Interactions between Software Product Lines and Adversarial Machine - - PowerPoint PPT Presentation

▶

Feb 01, 2024 23 likes •187 views

Interactions between Software Product Lines and Adversarial Machine Learning Paul TEMPLE 1 Gilles PERROUIN 1 , 2 Pierre-Yves SCHOBBENS 1 Patrick HEYMANS 1 1 NaDI, PReCISE, Faculty of Computer Science, University of Namur 2 FNRS April, 12 th 2019

SLIDE 1

Interactions between Software Product Lines and Adversarial Machine Learning

Paul TEMPLE 1 Gilles PERROUIN 1,2 Pierre-Yves SCHOBBENS 1 Patrick HEYMANS 1

1NaDI, PReCISE, Faculty of Computer Science, University of Namur 2FNRS

April, 12th 2019

this work is funded by the EOS VeriLearn project

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 1 / 7

SLIDE 2

Software Product Lines and Machine Learning

Machine Learning in Software Product Line

ML can help to deal with # of products (Linux kernel ≈ 215k) Inference mechanism → no generation; reason on configurations

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 2 / 7

SLIDE 3

Software Product Lines and Machine Learning

Machine Learning in Software Product Line

ML can help to deal with # of products (Linux kernel ≈ 215k) Inference mechanism → no generation; reason on configurations

Performance prediction with ML

Guo et al., Variability-Aware performance prediction: a statistical learning approach, ASE 2013 Siegmund et al., Performance-influence models for highly configurable systems, FSE 2015 Temple et al., Learning-based performance specialization of configurable systems, SPLC 2016

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 2 / 7

SLIDE 4

Machine Learning

Learning process

Sample a (small) number of configurations Generate associated products and measure performances Build a prediction model to infer performances Use the model on new configurations

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 3 / 7

SLIDE 5

Machine Learning

Learning process

Sample a (small) number of configurations Generate associated products and measure performances Build a prediction model to infer performances Use the model on new configurations

Machine Learning assumptions

Initial sample is representative of the configurations’ population New configurations follow the same distribution than initial sample

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 3 / 7

SLIDE 6

Machine Learning

But...

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 4 / 7

SLIDE 7

Machine Learning

But...

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 4 / 7

SLIDE 8

Machine Learning

But...

Craft configurations to artificially increase the number of errors of the prediction model → Adversarial Machine Learning

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 4 / 7

SLIDE 9

Breaking the rules of Machine Learning

Adversarial Machine Learning

Appeared in 2004 Popular around 2014 with GANsa Still popular todayb

Biggio and Roli, Wild patterns: Ten years after the rise of adversarial machine learning, Pattern Recognition,Vol. 84, 2018

aGoodfellow et al., Generative Adversarial Nets, NIPS 2014 bZhang et al., DeepRoad: GAN-based Metamorphic Autonomous Driving

System Testing, ASE’18

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 5 / 7

SLIDE 10

Breaking the rules of Machine Learning

Adversarial Machine Learning

Appeared in 2004 Popular around 2014 with GANsa Still popular todayb

Biggio and Roli, Wild patterns: Ten years after the rise of adversarial machine learning, Pattern Recognition,Vol. 84, 2018

aGoodfellow et al., Generative Adversarial Nets, NIPS 2014 bZhang et al., DeepRoad: GAN-based Metamorphic Autonomous Driving

System Testing, ASE’18

Goal of Adversarial Machine Learning

Better understand ML algorithms, their assumptions and weaknesses

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 5 / 7

SLIDE 11

Adversarial Machine Learning intuition

Add perturbations

⇒

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 6 / 7

SLIDE 12

Adversarial Machine Learning intuition

Add perturbations

⇒

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 6 / 7

SLIDE 13

Adversarial Machine Learning intuition

Add perturbations

⇒

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 6 / 7

SLIDE 14

Adversarial Machine Learning intuition

Add perturbations

⇒

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 6 / 7

SLIDE 15

SPL and Adversarial ML

SPL for AdvML

Various ways to create adversarial configurations → variability modeling Open questions:

Is it interesting? Can those attacks be composed? Can it help designing new adversarial techniques?

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 7 / 7

SLIDE 16

SPL and Adversarial ML

SPL for AdvML

Various ways to create adversarial configurations → variability modeling Open questions:

Is it interesting? Can those attacks be composed? Can it help designing new adversarial techniques?

AdvML for SPL testing

(Submitted to SPLC’19) Use AdvML for SPL Quality Assurance Adapted one adversarial technique to SPL Understand if the Variability Model is under/over constrained How to take into account constraints from the SPL in the process?

P TEMPLE (NaDI, PReCISE, UNamur) SPLs and advML April, 12th 2019 7 / 7