, 19 tweets, 20 min read Read on Twitter
#lca2019 @attacus_au on the non-consensual use of public tech use, and how to be evade it.

spoiler: reality is already scarier than scifi.

this is a live system adding metadata - pants vs skirt, direction of travel - of pedestrians
@attacus_au #lca2019 @attacus_au this chinese system matches faces to ones it knows, at a public escalator - v high accuracy, e.g. people travel the same way to work
@attacus_au #lca2019 @attacus_au researcher joy buolomwini found her face wasn’t picked up by face classifiers due to her dark skin, so tested a few algorithms. high accuracy: M presenting light skin, low accuracy F presenting dark skin.
@attacus_au #lca2019 @attacus_au ah yes, the gaydar face algorithm.

algorithms are made by biased humans, who are very bad at algorithm and experimental design.

i.e. don’t believe the hype
@attacus_au #lca2019 @attacus_au: @WestfieldAU shopping malls install cameras, in advertising screens, data used to profile demographics of users looking at it.
@attacus_au @WestfieldAU #lca2019 @attacus_au Have You Been Naughty Or Nice, the creepy church edition
@attacus_au @WestfieldAU #lca2019 @attacus_au facial recognition aimed at answering: who is this person?

how much do we trust the algorithm that is trying to answer that question?
@attacus_au @WestfieldAU #lca2019 @attacus_au
facial recognition systems have 3 parts:

detection
faceprint creation
recognition
@attacus_au @WestfieldAU #lca2019 @attacus_au
detection methods: meshes between dots, or projections
@attacus_au @WestfieldAU #lca2019 @attacus_au recognition:
DeepFace, an ill-advised name is 98% accurate according to FB

and then there are convolutional neural networks. lilly tested one with masks and fake beards - yep, it works
@attacus_au @WestfieldAU #lca2019 @attacus_au
and now, for the disappearance part:
- if you’re a hacker in a stock photo, wear a balaclava NOPE
- print a mask of a face - 2D or 3D YEP
- pull a weird face YEP
- $680NZ scarf YEP
- papparazzi visor YEP but a pain to wear
@attacus_au @WestfieldAU #lca2019 @attacus_au …or just wear a flashlight on your forehead, which dazzles the camera
@attacus_au @WestfieldAU #lca2019 @attacus_au in WWI adam harvey tried to camouflage ships to disguise their speed, size and direction of travel
@attacus_au @WestfieldAU #lca2019 @attacus_au show the cameras things that look like a lot of faces, i.e. fake news yourself
@attacus_au @WestfieldAU #lca2019 @attacus_au according to this neural network:
banana = banana
banana + sticker = toaster
@attacus_au @WestfieldAU #lca2019 @attacus_au shining infrared light in different patterns onto someone’s face tricked the FR to think it was four different people, including moby
@attacus_au @WestfieldAU #lca2019 @attacus_au the algorithms used in facial recognition are not
- independently audited
- tested for bias
- treated as speculative and as experimental as they actually are
@attacus_au @WestfieldAU #lca2019 @attacus_au The Australian #aabill Is Terrible News, the lilly edition: reasons too many to capture
@attacus_au @WestfieldAU #lca2019 @attacus_au
more skpeticism
more clarity
more auditing
Missing some Tweet in this thread?
You can try to force a refresh.

Like this thread? Get email updates or save it to PDF!

Subscribe to maia sauren @ linuxconf chch
Profile picture

Get real-time email alerts when new unrolls (>4 tweets) are available from this author!

This content may be removed anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!