Hacker News Reader

Comment by malf | Hacker News Reader

Paper seems to actually show things like “If you install the Fitness skill, Amazon shows you fitness ads”, but that’s less sexy, I guess.

View on HN

I think what it's actually showing (per Section 6.1) is that Amazon is using wishy-washy language to give the impression that they are not using voice recordings to build ad profiles on users[1]. They instead claim that it's for "personalized experiences" and "building inclusiveness".

This adorable little infographic on "the journey of a voice request" conveniently leaves out that it gets used for advertisement[2]. They have also made public statements that outright state that voice data doesn't get used for ad-targeting[3]

[1] https://www.amazon.com/Alexa-Privacy-Hub/b?ie=UTF8&node=1914...

[2] https://www.amazon.com/b/?node=23608618011

[3] In a statement, Amazon said the company took “privacy seriously” and did “not use customers’ voice recordings for targeted advertising.” https://www.nytimes.com/2018/03/31/business/media/amazon-goo...

> They have also made public statements that outright state that voice data doesn't get used for ad-targeting[3]

“not use customers’ voice recordings for targeted advertising.”

I guess it depends on how one reads that quote. A trusting sort could read that to mean, we don't use anything we learn from voice recordings for targeted advertising.

A skeptic might read that quote and determine:

Well we generate metadata from the recording, and we then use the metedata for targeted advertising, but we don't use the actual recording for advertising.

Which makes sense, if I was to implement something like this, I wouldn't use the actual recording, I'd process the recording(which I have to do anyway to answer the request) and if I happen to save some useful for advertising data along the way, well, more $$'s for me!

Which one is true? I guess it mostly depends on how hungry Amazon is to make a buck and what they think they can get away with. As a privacy snob, I'd prefer the trusting version to be true.

NSA uses the same tricky language. Clapper got before the American people and said "nobody is reading your emails." What he didn't say is, computers are processing them because there's too many for people to read. In this case, I suspect Amazon is technically telling the truth, and automatic transcription into text is what drives their ad engine.

Yes exactly, the text is arguably data derived from voice recordings. So they can say that they don't use the voice recordings for ads, while omitting that they do use the derived text that was recognized for ads. It's unclear language but I wouldn't be surprised if this was the correct interpretation.

99% of people who read your comment (including me) won't have read the paper. You should cite the parts of the paper which made you come to the conclusion that they really only showed that ads are shown based on what you install, not based on what you say otherwise.

Reading the introduction section should be enough. It’s pretty clear that they’re talking about Amazon targeting ads based on your interactions with the Alexa and not, as the title here implies, based on just listening in to other things you say.

The paper does talk about “voice data” which I think is a bit misleading. “Voice data” to me would imply an analysis of the sounds your voice makes directly for ad targeting purposes but it’s clear enough from the context what they actually mean.

Amazon does plenty of rubish things already, no need for us to make up extra things!

From the introduction, it seems like their main concern is that "smart speakers record audio from their environment and potentially share this data with other parties over the Internet—even when they should not". They provide two examples of how that happens:

1. "Smart speaker vendors or third-parties may infer users’ sensitive physical (e.g., age, health) and psychological (e.g., mood, confidence) traits from their voice."

2. "The set of questions and commands issued to a smart speaker can reveal sensitive information about users’ states of mind, interests, and concerns."

They also mention that "smart speaker platforms host malicious third-party apps", and "record users’ private conversations without their knowledge", but that's mentioned as examples of prior research and thus seems to serve more as background than something this paper is trying to prove.

Point 2 is the one you're focusing on, and yeah, that's not surprising. You'd expect Amazon to build a profile on you based on the stuff you ask Echo to do (though the ethics of this certainly warrants discussion).

Point 1 would be the surprising thing, that smart speakers infer information about people from their voice, rather than from the commands themselves.

Their methodology seems to be to create multiple personas and compare the sorts of ads they get. In order to prove that information is inferred from traits of the voice rather than the words in the commands, they would need two personas which are identical in which commands they send but with different voices (female vs male voice, healthy vs smoker voice, something like that). From skimming section 3, it doesn't seem like they did that, so I'm forced to agree that the thing they prove in this paper (if their statistical methods are valid) is that Amazon builds an advertisement profile based on your interests as expressed in terms of which commands you're sending the device.

bullet 2 is at odds with your initial statement. The data collected are from interactions with the smart speaker. Here is the opening sentence of the abstract: "Abstract—Smart speakers collect voice input that can be used to infer sensitive information about users".

Reply on news.ycombinator.com