How Risky Are Real Users IFTTT Applets? Camille Cobb ( - PowerPoint PPT Presentation

How Risky Are Real Users’ IFTTT Applets? Camille Cobb ( @camPossible) , Milijana Surbatovich, Anna Kawakami, Mahmood Sharif, Lujo Bauer, Anupam Das, Limin Jia

Smart-home devices can lead to risks End-User Programming 2

Smart-home devices can lead to risks 3

“applet” this that + + if then “trigger” “action” 4

SECRECY VIOLATION if then SmartThings post a sensor message is closed to Slack 19,323 unique INTEGRITY VIOLATION applets [Surbatovich et al. 2017] 5 5

SECRECY VIOLATION if then SmartThings post a sensor message is closed to Slack 50% Violating INTEGRITY 50% VIOLATION Safe [Surbatovich et al. 2017] 6 6

Moving from theory to practice if then SmartThings post a ● What fraction of users’ sensor message IFTTT applets are is closed to Slack violating, in practice? ● How much and what Mailbox? To coworkers? types of harm are Main entrance? To my family? Door to a safe? Just to me? IFTTT users actually exposed to? 7

We collected 743 rules from 28 IFTTT users if then SmartThings post a sensor message “If front Door is closed to Slack Sensor closed then post a message to a Slack Mailbox? To coworkers? service” [P28] Main entrance? To my family? Door to a safe? Just to me? 8

Evaluating participants’ rules: automated analysis finds similar results 59% Violating 41% Safe 9

Evaluating participants’ rules: considering context enables more accurate analysis Are all of these harmful? No! 59% Violating 743 unique rules Are any of 41% these potentially Safe harmful? 10

Evaluating participants’ rules: considering context enables more accurate analysis Are all of these harmful? No! 59% Violating 743 unique rules Are any of SmartThings add a row to 41% if then these potentially Google sensor Safe harmful? is closed spreadsheet 11

Evaluating participants’ rules: considering context enables more accurate analysis Are all of these harmful? Also safe No! 59% Violating 743 unique only ~14% rules Are any of 41% harmful these potentially Safe harmful? (not 59%) 12

Evaluating participants’ rules: considering context enables more accurate analysis Are all of these harmful? Also safe No! 59% Violating 743 unique only ~14% rules Are any of 41% harmful these potentially Safe harmful? (not 59%) Yes! 13

Evaluating participants’ rules: considering context enables more accurate analysis Are all of these harmful? if Alice’s then create Also safe No! 59% presence is journal Violating detected entry 743 unique only ~14% rules Are any of 41% harmful these potentially Safe harmful? (not 59%) Yes! 14

Evaluating participants’ rules: considering context enables more accurate analysis 20% surveillance Are all of these risks for incidental harmful? users Also safe No! 59% Violating 743 unique only ~14% rules Are any of 41% harmful these potentially Safe harmful? (not 59%) Yes! 15

How Risky Are Real Users’ IFTTT Applets? ● Real user data → New insights about risks & harms ○ Many “violating” rules are not harmful ○ “Non-violating” rules could be harmful ○ More in the paper! Camille Cobb ( @camPossible) , Milijana Surbatovich, Anna Kawakami, Mahmood Sharif, Lujo Bauer, Anupam Das, Limin Jia