4 ms·

Show HN: Rushline – Anonymously follow people's private conversations

6 points by danmorrison 10y ago

danmorrison 10y ago
I previously cofounded ITtoolbox, an online community with millions of users when acquired. I was disappointed to learn in most community models, less than 1% contribute their insights, and 99% don't. I believed there had to a better model so I started Rushline. Over 2 years, our team has developed, tested and received patents on a new way of sharing insights. It's based on shared interest conversations among trusted friends with a new concept of privacy. We just completed private beta and are excited to introduce it to you here today. We welcome feedback and questions.
jonny_storm 10y ago
The Rushline frontpage animation suggests anonymization involves dropping avatars and replacing names, but this actually does very little to ensure anonymity. Everything from geographic hints and topic involvement to word choice and punctuation make a person uniquely identifiable, and the vast assortment of public social networking feeds and blogs is a perfect source for training data. Right now, only state actors and service providers can intercept most private communications. It appears to me that Rushline will only lower the bar such that anyone can do so with a bit of Python. Moreover, I strongly doubt corporations, research institutes, or government agencies would permit secrets to be shared openly on any public communications platform, even with the pretense of anonymity. These organizations would be ideal sources of "insight," and you won't have them. For all the things about which people do want to converse openly, there is IRC, Twitter, Reddit, and so on, but on these platforms, it is at least implicitly understood that anonymity is something users create for themselves. Ultimately, you could have made a pretty GUI on top of existing XMPP infrastructure, and there would almost certainly be no difference in the features provided. What value does Rushline actually deliver?
- danmorrison 10y ago
  johnny_storm Thank you for your questions. I'll do my best to answer starting with your last question because I think it informs our thinking on the former. The value we bring with our unique form of privacy is freedom for the common person to share and post publicly with greatly reduced reputation risk and anxiety. Rather than sharing actual secrets (not what we're aiming for), we want people to be comfortable sharing their honest insights and knowledge on topics ranging from work advice to relationships and hobbies. I've talked with many among the 99% who don't openly share their knowledge, and the two most common reasons they give all tie back to a combination of privacy and the fact they prefer to interact with people they know and trust vs. strangers (even though the best answers might've come from someone they don't know). Here's a Medium post sharing more behind my thoughts on what value Rushline is providing. http://bit.ly/2eaZxXB http://bit.ly/2eaZxXB A key difference between anonymity at Rushline vs. personally managed anonymity at Reddit or Twitter is that at Rushline, the core of all conversations are a group of trusted friends who do know and see each other within a conversation. Generally, people who know each other behave differently from strangers, especially true of anonymous strangers. When you see anonymous interaction on Rushline, you're watching people who understand each other and have their best interests in mind. As for re-identifying people through behavior patterns, I suppose this will become increasingly possible, just like video will detect a person by their gate. What kind of challenge it introduces for Rushline will depend on what the landscape looks like as those tools become more readily available. I believe there's a way to address any problem by looking at the pieces in front of you and reading where things are going, and we'll be doing that. At the very simplest level, we start by restricting who has access to Rushline--we don't intend to be publicly searchable or indexable.
  - jonny_storm 10y ago
    I suspect you may be unaware of the ease with which people can be identified by text, alone. Here's a Python library: https://github.com/jpotts18/stylometry https://github.com/jpotts18/stylometry. Of course, this doesn't account for keywords that represent places, but I imagine tools exist for that, as well, and the intersection of several such tools would provide good approximations, indeed. The Facebooks and Googles of the world already perform very sophisticated analyses to track users, and even seemingly innocuous ads (e.g. emitting inaudible sound to correlate devices in the same room) have become insidious. It would be fair to point out that these techniques only yield good results for targeted analysis (you have to provide training data). Yet I think this raises the stakes rather than lowering them, as a targeted analysis would only be conducted by a motivated adversary. Privacy concerns aside, I conjecture that filtering for author experience will more often result in monoculture as the user's own experience increases. A user of 10 years experience seems likely to filter at their own level of expertise, insulating themselves from genuine insights arising from conversations involving non-experts. That said, there are surely ways to correct this; just be aware of the possibility. All in all, I'm sympathetic to your aims. Best of luck.
    - danmorrison 10y ago
      I think your points are all good. "a targeted analysis would only be conducted by a motivated adversary" For the types of conversations we envision, it's hard to see that motivation being very high. It's not likely to reveal much to know who it is that's trying to lose 20 pounds, improve their A1C levels, deal with a problem student as a teacher, or cope with a cheating spouse. These are things we're comfortable sharing with a select group, but we wouldn't necessarily want the world to know we're dealing with. With regard to what we do aim to do, here's a bit of insight into how we look at user privacy (not intending that these alone fully overcome each of the bars you've set, but a summary) We intentionally don't do the following: -We don't use Google Analytics or any big data analytics -We don't use Google, FB, or Twitter sign in -We don't use FB thumbs up or other tracking mechanisms -We don't allow crawling of our database or searching Rushline without being signed in (harder to get large samples for text analysis or re-identification, among other things) We made these decisions in an effort to protect users conversations from being re-identified and combined with the database in the sky. People should be able to talk and read about their personal interests with trusted friends and others, without someone else keeping track of it and adding to their master profile. That's getting much harder everyday, but to the extent we can protect our users from it, we will continue to try our best. Regarding the experience filter, your concern about creating a monoculture makes sense logically. Unfortunately what we learned at ITtoolbox is that when the more experienced people can't protect themselves from certain types of posts, they simply leave the community, taking their insights with them. A far worse outcome. The alternatives to that tend to come in the form of communities with heavy self-policing. There it's novices who feel anxiety and often avoid posting for fear of being yelled at for asking what someone thinks was a poor question. The solution we've created allows experts to easily adjust their own signal to noise threshold. While still not perfect, we think it's a pretty good solution. Newbies and novices get the benefit of seeing all conversation, including that of the experts, while experts can choose to dial in and participate anywhere within the structure they want. Even if they stay mostly up at the top layer of the wedding cake (as we look at it), they're still there and sharing. This is one of the features we felt was important enough to include in our patents pending.

6 points by danmorrison 10y ago

I previously cofounded ITtoolbox, an online community with millions of users when acquired. I was disappointed to learn in most community models, less than 1% contribute their insights, and 99% don't. I believed there had to a better model so I started Rushline. Over 2 years, our team has developed, tested and received patents on a new way of sharing insights. It's based on shared interest conversations among trusted friends with a new concept of privacy. We just completed private beta and are excited to introduce it to you here today. We welcome feedback and questions.

jonny_storm 10y ago

The Rushline frontpage animation suggests anonymization involves dropping avatars and replacing names, but this actually does very little to ensure anonymity. Everything from geographic hints and topic involvement to word choice and punctuation make a person uniquely identifiable, and the vast assortment of public social networking feeds and blogs is a perfect source for training data. Right now, only state actors and service providers can intercept most private communications. It appears to me that Rushline will only lower the bar such that anyone can do so with a bit of Python. Moreover, I strongly doubt corporations, research institutes, or government agencies would permit secrets to be shared openly on any public communications platform, even with the pretense of anonymity. These organizations would be ideal sources of "insight," and you won't have them. For all the things about which people do want to converse openly, there is IRC, Twitter, Reddit, and so on, but on these platforms, it is at least implicitly understood that anonymity is something users create for themselves. Ultimately, you could have made a pretty GUI on top of existing XMPP infrastructure, and there would almost certainly be no difference in the features provided. What value does Rushline actually deliver?

danmorrison 10y ago
johnny_storm Thank you for your questions. I'll do my best to answer starting with your last question because I think it informs our thinking on the former. The value we bring with our unique form of privacy is freedom for the common person to share and post publicly with greatly reduced reputation risk and anxiety. Rather than sharing actual secrets (not what we're aiming for), we want people to be comfortable sharing their honest insights and knowledge on topics ranging from work advice to relationships and hobbies. I've talked with many among the 99% who don't openly share their knowledge, and the two most common reasons they give all tie back to a combination of privacy and the fact they prefer to interact with people they know and trust vs. strangers (even though the best answers might've come from someone they don't know). Here's a Medium post sharing more behind my thoughts on what value Rushline is providing. http://bit.ly/2eaZxXB http://bit.ly/2eaZxXB A key difference between anonymity at Rushline vs. personally managed anonymity at Reddit or Twitter is that at Rushline, the core of all conversations are a group of trusted friends who do know and see each other within a conversation. Generally, people who know each other behave differently from strangers, especially true of anonymous strangers. When you see anonymous interaction on Rushline, you're watching people who understand each other and have their best interests in mind. As for re-identifying people through behavior patterns, I suppose this will become increasingly possible, just like video will detect a person by their gate. What kind of challenge it introduces for Rushline will depend on what the landscape looks like as those tools become more readily available. I believe there's a way to address any problem by looking at the pieces in front of you and reading where things are going, and we'll be doing that. At the very simplest level, we start by restricting who has access to Rushline--we don't intend to be publicly searchable or indexable.
- jonny_storm 10y ago
  I suspect you may be unaware of the ease with which people can be identified by text, alone. Here's a Python library: https://github.com/jpotts18/stylometry https://github.com/jpotts18/stylometry. Of course, this doesn't account for keywords that represent places, but I imagine tools exist for that, as well, and the intersection of several such tools would provide good approximations, indeed. The Facebooks and Googles of the world already perform very sophisticated analyses to track users, and even seemingly innocuous ads (e.g. emitting inaudible sound to correlate devices in the same room) have become insidious. It would be fair to point out that these techniques only yield good results for targeted analysis (you have to provide training data). Yet I think this raises the stakes rather than lowering them, as a targeted analysis would only be conducted by a motivated adversary. Privacy concerns aside, I conjecture that filtering for author experience will more often result in monoculture as the user's own experience increases. A user of 10 years experience seems likely to filter at their own level of expertise, insulating themselves from genuine insights arising from conversations involving non-experts. That said, there are surely ways to correct this; just be aware of the possibility. All in all, I'm sympathetic to your aims. Best of luck.
  - danmorrison 10y ago
    I think your points are all good. "a targeted analysis would only be conducted by a motivated adversary" For the types of conversations we envision, it's hard to see that motivation being very high. It's not likely to reveal much to know who it is that's trying to lose 20 pounds, improve their A1C levels, deal with a problem student as a teacher, or cope with a cheating spouse. These are things we're comfortable sharing with a select group, but we wouldn't necessarily want the world to know we're dealing with. With regard to what we do aim to do, here's a bit of insight into how we look at user privacy (not intending that these alone fully overcome each of the bars you've set, but a summary) We intentionally don't do the following: -We don't use Google Analytics or any big data analytics -We don't use Google, FB, or Twitter sign in -We don't use FB thumbs up or other tracking mechanisms -We don't allow crawling of our database or searching Rushline without being signed in (harder to get large samples for text analysis or re-identification, among other things) We made these decisions in an effort to protect users conversations from being re-identified and combined with the database in the sky. People should be able to talk and read about their personal interests with trusted friends and others, without someone else keeping track of it and adding to their master profile. That's getting much harder everyday, but to the extent we can protect our users from it, we will continue to try our best. Regarding the experience filter, your concern about creating a monoculture makes sense logically. Unfortunately what we learned at ITtoolbox is that when the more experienced people can't protect themselves from certain types of posts, they simply leave the community, taking their insights with them. A far worse outcome. The alternatives to that tend to come in the form of communities with heavy self-policing. There it's novices who feel anxiety and often avoid posting for fear of being yelled at for asking what someone thinks was a poor question. The solution we've created allows experts to easily adjust their own signal to noise threshold. While still not perfect, we think it's a pretty good solution. Newbies and novices get the benefit of seeing all conversation, including that of the experts, while experts can choose to dial in and participate anywhere within the structure they want. Even if they stay mostly up at the top layer of the wedding cake (as we look at it), they're still there and sharing. This is one of the features we felt was important enough to include in our patents pending.