes with more users moving from desktop PCs to battery-operated mobile devices. These dynamics have prompted chip-design companies to evaluate ways to lower the power consumption of devices, thereby elongating battery life. With its lower power microprocessor-core architecture, a newer and smaller company, ARM Holdings, has been able to challenge the much bigger incumbent Intel and capture significant market share by creating a powerful ecosystem based on strategic partnerships. This thesis will evaluate how ARM's 'design and license' business model based on a platform ecosystem- partnerships with Original Equipment Manufacturers (OEMs), semiconductor companies, and software developers, has been able to counter Intel's vertically-integrated business model.
    [Show full text]
  • A Study on Hardware Trojan Detection and Taxonomy
    A STUDY ON HARDWARE TROJAN DETECTION AND TAXONOMY Abhishek Gaikwad* Electrical Engineering California State University, Fullerton CA 92831, USA E-mail: abgaikwad123@csu.fullerton.edu *Corresponding Author Sunil Kumar Anasuri Computer Engineering California State University, Fullerton CA 92831, USA E-mail: Sunil2308@csu.fullerton.edu Yoonsuk Choi Computer Engineering California State University, Fullerton CA 92831,USA E-mail: yochoi@fullerton.edu ABSTRACT Hardware Trojan (HT) is a malicious modification of the circuitry of an integrated circuit. Hardware Trojan is completely characterized by its physical representation and its behavior. Thus there should be some means of Trojan detection and timely elimination of such threats in the designed circuitry. Through this research, we analyzed detail taxonomy of Trojans. Then perform side channel analysis on the performance of the designed system before and after introduction of Trojans in the programming logic embedded in the Integrated circuit. Then the degree of corruption will be analyzed by comparing the bandwidth of side channel of the signals obtained before and after introduction of Trojans. Also protocols will be defined to eliminate such introduced Trojans and to define methods to recover from potential Trojan attacks. Also a case study about Trojan attacks in fiber optic system and its effects will be discussed in the paper. Keywords- Bandwidth, embedded, elimination, fiber optic system, integrated circuit, protocols, side channel analysis. 1. INTRODUCTION In today’s era, digital systems are an integral part of almost all relevant mechanisms and phenomena driving them. Thus with the increase in reliance on digital systems, their security remains an issue of utmost importance. There could be a breach in the security and performance of these systems and indirectly their performance through myriad of ways.
    [Show full text]
  • Enabling the Next Mobile Computing Revolution with Highly Integrated Armv8-A Based Socs Introduction
    ARM AND QUALCOMM: Enabling the Next Mobile Computing Revolution with Highly Integrated ARMv8-A based SoCs Introduction In the last five years, there has been a major revolution in how consumers access their digital life. People have moved away from bulky laptops with limited battery life, to smartphones and tablets. This revolution has been enabled by the availability of highly integrated SoCs (Systems on Chip) utilizing the ARM® architecture. This unique ecosystem has been facilitated by the ARM business model combined with the SoC skills of ARM licensees such as Qualcomm Technologies, Inc. This white paper examines the forces behind this mobile revolution and what will be required for companies to succeed in future evolutions of mobile computing. This paper will include how the next generation of the ARM architecture, ARMv8-A, will enable the next mobile revolution based around the ARM AArch64 64-bit instruction set while providing full support for today’s mobile ecosystem. This includes the recently announced Android L (developer preview) release with ARMv8-A 64-bit support. The ARM Business Model ARM is at the heart of mobile computing, powering more than 951 percent of smartphones in the main application processor as well as other critical support and communication chips. The ARM business model and the energy-efficient ARM architecture have fostered a wave of innovation in mobile devices from the earliest days of feature phones, to smartphones, and now tablets and new compute form factors. ARM believes its collaborative business model is a unique differentiator in which ARM licenses intellectual property (IP) designs for CPU processors, graphics processors, system components, and physical libraries to a network of ARM partners who deliver SoC solutions for a wide range of applications.
    [Show full text]
  • Anatomy of a Globally Recursive Embedded LINPACK Benchmark
    AnatomyAnatomy ofof aa GloballyGlobally RecursiveRecursive EmbeddedEmbedded LINPACKLINPACK BenchmarkBenchmark Jack Dongarra and Piotr Luszczek Batteries included. Some assembly required. ARMARM LandscapeLandscape ● Architecture – ARM11, Cortex A8, A9, A15 ● ISA – ARMv6, ARMv7 ● Floating-point – VFP11, VFPv3, VFPv4 – Pipelined and not pipelined ● Profiles – Application, Real-time, Microcontroller ● Implementations (with options and secrete sauce) – Qualcomm Snapdragon ● Scorpion, Krait – OMAP – Samsung Exynos 3, 4, 5 HPEC 2012 Waltham, MA September 10-12, 2012 Why? HPEC 2012 Waltham, MA September 10-12, 2012 FamousFamous MATLAB'sMATLAB's EasterEaster EggEgg ● MATLAB>why Jack made me do it. HPEC 2012 Waltham, MA September 10-12, 2012 1E+18 1E+16 1E+14 1E+12 1E+10 1E+8 1E+6 1E+4 1E+2 1E+0 1950 1960 1970 1980 1990 2000 2010 2020 HPEC 2012 Waltham, MA September 10-12, 2012 II ThoughtThought iOSiOS 44 hadhad AccelerateAccelerate FrameworkFramework ● With iOS 4 Apple brought Accelerate Framework ● Mac OS X includes optimized ATLAS inside Accelerate ● ATLAS port to ARM is fast enough on Linux ● But in reality... HPEC 2012 Waltham, MA September 10-12, 2012 BLASBLAS onon iPad?iPad? ● Accelerate Framework on iOS 4 did not work ● No ATLAS for iDevices – Cross compilation – No explicit compiler invocation – App signing requirements HPEC 2012 Waltham, MA September 10-12, 2012 WhatWhat isis thethe iPadiPad Hardware?Hardware? ● We know something about ISA – Must be disclosed for low-level programming ● Hardware features are only approximate – May give
    [Show full text]