owasp-top-5-machine-learning-risks.md

---

title        : Owasp Top 5 Machine Learning risks
type         : working-session      # working-session, user-session
track        : OWASP Projects
topics       : ["Owasp Project", "AI & ML"]
technology   :
categories   : ["OWASP projects"]
featured     :
when_day     : Fri
when_time    : AM-1
room_layout  :                    #
room_id      : room-2
session_slack: https://os-summit.slack.com/messages/CAVABULJF
status       : review-content              # draft, review-content, done
description  :
organizers   :
    - Talal Albacha
    - Jean-Noël Colin
    
participants :
    - Sebastien Deleersnyder

---

Why

Deep Learning and Machine Learning become vital part of critical systems like self-driving cars, advanced authentication and automated detection of lesions/tumors. However, research shows that such technologies have inherent risks originated from the process of how the models are being learnt or used. In this session we will learn about OWASP project (Top 5 Machine Learning Risks) which tries to identify and document these risks in general, and then we will discuss one case study about specific risk and how to address it.

What

• Top 5 Machine Learning Risks Project Introduction

  • project team

  • update about current state of document

• Developing attacks against machine learning models.

  • Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning (Chen et al. 2017)

Outcomes

Define risk rating approach for this type of attacks and suggest defence techniques

Who

• Application security professionals

• AI professionals

Working materials

• project documentation file

• paper file https://arxiv.org/abs/1712.05526

• https://www.owasp.org/index.php/OWASP_Top_5_Machine_Learning_Risks

• https://owaspsummit.org/Outcomes/machine-learning-and-security/machine-learning-and-security.html