- From: Harry Halpin <hhalpin@ibiblio.org>
- Date: Tue, 14 Sep 2010 20:09:12 +0200
- To: public-xg-socialweb@w3.org
>From http://www.w3.org/2010/api-privacy-ws/papers/privacy-ws-22.txt Just cut and pasting it below so people can read before telecon: = Privacy: A Pictographic Approach = Submitted by: Aza Raskin <aza@mozilla.com> Arun Ranganathan <arun@mozilla.com> On behalf of Mozilla Mozilla believes that privacy policies are long documents written in legalese that obfuscate meaning. Nobody reads them because they are indecipherable and obtuse. Yet, these are the documents that tell you what’s going on with your data — how, when, and by whom your information will used. To put it another way, the privacy policy lets you know if some company can make money from information (like selling your email address to a spammer). Creative Commons [1] did an amazing thing for copyright law. It made it understandable. Creative commons reduced the complexity of letting others use your work with a set of combinable, modular icons. In order for privacy to have meaning for actual people, we need to follow in Creative Commons' footsteps. We need to reduce the complexity of privacy policies to an indicator scannable in seconds. Mozilla believes that solving all of the problems with privacy in one go is tilting at windmills. P3P [2] took a taxonomic approach and languished in the exponential complexity. Instead, we should be seeking to answer the question: "What attributes of privacy policies and terms of service should people care about?" and then highlight those attributes. Further, we should only be highlighting attributes of privacy which are not "business as usual" so that users do not become inured by constant warnings. Finally, Mozilla believes that the attributes should be machine readable to encourage user-agent and other innovation. == Only What People Should Care About == The “should” is critical. Privacy policies are often complex documents that deal with subtle and expansive issues. A set of easily understood and universal icons cannot possible encode everything. Instead, we should call out only the attributes which are not “business as usual”: the warning flags that your privacy and data are at risk. Here's an example. Should we have an icon that lets the you know that your data will be shared with 3rd parties? Isn’t 3rd party sharing intrinsically a bit suspect? The answer is a subtle no. Sharing with 3rd parties should raise a warning flag but only if that sharing isn’t required. The classic example is buying a book on Amazon.com and getting it shipped to your home. Amazon needs to share your home address with UPS and Privacy Icons shouldn’t penalize them for that necessary disclosure. In other words, Privacy Icons should only highlight 3rd party data sharing when you do not have a reasonable expectation that your data is being shared. The “should” is a major differentiator from many of the prior approaches, like the taxonomic P3P or Lorrie Cranor’s crowd-sourced Privacy Duck [3]. == Bolt-on Approach == Privacy policies and Terms of Services are complex documents that encapsulate a lot of situation-specific detail. The Creative Commons approach is to reduce the complexity of sharing to a small number of licenses from which you choose. That simply doesn’t work here: there are too many edge-cases and specifics that each company has to put into their privacy policy. There can be no catch-all boiler-plate. We seem to have lost before we have even begun. There’s another approach. Here’s where we stand: companies need to write their own privacy policies/terms of service, replete with company-specific detail. Why? Because a small number of licenses can’t capture the required complexity. The problem is that for everyday people, reading and understanding those necessarily custom privacy policies is time consuming and nigh impossible. Here’s a solution: create a set of easily-understood Privacy Icons that “bolt on to” a privacy policy. When you add a Privacy Icon to your privacy policy it says the equivalent of “No matter what the rest of this privacy policy says, the following is true and preempts anything else in this document…”. The Privacy Icon makes an iron-clad guarantee about some portion of how a company treats your data. For example, if a privacy policy includes the icon for “None of your data is sold or shared with 3rd parties”, then no matter what the privacy policy says in the small print, it gets preempted by the icon and the company is legally bound to never sharing or selling your data. Of course, the set of icons still needs to be decided. Mozilla held a workshop on the 27th of January 2010 to help decide these kinds of questions, and will host further events in the future. == Lawyer Selected, Reader Approved == Since its release, Creative Commons has continually pared down the number of licenses it provides and is now down to just two icons, one with two states and one with three. It has to be so simple because everyday people choose their own license. Privacy Icons don’t have that constraint. A qualified lawyer chooses what icons to bind to their privacy policy, and so there can be substantially more icons to choose from allowing the creation of a rich privacy story. As long as the icons are understandable by an everyday person, we are golden. == Machine Readable == Some of the attributes will have potentially a bad normative value, like an icon that indicates your data may be sold to third parties. The question becomes, why would any company display such an icon in their privacy policy? Wouldn’t they instead opt to not use the Privacy Icons at all? This is the largest problem facing the Privacy Icons idea. Aren’t we are creating an incentive system whereby good companies/services will display Privacy Icons and bad companies/services will not? If attributes become widely adopted then the correlation of good companies using the icons and bad companies not using the icons becomes rather strong. If a privacy policy doesn’t include any icons it runs the risk of becoming synonymous with that policy making no guarantees for not using your data for evil. The absence of Privacy Icons becomes stigmatic. Asking people to notice the absence of something is asking the implausible. People don’t generally don’t notice an absence; just a presence. The solution hinges on Privacy Icons being machine readable and Firefox being used by 400 million people world-wide. If Firefox encounters a privacy policy that doesn’t have called-out attributes, we’ll automatically display the icons with the poorest guarantees: your data may be sold to 3rd parties, your data may be stored indefinitely, and your data may be turned over to law enforcement without a warrant, etc. This way, companies are incentivized to display these attributes and thereby be bound to protecting user privacy appropriately. == Attribute Strawperson == As a strawperson, here are the 7 things that we believe matter most in privacy: * Is you data used for secondary use? And is it shared with 3rd parties? * Is your data bartered? * Under what terms is your data shared with the government and with law enforcement? * Does the company take reasonable measures to protect your data in all phases of collection and storage? * Does the service give you control of your data? * Does the service use your data to build and save a profile for non-primary use? * Are ad networks being used and under what terms? === Secondary Use of Data === *Is your data used for secondary use?* The European Union has spent time codifying and refining the idea of “secondary use”; the use of data for something other than the purpose for which the collectee believes it was collected. Mint.com uses your login information to import your financial data from your banks — with your explicit permission. That’s primary use and shouldn’t be punished. The RealAge [4] tests poses as a cute questionnaire and then turns around and sells your data. That’s secondary use and is fishy. When you sign up to use a service you should care if your data will only be used for that service. If the service does use your data for secondary use, they should disclose those uses. If they share your data with 3rd parties, then they should disclose that list too. === Bartering User Data === *Is your data bartered?* You should know when someone is making a gain off your back. You should also know roughly how and for what that data is being bartered. === Goverment and Law Enforcement as Third Parties === *Under what terms is your data shared with the government and with law enforcement?* Do they just hand it over without a warrant or a subpoena? === Private Data Storage Considerations === *Does the company take reasonable measures to protect your data in all phases of collection and storage.* There are numerous ways that your data can be protected: from using SSL during transmission, to encryption on the server, to deleting your data after it is no longer needed. Does the company protect your data during transmission, storage, and from employees? This icon should tell you what the weak link is. === User Control of Data === *Does the service give you control of your data?* Can you delete your data if you choose? Can you edit it? What level of control do you have over the data stored on their server. === User Profiles for Non-Primary Use === *Does the service use your data to build and save a profile for non-primary use?* This is a subtle one, as we want to include the concept of PII (personally identifiable information). What we are worried about are companies secretly building a dossier on you — say by taking your email address and then buying more information from a 3rd party about that email address to get, say, your credit rating. Then using that profile for uses with which you haven’t agreed. === Use of Advertising Networks === *Are ad networks being used and under what terms?* On the web most pages include ads of some form, and the prevalence of behavioral tracking is on the rise. While letting users get a handle on ad networks is important, raising the alarm on every page would be counter-productive. We haven’t figured out yet how to handle ad networks and are looking for more thought here. == Conclusion == Mozilla believes that it is possible to have a small set of easily understood icons that can represent actionable privacy choices to a user, and is open to discussing standardizing these across user agents. Additionally, we should consider making these icons machine readable, and discuss the best way to do so. [1] http://creativecommons.org/ [2] http://www.w3.org/P3P/ [3] http://lorrie.cranor.org/ [4] http://www.zdnet.com/blog/healthcare/is-realage-a-scam/2040
Received on Tuesday, 14 September 2010 18:09:41 UTC