book

Machine Learning Security Principles

Name: Machine Learning Security Principles
Author: John Paul Mueller
ISBN: 9781804618851

by John Paul Mueller

December 2022

Beginner to intermediate

450 pages

12h 37m

English

Packt Publishing

Audiobook available

Read now

Unlock full access

Machine Learning Security Principles
ForewordContributorsAbout the authorAcknowledgementsAbout the reviewers
Preface
Who this book is forWhat this book coversTo get the most out of this bookDownload the example code filesConventions usedGet in touchShare Your ThoughtsDownload a free PDF copy of this book
Part 1 – Securing a Machine Learning System
Chapter 1: Defining Machine Learning Security
Building a picture of MLWhy is ML important?Identifying the ML security domainDistinguishing between supervised and unsupervisedUsing ML from development to productionAdding security to MLDefining the human elementCompromising the integrity and availability of ML modelsDescribing the types of attacks against MLConsidering what ML security can achieveSetting up for the bookWhat do you need to know?Considering the programming setupSummary
Chapter 2: Mitigating Risk at Training by Validating and Maintaining Datasets
Technical requirementsDefining dataset threatsLearning about the kinds of database threatsConsidering dataset threat sourcesDelving into data changeDelving into data corruptionUncovering feature manipulationExamining source modificationThwarting privacy attacksDetecting dataset modificationAn example of relying on traditional methodsWorking with hashes and larger filesUsing a data version control system exampleMitigating dataset corruptionThe human factor in missingnessAn example of recreating the datasetUsing an imputerHandling missing or corrupted dataSummary
Chapter 3: Mitigating Inference Risk by Avoiding Adversarial Machine Learning Attacks
Defining adversarial MLCategorizing the attack vectorsExamining the hacker mindsetConsidering security issues in ML algorithmsDefining attacker motivationsEmploying CAPTCHA bypass techniquesConsidering common hacker goalsRelying on trial and errorAvoiding helping the hackerIntegrating new research quicklyUnderstanding the Black Swan TheoryDescribing the most common attack techniquesEvasion attacksModel poisoningUnderstanding membership inference attacksUnderstanding Trojan attacksUnderstanding backdoor (neural) attacksSeeing adversarial attacks in actionMitigating threats to the algorithmDeveloping principles that help protect against every threatDetecting and mitigating an evasion attackDetecting and mitigating a model poisoning attackDetecting and mitigating a membership inference attackDetecting and mitigating a Trojan attackDetecting and mitigating backdoor (neural) attacksSummaryFurther reading
Part 2 – Creating a Secure System Using ML
Chapter 4: Considering the Threat Environment
Technical requirementsDefining an environmentUnderstanding business threatsProtecting consumer sitesUnderstanding malwareUnderstanding network attacksEyeing the small stuffDealing with web APIsDealing with the hype cycleConsidering social threatsSpamIdentity theftUnwanted trackingRemote storage data loss or corruptionAccount takeoverEmploying ML in security in the real worldUnderstanding the kinds of application securityConsidering the realities of the machineAdding human interventionDeveloping a simple authentication exampleDeveloping a simple spam filter exampleSummaryFurther reading
Chapter 5: Keeping Your Network Clean
Technical requirementsDefining current network threatsDeveloping a sense of control over chaosImplementing access controlEnsuring authenticationDetecting intrusionsDefining localized attacksUnderstanding botnetsConsidering traditional protectionsWorking with honeypotsUsing data-centric securityLocating subtle intrusion indicatorsUsing alternative identity strategiesObtaining data for network traffic testingAdding ML to the mixDeveloping an updated security planDetermining which features to trackCreating real-time defensesUsing supervised learning exampleUsing a subprocess in Python exampleWorking with Flask exampleAsking for human interventionDeveloping predictive defensesDefining what is available todayDownsides of predicting the futureCreating a realistic network modelSummary
Chapter 6: Detecting and Analyzing Anomalies
Technical requirementsDefining anomaliesSpecifying the causes and effects of anomaly detectionConsidering anomaly sourcesUnderstanding when anomalies occurUsing and combining anomaly detection and signature detectionDetecting data anomaliesChecking data validityForecasting potential anomalies exampleUsing anomaly detection effectively in MLConsidering other mitigation techniquesSummaryFurther reading

Chapter 7: Dealing with Malware
Technical requirementsDefining malwareSpecifying the malware typesUnderstanding the subtleties of malwareDetermining malware goalsGenerating malware detection featuresGetting the required disassemblerCollecting data about any applicationExtracting strings from an executableExtracting images from an executableGenerating a list of application featuresSelecting the most important featuresConsidering speed of detectionBuilding a malware detection toolboxClassifying malwareObtaining malware samples and labelsDevelopment of a simple malware detection scenarioSummaryFurther reading
Chapter 8: Locating Potential Fraud
Technical requirementsUnderstanding the types of fraudDefining fraud sourcesConsidering fraudstersConsidering hackersConsidering other organizationsConsidering company insidersConsidering customers (or fraudsters posing as customers)Obtaining fraud datasetsConsidering fraud that occurs in the backgroundDetecting fraud that occurs when you’re not lookingBuilding a background fraud detection applicationConsidering fraud that occurs in real timeConsidering the types of real-time fraudDetecting real-time fraudBuilding a fraud detection exampleGetting the dataSetting the example upSplitting the data into train and test setsBuilding the modelPerforming the analysisChecking another modelCreating a ROC curve and calculating AUCSummaryFurther reading
Chapter 9: Defending against Hackers
Technical requirementsConsidering hacker targetsHosted systemsNetworksMobile devicesCustomersPublic venues and social mediaDefining hacker goalsData stealingData modificationAlgorithm modificationSystem damageMonitoring and alertingConsidering the importance of lagAn example of detecting behaviorBuilding and testing an XGBoost regressorPutting the data in perspectivePredicting new behavior based on the pastLocating other behavioral datasetsImproving security and reliabilitySummaryFurther reading
Part 3 – Protecting against ML-Driven Attacks
Chapter 10: Considering the Ramifications of Deepfakes
Technical requirementsDefining a deepfakeModifying mediaCommon deepfake typesThe history of deepfakesCreating a deepfake computer setupInstalling TensorFlow on a desktop systemChecking for a GPUUnderstanding autoencodersDefining the autoencoderWorking with an autoencoder exampleUnderstanding CNNs and implementing GANsAn overview of a Pix2Pix GANObtaining and viewing the imagesManipulating the imagesDeveloping datasets from the modified imagesCreating the generatorCreating the discriminatorPerforming optimization of both generator and discriminatorMonitoring the training processTraining the modelSummaryFurther reading
Chapter 11: Leveraging Machine Learning for Hacking
Making attacks automatic and personalizedGaining unauthorized access bypassing CAPTCHAAutomatically harvesting informationEnhancing existing capabilitiesRendering malware less effective using GANsPutting artificial intelligence in spear-phishingGenerating smart bots for fake news and reviewsSummaryFurther reading
Part 4 – Performing ML Tasks in an Ethical Manner
Chapter 12: Embracing and Incorporating Ethical Behavior
Technical requirementsSanitizing data correctlyObtaining benefits from data sanitizationConsidering the current datasetRemoving PIIAdding traits together to make them less identifiableEliminating unnecessary featuresDefining data source awarenessValidating user permissionsUsing recognizable datasetsVerifying third-party datasetsObtaining required third-party permissionsUnderstanding ML fairnessDetermining what fairness meansUnderstanding Simpson’s paradoxRemoving personal biasDefining algorithmic biasAddressing fairness concernsComputing fairness indicators with TensorFlowSolving fairness problems with TensorFlow-constrained optimizationMitigating privacy risks using federated learning and differential privacyDistributing data and privacy risks using federated learningRelying on differential privacySummaryFurther reading
Index
Why subscribe?
Other Books You May EnjoyPackt is searching for authors like youShare Your ThoughtsDownload a free PDF copy of this book

Overview

Machine Learning Security Principles explores the landscape of securing machine learning systems against modern threats. Inside, you'll learn how to detect and prevent unauthorized access, mitigate risks from adversarial machine learning, and safeguard datasets ethically. This book equips you with actionable techniques to secure data and applications, ensuring privacy and robustness.

What this Book will help me do

Understand how to detect and prevent unauthorized access to machine learning systems.
Learn methods to mitigate adversarial machine learning threats effectively.
Gain proficiency in ethical data management to reduce privacy risks.
Develop strategies to detect and respond to fraud and deepfake threats.
Enhance your machine learning security measures to protect applications.

Author(s)

John Paul Mueller is an experienced author and technologist specializing in security, programming, and data science. With numerous books on technical topics, he brings clarity and practical insights to complex subjects. His engaging style makes advanced concepts accessible, and he writes with the dedicated purpose of empowering readers to master new skills.

Who is it for?

This book is ideal for data scientists, researchers, and managers working with machine learning systems who want to strengthen their security knowledge. It assumes a basic understanding of machine learning, with some benefit from familiarity with Python. If you're looking to secure ML systems and enhance your understanding of ethical data practices, this book is for you.

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.

Read now

Unlock full access

More than 5,000 organizations count on O’Reilly

O’Reilly covers everything we've got, with content to help us build a world-class technology community, upgrade the capabilities and competencies of our teams, and improve overall team performance as well as their engagement.

Julian F.

Head of Cybersecurity

I wanted to learn C and C++, but it didn't click for me until I picked up an O'Reilly book. When I went on the O’Reilly platform, I was astonished to find all the books there, plus live events and sandboxes so you could play around with the technology.

Addison B.

Field Engineer

I’ve been on the O’Reilly platform for more than eight years. I use a couple of learning platforms, but I'm on O'Reilly more than anybody else. When you're there, you start learning. I'm never disappointed.

Amir M.

Data Platform Tech Lead

I'm always learning. So when I got on to O'Reilly, I was like a kid in a candy store. There are playlists. There are answers. There's on-demand training. It's worth its weight in gold, in terms of what it allows me to do.

Mark W.

Embedded Software Engineer

Publisher Resources

ISBN: 9781804618851

Cloud Computing

Data Engineering

Data Science

AI & ML

Programming Languages

Software Architecture

IT/Ops

Security

Design

Business

Soft Skills