SlideShare a Scribd company logo
1 of 21
fortiss GmbH
An-
Formal Methods for Dependable Neural Networks
Towards Certifying Dependable Neural Networks and the role of Formal Methods
ACM Chapters Computer Science in Cars Symposium (CSCS 2017), Munich
Chih-Hong Cheng, Georg
Content of this work is based on a project contributed also by other researchers within fortiss:
Federik Diehl, Gereon Hinz, Michale Troung Le, Markus Rickert, Harald Ruess
Landesinstitut des Freistaats Bayern
2
• Non-profit academic research institute
• Associated with TU Munich
fortiss - An-Institut Technische
Data-driven engineering
Classical approach from specification to
implementation
• Costly to execute process-based
certification (ISO 26262, DO-178C, IEC
61508)
fortiss GmbH3
Artificial Neural Network (ANN) approach learning
from data
• Fast to develop (quick win)
•
• Implementation behaves like a black box
• No certification method exists to deal with ANN
Shall we just completely drop off existing assurance
approaches (ISO-26262) and embrace the new era?
ISO-26262
DO-178C
Requirement-to-code traceability
Architecture deign
Testing and verification
Maybe not!
Dependable neural networks are crucial for safe and
secure autonomous and decision systems
4
Towards dependable ANN
from a certification perspective
• Goal of process-based certification as in DO-178C, ISO 26262,
IEC 61508:
– Specification
• Assume that specification is correct, or
• Provide evidence that specification is correct
– Provide evidence that implementation realizes specification
• Via providing understandability guarantees (e.g., code block X realizes
specification Y)
• Via testing and coverage criteria, or static analysis (e.g., DO-333)
5
Deep / Convolutional / Recurrent
Neural Network for
control policy decision
Towards dependable Neural Networks
From a certification perspective, we need to have
• Understandability: associate each substructure of an ANN
with a partial specification/functionality
– E.g., research works regarding deconvolution or heat-maps are
approaches towards this direction
• Correctness: provide (best effort) correctness claims over
partial classical specification (e.g., road safety, traffic rules)
• Accountability: the infrastructure allows to, whenever an
undesired behavior occurs in run-time, to backtrack and
understand if
– modern specification is correct and complete
– implementation realizes modern specification (due to limitation
of best effort approaches)6
Can formal methods help?
fortiss GmbH7
might bring benefit [1]
[1] http://spectrum.ieee.org/cars-that-think/transportation/self-
driving/toyota-gill-pratt-on-the-reality-of-full-autonomy
https://www.mathworks.com/products/sldesignverifier.html
https://shemesh.larc.nasa.gov/people/bld/ftp/NASA-CR-
2014-218244.pdf
http://fbinfer.com/
http://www.astree.ens.fr/
Success stories!
Formal methods for dependable neural networks
fortiss in-house projects
Dependable
Neural Network
Understandability
AccountabilityCorrectness
8
Formal methods for dependable neural networks
In-house projects
Dependable
Neural Network
Understandability
AccountabilityCorrectness
9
nn-verifier: formal verification of
neural networks
Formal verification via constraint programming
• Neural networks with piece-wise linear activation
functions can be modeled as mixed-integer-
linear constraints.
• By presenting the property of the network to be
constraints or objectives, we reduce the
verification problem to a MILP problem.
– By solving the optimization problem we compute the
robustness or prove a property of the neural network.
Examples of using this technique?
10
Example 1
Resilience bound for neural networks
11
Neural Network
Neural Network
good can your neural
network resist sensor
A formal, computable, and comparable
measure can act as an indicator or as a
differentiator
Defining Resilience
• We define a resilience metric that can be computed
precisely
– F maximal
allowed perturbation
•
we go beyond a single image and noise
• Equivalently, we ask what is the minimal perturbation to lead to bad
behavior
• The tricky part is about the output layer
–
into MILP
– softmax: a function that involves the computation of
exponentials e^{x}
fortiss GmbH12
Preprint available at
https://arxiv.org/abs/1705.01040
95 % : 3
5 % : 8
21 % : 2
21 % : 3
57 % : 8
image Mininumly perturbed image
Neural
Network description
formal reasoning engine
(nn-verifier)
outputs
input
Example 2
Safety of highway motion predictor
Properties under consideration:
• [Problematic decision] Is it possible for the
controller to suggest go left, while there is
already car in the left?
• [Strange speed range] Is it possible, for all
cars to be between 100~110 km/h, that the
controller suggests to run 200 km/h?
• [Effect of output difference] By introducing
sensor imprecision, what can be the maximal
speed difference suggested by the neural
network?
13
Highway motion predictior, being trained under the NGSim dataset
SlideShare users: Move to the next page
for accessing the Youtube Video!
https://www.youtube.com/watch?v=C_Z2s-fauKY
Demonstration
nn-verifier (research prototype from fortiss)
14
The verification technique is implemented in a tool called
nn-verifier, using IBM CPLEX as its underlying MILP solver.
SlideShare users: Move to the next page for accessing the Youtube Video!
http://www.youtube.com/watch?v=BK825-_ScCU
Formal methods for dependable neural networks
In-house projects
Dependable
Neural Network
Understandability
AccountabilityCorrectness
15
Formal synthesis for pervasive
controllers for run-time
monitoring/enforcement
Overlaying a neural network by a monitor / regulator
You may not always want to trust neural network
• This design of monitor/regulator is fine to take partial specifications, such as safety rules
– It constrains some output values created by the neural network
fortiss GmbH16
Controller
being trained
using neural
network
Sensor input
Prohibit
some
actions
Actuation
Controller
(synthesized)
from formal,
classical
specification
Allowed:
{speed-up,
go-right}
speed-up: 0%
go-right: 45%
go-left: 47%
go-right
The
action
with 2nd
largest
prob is
selected!e
Run-time monitors / enforcement
• Create components from formal specifications
– Runtime monitoring units
• For examining if current output is has high confidence
• For examining if current output is consistent with actions regulated by the partial specification
– Runtime enforcement units
• Perform corrective actions
• We want to use formal specification, formal synthesis, and model checking to guarantee highest
safety requirement such as SIL-4 or ASIL-D
17
Synthesizing monitors = finding maximal pervasive controller
• The basic concept is about maximum pervasive
controller in a safety game
• It is more complicated when numeric is involved,
i.e., to have specification that goes beyond
Boolean variables
risk attractors
(states which eventually leads to risk)
Risk
C8
C2
b a
c
a
c
Maximum
pervasive controller
Demonstration
Formal synthesis of pervasive controllers (research prototype from fortiss)
19
SlideShare users: Please
YouTube video to see the
explanation!
http://www.youtube.com/watch?v=p26rfsl-ohk
The gamified simulator is modified from the highway overtaking simulator
from the MIT 6.S094 course http://selfdrivingcars.mit.edu/deeptrafficjs/
Outlook
• Formal methods can help introducing neural networks in critical environments
– From formal verification to run-time verification/enforcement
• Further research directions
– A certification roadmap, as well as formal method complements
• Analogous to DO-178C (safety for civil avionics) and DO-333 (formal method complement)
– Understandablity of neural networks by formal methods
– Scalability of verification by combining approaches (e.g., by taking knowledge of de-convolution)
–
20
Dr. Chih-Hong Cheng, Georg
fortiss GmbH
An-
tel +49 89 3603522 11 fax +49 89 3603522 50
info@fortiss.org
www.fortiss.org
21

More Related Content

Similar to Formal Methods Dependable Neural Networks

Case study of dcs upgrade how to reduce stress during execution
Case study of dcs upgrade how to reduce stress during executionCase study of dcs upgrade how to reduce stress during execution
Case study of dcs upgrade how to reduce stress during executionJohn Kingsley
 
[SiriusCon 2020] Realization of Model-Based Safety Analysis and Integration w...
[SiriusCon 2020] Realization of Model-Based Safety Analysis and Integration w...[SiriusCon 2020] Realization of Model-Based Safety Analysis and Integration w...
[SiriusCon 2020] Realization of Model-Based Safety Analysis and Integration w...Obeo
 
Asset Utilization Metrics Propel a Revival in Safety Solutions
Asset Utilization Metrics Propel a Revival in Safety SolutionsAsset Utilization Metrics Propel a Revival in Safety Solutions
Asset Utilization Metrics Propel a Revival in Safety SolutionsARC Advisory Group
 
It‘s Math That Drives Things – Simulink as Simulation and Modeling Environment
It‘s Math That Drives Things – Simulink as Simulation and Modeling EnvironmentIt‘s Math That Drives Things – Simulink as Simulation and Modeling Environment
It‘s Math That Drives Things – Simulink as Simulation and Modeling EnvironmentJoachim Schlosser
 
Automatized testing hil system for agile product-design environment
Automatized testing hil system for agile product-design environmentAutomatized testing hil system for agile product-design environment
Automatized testing hil system for agile product-design environmentTritem
 
Model-Based Design & Analysis.ppt
Model-Based Design & Analysis.pptModel-Based Design & Analysis.ppt
Model-Based Design & Analysis.pptRajuRaju183149
 
wp-01070-hardware-software-coverification-fpga
wp-01070-hardware-software-coverification-fpgawp-01070-hardware-software-coverification-fpga
wp-01070-hardware-software-coverification-fpgaGautam Kavipurapu
 
Innovate with confidence – Functional Verification of Embedded Algorithms
Innovate with confidence – Functional Verification of Embedded AlgorithmsInnovate with confidence – Functional Verification of Embedded Algorithms
Innovate with confidence – Functional Verification of Embedded AlgorithmsJoachim Schlosser
 
HiPEAC2023-DL4IoT Workshop_Jean Hagemeyer presentation
HiPEAC2023-DL4IoT Workshop_Jean Hagemeyer presentationHiPEAC2023-DL4IoT Workshop_Jean Hagemeyer presentation
HiPEAC2023-DL4IoT Workshop_Jean Hagemeyer presentationVEDLIoT Project
 
Defesa de Doutorado: HAMSTER - healthy, mobility and security-based data comm...
Defesa de Doutorado: HAMSTER - healthy, mobility and security-based data comm...Defesa de Doutorado: HAMSTER - healthy, mobility and security-based data comm...
Defesa de Doutorado: HAMSTER - healthy, mobility and security-based data comm...Daniel Fernando Pigatto
 
SourceWarp AST 2023.pdf
SourceWarp AST 2023.pdfSourceWarp AST 2023.pdf
SourceWarp AST 2023.pdfJulian Thome
 
Towards 0-bug software in the automotive industry
Towards 0-bug software in the automotive industryTowards 0-bug software in the automotive industry
Towards 0-bug software in the automotive industryAshley Zupkus
 
Face detection on_embedded_systems
Face detection on_embedded_systemsFace detection on_embedded_systems
Face detection on_embedded_systemsr_sadoun
 
Industrial Ethernet Facts - The 5 major technologies
Industrial Ethernet Facts - The 5 major technologiesIndustrial Ethernet Facts - The 5 major technologies
Industrial Ethernet Facts - The 5 major technologiesStephane Potier
 
Time is ready for the Civil Infrastructure Platform
Time is ready for the Civil Infrastructure PlatformTime is ready for the Civil Infrastructure Platform
Time is ready for the Civil Infrastructure PlatformYoshitake Kobayashi
 
Real-Time Engineering Simulators
Real-Time Engineering SimulatorsReal-Time Engineering Simulators
Real-Time Engineering SimulatorsGSE Systems, Inc.
 
Formal Verification Of An Intellectual Property In a Field Programmable Gate ...
Formal Verification Of An Intellectual Property In a Field Programmable Gate ...Formal Verification Of An Intellectual Property In a Field Programmable Gate ...
Formal Verification Of An Intellectual Property In a Field Programmable Gate ...IRJET Journal
 
DO-254 for dummies 7
DO-254 for dummies 7DO-254 for dummies 7
DO-254 for dummies 7DMAP
 

Similar to Formal Methods Dependable Neural Networks (20)

Case study of dcs upgrade how to reduce stress during execution
Case study of dcs upgrade how to reduce stress during executionCase study of dcs upgrade how to reduce stress during execution
Case study of dcs upgrade how to reduce stress during execution
 
[SiriusCon 2020] Realization of Model-Based Safety Analysis and Integration w...
[SiriusCon 2020] Realization of Model-Based Safety Analysis and Integration w...[SiriusCon 2020] Realization of Model-Based Safety Analysis and Integration w...
[SiriusCon 2020] Realization of Model-Based Safety Analysis and Integration w...
 
Asset Utilization Metrics Propel a Revival in Safety Solutions
Asset Utilization Metrics Propel a Revival in Safety SolutionsAsset Utilization Metrics Propel a Revival in Safety Solutions
Asset Utilization Metrics Propel a Revival in Safety Solutions
 
It‘s Math That Drives Things – Simulink as Simulation and Modeling Environment
It‘s Math That Drives Things – Simulink as Simulation and Modeling EnvironmentIt‘s Math That Drives Things – Simulink as Simulation and Modeling Environment
It‘s Math That Drives Things – Simulink as Simulation and Modeling Environment
 
Automatized testing hil system for agile product-design environment
Automatized testing hil system for agile product-design environmentAutomatized testing hil system for agile product-design environment
Automatized testing hil system for agile product-design environment
 
Model-Based Design & Analysis.ppt
Model-Based Design & Analysis.pptModel-Based Design & Analysis.ppt
Model-Based Design & Analysis.ppt
 
wp-01070-hardware-software-coverification-fpga
wp-01070-hardware-software-coverification-fpgawp-01070-hardware-software-coverification-fpga
wp-01070-hardware-software-coverification-fpga
 
Innovate with confidence – Functional Verification of Embedded Algorithms
Innovate with confidence – Functional Verification of Embedded AlgorithmsInnovate with confidence – Functional Verification of Embedded Algorithms
Innovate with confidence – Functional Verification of Embedded Algorithms
 
HiPEAC2023-DL4IoT Workshop_Jean Hagemeyer presentation
HiPEAC2023-DL4IoT Workshop_Jean Hagemeyer presentationHiPEAC2023-DL4IoT Workshop_Jean Hagemeyer presentation
HiPEAC2023-DL4IoT Workshop_Jean Hagemeyer presentation
 
5 Thomas Magedanz
5  Thomas Magedanz5  Thomas Magedanz
5 Thomas Magedanz
 
Defesa de Doutorado: HAMSTER - healthy, mobility and security-based data comm...
Defesa de Doutorado: HAMSTER - healthy, mobility and security-based data comm...Defesa de Doutorado: HAMSTER - healthy, mobility and security-based data comm...
Defesa de Doutorado: HAMSTER - healthy, mobility and security-based data comm...
 
SourceWarp AST 2023.pdf
SourceWarp AST 2023.pdfSourceWarp AST 2023.pdf
SourceWarp AST 2023.pdf
 
Towards 0-bug software in the automotive industry
Towards 0-bug software in the automotive industryTowards 0-bug software in the automotive industry
Towards 0-bug software in the automotive industry
 
Profile tulasi v1.1
Profile tulasi v1.1Profile tulasi v1.1
Profile tulasi v1.1
 
Face detection on_embedded_systems
Face detection on_embedded_systemsFace detection on_embedded_systems
Face detection on_embedded_systems
 
Industrial Ethernet Facts - The 5 major technologies
Industrial Ethernet Facts - The 5 major technologiesIndustrial Ethernet Facts - The 5 major technologies
Industrial Ethernet Facts - The 5 major technologies
 
Time is ready for the Civil Infrastructure Platform
Time is ready for the Civil Infrastructure PlatformTime is ready for the Civil Infrastructure Platform
Time is ready for the Civil Infrastructure Platform
 
Real-Time Engineering Simulators
Real-Time Engineering SimulatorsReal-Time Engineering Simulators
Real-Time Engineering Simulators
 
Formal Verification Of An Intellectual Property In a Field Programmable Gate ...
Formal Verification Of An Intellectual Property In a Field Programmable Gate ...Formal Verification Of An Intellectual Property In a Field Programmable Gate ...
Formal Verification Of An Intellectual Property In a Field Programmable Gate ...
 
DO-254 for dummies 7
DO-254 for dummies 7DO-254 for dummies 7
DO-254 for dummies 7
 

Recently uploaded

Pigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping ElbowsPigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping ElbowsPigging Solutions
 
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)Mark Simos
 
Streamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupStreamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupFlorian Wilhelm
 
WordPress Websites for Engineers: Elevate Your Brand
WordPress Websites for Engineers: Elevate Your BrandWordPress Websites for Engineers: Elevate Your Brand
WordPress Websites for Engineers: Elevate Your Brandgvaughan
 
New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024BookNet Canada
 
Nell’iperspazio con Rocket: il Framework Web di Rust!
Nell’iperspazio con Rocket: il Framework Web di Rust!Nell’iperspazio con Rocket: il Framework Web di Rust!
Nell’iperspazio con Rocket: il Framework Web di Rust!Commit University
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesSinan KOZAK
 
My Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 PresentationMy Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 PresentationRidwan Fadjar
 
Scanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsScanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsRizwan Syed
 
"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii SoldatenkoFwdays
 
Benefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other FrameworksBenefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other FrameworksSoftradix Technologies
 
"ML in Production",Oleksandr Bagan
"ML in Production",Oleksandr Bagan"ML in Production",Oleksandr Bagan
"ML in Production",Oleksandr BaganFwdays
 
Key Features Of Token Development (1).pptx
Key  Features Of Token  Development (1).pptxKey  Features Of Token  Development (1).pptx
Key Features Of Token Development (1).pptxLBM Solutions
 
Gen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdfGen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdfAddepto
 
Install Stable Diffusion in windows machine
Install Stable Diffusion in windows machineInstall Stable Diffusion in windows machine
Install Stable Diffusion in windows machinePadma Pradeep
 
Build your next Gen AI Breakthrough - April 2024
Build your next Gen AI Breakthrough - April 2024Build your next Gen AI Breakthrough - April 2024
Build your next Gen AI Breakthrough - April 2024Neo4j
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 3652toLead Limited
 
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmaticsKotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmaticscarlostorres15106
 
Bluetooth Controlled Car with Arduino.pdf
Bluetooth Controlled Car with Arduino.pdfBluetooth Controlled Car with Arduino.pdf
Bluetooth Controlled Car with Arduino.pdfngoud9212
 

Recently uploaded (20)

Hot Sexy call girls in Panjabi Bagh 🔝 9953056974 🔝 Delhi escort Service
Hot Sexy call girls in Panjabi Bagh 🔝 9953056974 🔝 Delhi escort ServiceHot Sexy call girls in Panjabi Bagh 🔝 9953056974 🔝 Delhi escort Service
Hot Sexy call girls in Panjabi Bagh 🔝 9953056974 🔝 Delhi escort Service
 
Pigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping ElbowsPigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping Elbows
 
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
 
Streamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupStreamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project Setup
 
WordPress Websites for Engineers: Elevate Your Brand
WordPress Websites for Engineers: Elevate Your BrandWordPress Websites for Engineers: Elevate Your Brand
WordPress Websites for Engineers: Elevate Your Brand
 
New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
 
Nell’iperspazio con Rocket: il Framework Web di Rust!
Nell’iperspazio con Rocket: il Framework Web di Rust!Nell’iperspazio con Rocket: il Framework Web di Rust!
Nell’iperspazio con Rocket: il Framework Web di Rust!
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen Frames
 
My Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 PresentationMy Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 Presentation
 
Scanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsScanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL Certs
 
"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko
 
Benefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other FrameworksBenefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other Frameworks
 
"ML in Production",Oleksandr Bagan
"ML in Production",Oleksandr Bagan"ML in Production",Oleksandr Bagan
"ML in Production",Oleksandr Bagan
 
Key Features Of Token Development (1).pptx
Key  Features Of Token  Development (1).pptxKey  Features Of Token  Development (1).pptx
Key Features Of Token Development (1).pptx
 
Gen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdfGen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdf
 
Install Stable Diffusion in windows machine
Install Stable Diffusion in windows machineInstall Stable Diffusion in windows machine
Install Stable Diffusion in windows machine
 
Build your next Gen AI Breakthrough - April 2024
Build your next Gen AI Breakthrough - April 2024Build your next Gen AI Breakthrough - April 2024
Build your next Gen AI Breakthrough - April 2024
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
 
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmaticsKotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
 
Bluetooth Controlled Car with Arduino.pdf
Bluetooth Controlled Car with Arduino.pdfBluetooth Controlled Car with Arduino.pdf
Bluetooth Controlled Car with Arduino.pdf
 

Formal Methods Dependable Neural Networks

  • 1. fortiss GmbH An- Formal Methods for Dependable Neural Networks Towards Certifying Dependable Neural Networks and the role of Formal Methods ACM Chapters Computer Science in Cars Symposium (CSCS 2017), Munich Chih-Hong Cheng, Georg Content of this work is based on a project contributed also by other researchers within fortiss: Federik Diehl, Gereon Hinz, Michale Troung Le, Markus Rickert, Harald Ruess
  • 2. Landesinstitut des Freistaats Bayern 2 • Non-profit academic research institute • Associated with TU Munich fortiss - An-Institut Technische
  • 3. Data-driven engineering Classical approach from specification to implementation • Costly to execute process-based certification (ISO 26262, DO-178C, IEC 61508) fortiss GmbH3 Artificial Neural Network (ANN) approach learning from data • Fast to develop (quick win) • • Implementation behaves like a black box • No certification method exists to deal with ANN Shall we just completely drop off existing assurance approaches (ISO-26262) and embrace the new era? ISO-26262 DO-178C Requirement-to-code traceability Architecture deign Testing and verification
  • 4. Maybe not! Dependable neural networks are crucial for safe and secure autonomous and decision systems 4
  • 5. Towards dependable ANN from a certification perspective • Goal of process-based certification as in DO-178C, ISO 26262, IEC 61508: – Specification • Assume that specification is correct, or • Provide evidence that specification is correct – Provide evidence that implementation realizes specification • Via providing understandability guarantees (e.g., code block X realizes specification Y) • Via testing and coverage criteria, or static analysis (e.g., DO-333) 5 Deep / Convolutional / Recurrent Neural Network for control policy decision
  • 6. Towards dependable Neural Networks From a certification perspective, we need to have • Understandability: associate each substructure of an ANN with a partial specification/functionality – E.g., research works regarding deconvolution or heat-maps are approaches towards this direction • Correctness: provide (best effort) correctness claims over partial classical specification (e.g., road safety, traffic rules) • Accountability: the infrastructure allows to, whenever an undesired behavior occurs in run-time, to backtrack and understand if – modern specification is correct and complete – implementation realizes modern specification (due to limitation of best effort approaches)6
  • 7. Can formal methods help? fortiss GmbH7 might bring benefit [1] [1] http://spectrum.ieee.org/cars-that-think/transportation/self- driving/toyota-gill-pratt-on-the-reality-of-full-autonomy https://www.mathworks.com/products/sldesignverifier.html https://shemesh.larc.nasa.gov/people/bld/ftp/NASA-CR- 2014-218244.pdf http://fbinfer.com/ http://www.astree.ens.fr/ Success stories!
  • 8. Formal methods for dependable neural networks fortiss in-house projects Dependable Neural Network Understandability AccountabilityCorrectness 8
  • 9. Formal methods for dependable neural networks In-house projects Dependable Neural Network Understandability AccountabilityCorrectness 9 nn-verifier: formal verification of neural networks
  • 10. Formal verification via constraint programming • Neural networks with piece-wise linear activation functions can be modeled as mixed-integer- linear constraints. • By presenting the property of the network to be constraints or objectives, we reduce the verification problem to a MILP problem. – By solving the optimization problem we compute the robustness or prove a property of the neural network. Examples of using this technique? 10
  • 11. Example 1 Resilience bound for neural networks 11 Neural Network Neural Network good can your neural network resist sensor A formal, computable, and comparable measure can act as an indicator or as a differentiator
  • 12. Defining Resilience • We define a resilience metric that can be computed precisely – F maximal allowed perturbation • we go beyond a single image and noise • Equivalently, we ask what is the minimal perturbation to lead to bad behavior • The tricky part is about the output layer – into MILP – softmax: a function that involves the computation of exponentials e^{x} fortiss GmbH12 Preprint available at https://arxiv.org/abs/1705.01040 95 % : 3 5 % : 8 21 % : 2 21 % : 3 57 % : 8 image Mininumly perturbed image Neural Network description formal reasoning engine (nn-verifier) outputs input
  • 13. Example 2 Safety of highway motion predictor Properties under consideration: • [Problematic decision] Is it possible for the controller to suggest go left, while there is already car in the left? • [Strange speed range] Is it possible, for all cars to be between 100~110 km/h, that the controller suggests to run 200 km/h? • [Effect of output difference] By introducing sensor imprecision, what can be the maximal speed difference suggested by the neural network? 13 Highway motion predictior, being trained under the NGSim dataset SlideShare users: Move to the next page for accessing the Youtube Video! https://www.youtube.com/watch?v=C_Z2s-fauKY
  • 14. Demonstration nn-verifier (research prototype from fortiss) 14 The verification technique is implemented in a tool called nn-verifier, using IBM CPLEX as its underlying MILP solver. SlideShare users: Move to the next page for accessing the Youtube Video! http://www.youtube.com/watch?v=BK825-_ScCU
  • 15. Formal methods for dependable neural networks In-house projects Dependable Neural Network Understandability AccountabilityCorrectness 15 Formal synthesis for pervasive controllers for run-time monitoring/enforcement
  • 16. Overlaying a neural network by a monitor / regulator You may not always want to trust neural network • This design of monitor/regulator is fine to take partial specifications, such as safety rules – It constrains some output values created by the neural network fortiss GmbH16 Controller being trained using neural network Sensor input Prohibit some actions Actuation Controller (synthesized) from formal, classical specification Allowed: {speed-up, go-right} speed-up: 0% go-right: 45% go-left: 47% go-right The action with 2nd largest prob is selected!e
  • 17. Run-time monitors / enforcement • Create components from formal specifications – Runtime monitoring units • For examining if current output is has high confidence • For examining if current output is consistent with actions regulated by the partial specification – Runtime enforcement units • Perform corrective actions • We want to use formal specification, formal synthesis, and model checking to guarantee highest safety requirement such as SIL-4 or ASIL-D 17
  • 18. Synthesizing monitors = finding maximal pervasive controller • The basic concept is about maximum pervasive controller in a safety game • It is more complicated when numeric is involved, i.e., to have specification that goes beyond Boolean variables risk attractors (states which eventually leads to risk) Risk C8 C2 b a c a c Maximum pervasive controller
  • 19. Demonstration Formal synthesis of pervasive controllers (research prototype from fortiss) 19 SlideShare users: Please YouTube video to see the explanation! http://www.youtube.com/watch?v=p26rfsl-ohk The gamified simulator is modified from the highway overtaking simulator from the MIT 6.S094 course http://selfdrivingcars.mit.edu/deeptrafficjs/
  • 20. Outlook • Formal methods can help introducing neural networks in critical environments – From formal verification to run-time verification/enforcement • Further research directions – A certification roadmap, as well as formal method complements • Analogous to DO-178C (safety for civil avionics) and DO-333 (formal method complement) – Understandablity of neural networks by formal methods – Scalability of verification by combining approaches (e.g., by taking knowledge of de-convolution) – 20
  • 21. Dr. Chih-Hong Cheng, Georg fortiss GmbH An- tel +49 89 3603522 11 fax +49 89 3603522 50 info@fortiss.org www.fortiss.org 21