hocos SLT Programs hardware-oriented computer science Why This - - PowerPoint PPT Presentation

▶
hocos
SMART_READER_LITE
LIVE PREVIEW

hocos SLT Programs hardware-oriented computer science Why This - - PowerPoint PPT Presentation

Jan 09, 2024 447 likes •711 views

Steganography for Our Research Hardware Security DeepFake Detection Foci and Your Secure Opportunities Memristive Composition Cryptography Fault attacks and Ilia Polian Stochastic countermeasures Hardware-oriented Computer Science

slide-1
SLIDE 1

hocos

hardware-oriented computer science

Hardware Security Robustness

Fault attacks and countermeasures

Emerging Architectures

Secure Composition Memristive Cryptography Steganography for DeepFake Detection Analysis of System-Level Test (SLT) Fails Automated Generation of SLT Programs Robust Quantum Computing Stochastic Computing

Our Research Foci and Your Opportunities

June 2020 Ilia Polian Hardware-oriented Computer Science

slide-2
SLIDE 2

hocos

hardware-oriented computer science

2

Why This Presentation?

  • Lectures must cover (somewhat) well-established

knowledge; new research results often lack maturity to be included into regular teaching.

  • We want to show you what we are working on now.

– The topics change over time; this material is from 2020.

  • Useful especially for those of you who are planning

to focus in our area, write a thesis with us, etc.

  • Brief sketches rather than fully-fledged coverage,

but providing references to further probe yourself.

slide-3
SLIDE 3

hocos

hardware-oriented computer science

3

Outline

  • 1. Topics related to hardware-oriented security
  • 2. Topics related to emerging technologies
  • 3. Topics related to robustness
  • 4. Connection to our teaching program
slide-4
SLIDE 4

hocos

hardware-oriented computer science

4

Hardware Security Robustness

Fault attacks and countermeasures

Emerging Architectures

Secure Composition Memristive Cryptography Steganography for DeepFake Detection

slide-5
SLIDE 5

hocos

hardware-oriented computer science

5

Fault Attacks

  • Automatic construction
  • f fault attacks.

– Tool AutoFault: Reads cipher description, produces algebraic attack. – Future: Incorporate countermeasures, protection against other attacks.

  • Security-oriented error-detecting codes.

– Compact protection codes, Rabii-Keren codes (with correction), codes incorporating randomness.

  • New attacks: Statistical Impossible Fault Attack.
slide-6
SLIDE 6

hocos

hardware-oriented computer science

6

Secure Composition

  • How to protect circuits against fault attacks, side-

channel attacks, counterfeiting, at the same time?

– Does error-detecting circuitry leak information? – Do fault attacks work on circuits with masking, locking,

  • r further countermeasures against other threats?

Circuit Error-det.

Info leakage?

P1 P128

K1, … K128

C1 C128

Locking AutoFault

P1 P128

K1, … K128

C1 C128 l1 l128

slide-7
SLIDE 7

hocos

hardware-oriented computer science

7

DeepFake Detection

  • DeepFake: Authentically looking fake video.

– E.g., face-swap, lip-synchronization, puppet-master.

  • Detect, combining deep-learning + steganography.

– Challenge: Video can be modified in legitimate ways!

Steganographic Additions

DeepFake

Detection using deep NN Legitimate Processing (compression, scaling, transmission errors …)

slide-8
SLIDE 8

hocos

hardware-oriented computer science

8

Memristive Cryptography

  • How to implement crypto functions using

memristors (emerging nano-devices)?

– Focus on novel electroforming-free BFO memristors. – Also investigate physical attacks + countermeasures.

Au BFO (600nm) Pt

A vin

  • 8
  • 4

4 8 1E-10 1E-8 1E-6 1E-4

0 V +8 V -8 V 0 V

  • n Pt/Ti/Sapphire

|Current| (A) Voltage (V) 650

  • C

(1) (2) (3) (4)

HRS LRS SET RESET

(1) (2) (3) (4)

0 V -8 V +8 V 0 V

slide-9
SLIDE 9

hocos

hardware-oriented computer science

9

To Probe Further…

  • AutoFault: https://www.doi.org/10.1109/FDTC.2019.00012
  • Error-detecting codes: https://doi.org/10.29007/w37p
  • SIFA: https://www.doi.org/10.13154/tches.v2018.i3.547-572
  • Information leakage: http://www.proofs-

workshop.org/2019/doc/PROOFS2019-Paper1.pdf

  • Camouflaging, locking, obfuscation:

https://dl.acm.org/doi/10.1145/2508859.2516656 https://ieeexplore.ieee.org/document/8203496

https://link.springer.com/article/10.1007/s10836-019-05800-4

https://ieeexplore.ieee.org/document/7546854

  • Steganography: https://doi.org/10.2352/ISSN.2470-

1173.2020.4.MWSF-076

  • DeepFake: https://doi.org/10.1109/AVSS.2018.8639163
  • Memristors: https://doi.org/10.1109/IVSW.2019.8854394
slide-10
SLIDE 10

hocos

hardware-oriented computer science

10

Outline

  • 1. Topics related to hardware-oriented security
  • 2. Topics related to emerging technologies
  • 3. Topics related to robustness
  • 4. Connection to our teaching program
slide-11
SLIDE 11

hocos

hardware-oriented computer science

11

Hardware Security Robustness

Robust Quantum Computing

Emerging Architectures

Stochastic Computing

slide-12
SLIDE 12

hocos

hardware-oriented computer science

12

Robust Quantum Computing

  • Transpilation of quantum circuits: map a circuit to

a quantum architecture with known noise levels.

  • Investigate and improve robustness of “noisy

intermediate-scale quantum” (NISQ) circuits.

– E.g., Variational Quantum Eigensolver.

Ansatz Circuit Hermi- tian H Classical Optimizer

λmin

Vector of parameters θ Compute trial state |ψ(θ)〉 Measure exp. value λθ ≥ λmin

  • f H on |ψ(θ)〉

Repeat with new θ until λθ converges

random errors

slide-13
SLIDE 13

hocos

hardware-oriented computer science

13

Stochastic Computing

  • Stochastic computing for multimodal tasks, e.g.,

image/video classification.

  • Robustness of stochastic circuits under errors.
  • Biomedical systems using stochastic computing

– E.g., X-ray image segmentation by convolutional NNs.

slide-14
SLIDE 14

hocos

hardware-oriented computer science

14

To Probe Further:

  • QC transpilation: https://arxiv.org/pdf/2002.09783.pdf ,

https://arxiv.org/pdf/1809.02573.pdf , https://arxiv.org/pdf/1712.04722.pdf

  • QC robustness: https://doi.org/10.1038/nature23879 ,

https://doi.org/10.1038/ncomms5213

  • SC basics: https://doi.org/10.1109/TCAD.2017.2778107
  • SC under errors: https://doi.org/10.1145/2990503
  • SC-based NNs: 10.1109/ICRC.2019.8914706
  • Multimodal NNs:

https://cs.stanford.edu/people/karpathy/cvpr2015.pdf

  • X-ray segment.: https://doi.org/10.1109/CHASE.2017.59
slide-15
SLIDE 15

hocos

hardware-oriented computer science

15

Outline

  • 1. Topics related to hardware-oriented security
  • 2. Topics related to emerging technologies
  • 3. Topics related to robustness
  • 4. Connection to our teaching program
slide-16
SLIDE 16

hocos

hardware-oriented computer science

16

Hardware Security Robustness

Analysis of System-Level Test (SLT) Fails

Emerging Architectures

Automated Generation of SLT Programs

slide-17
SLIDE 17

hocos

hardware-oriented computer science

17

System-Level Test (SLT)

  • Why are there SLT-unique fails, and how to prevent them?

– Complex defects? Coverage holes? System-level interactions?

  • How to generate SLT programs with desired characteristics?

– E.g., software-based stress test from high-level architecture models.

  • How to incorporate self-awareness of SoC-under-test?

Manufactured circuit with defects Post-manufacturing test System-level test SLT-unique fails reported

slide-18
SLIDE 18

hocos

hardware-oriented computer science

18

Outline

  • 1. Topics related to hardware-oriented security
  • 2. Topics related to emerging technologies
  • 3. Topics related to robustness
  • 4. Connection to our teaching program
slide-19
SLIDE 19

hocos

hardware-oriented computer science

19

Our Teaching Offer

Semester

1 (winter)

Computer Architecture and Organization Electronic Design Automation Robust System Design Hardware Oriented Security Seminar Emerging Archtiectures Seminar HW Oriented Security Master Thesis Study Project

2 (summer) 3 (winter) 4 (summer)

Lab course RISC-V Proces- sor Design Shifted to winter due to COVID- 19 restrictions

slide-20
SLIDE 20

hocos

hardware-oriented computer science

20

Teaching Offer Details

  • This is a generic plan; we cannot guarantee it for

each semester (we may have to skip some courses).

  • We can also offer new courses, e.g., other seminars.
  • CAO and EDA are more general-interest lectures;

HOS and RSD are closer to our own research.

  • RISC-V Processor Design is a new lab course. It

requires CAO. Other courses have no prerequisites.

  • If you have to choose between a seminar and a

project, we recommend taking a seminar.

  • Take some of our courses before asking for a project.
slide-21
SLIDE 21

hocos

hardware-oriented computer science

21

Thesis, Projects & Co

  • We do not have a list of pre-defined topics.

– We want to define your dream topic (and no, we cannot supervise a topic that we do not understand ourselves). – Fill out the questionnaire on the HOCOS website and send it with your transcript of records.

  • You are encouraged to talk to group members if a

topic is of interest to you (see list on next slide).

– We expect some pre-existing knowledge on that topic, e.g., HOS for security, Prof. Leymann’s / Prof. Barz’s lecture for quantum, deep-learning lecture for DeepFake. – We prefer people who did a seminar with us (your thesis topic can but doesn’t have to extend your seminar topic).

slide-22
SLIDE 22

hocos

hardware-oriented computer science

22

Your Main Contacts

  • Mael Gay: AutoFault, Error-detecting codes, fault
  • vs. side-channel attacks, masking.
  • Devanshi Upadhyaya: AutoFault, locking/obfusc.
  • Swaroop Shankar Prasad: Stego, DeepFake.
  • N.N. (we are currently hiring): Memristors.
  • Sebastian Brandhofer: Quantum circuits.
  • Florian Neugebauer: Robust stochastic circuits.
  • Roshwin Sengupta: Stochastic multimodal NNs.
  • Nourhan Elhamawy: System-level test.
slide-23
SLIDE 23

hocos

hardware-oriented computer science

23

Recent Master Thesis Topics

  • Preliminary Hazard Analysis and Fault Handling Methods in

Solar Thermal Power Plant Control Systems.

  • Hardware Optimization of Code-based Post-quantum

Cryptosystem based on Quasi-dyadic Goppa Codes.

  • Framework for mapping a given neural network onto a

stochastic circuit.

  • FPGA-Based Elliptic Curve Fault Attacks.
  • Detection of Malicious Spatial-Domain Steganography over

Noisy Channels Using Convolutional Neural Networks.

  • Implementation and Analysis of Stochastic Convolutional

Neural Network (LeNet-5) on FPGA.

  • Evaluating Robustness of Stochastic Neural Networks

against Adversarial Learning Attacks.

slide-24
SLIDE 24

hocos

hardware-oriented computer science

24

Hardware Security Robustness

Fault attacks and countermeasures

Emerging Architectures

Secure Composition Memristive Cryptography Steganography for DeepFake Detection Analysis of System-Level Test (SLT) Fails Automated Generation of SLT Programs Robust Quantum Computing Stochastic Computing