Re: [Mailman-Developers] [GSoC14] Full Anonymization Project Idea
Hi I am a prospective Gsoc'15 applicant. I found the Anonymous mailing list idea for project interesting. While researching about the same I found this thread. I have a queries regarding this : Can't we keep our encryption application away from the site admins by creating an interface between sender and the switchboard ,what I'm trying to imply here is, can't we process the encryption before the mail goes to switchboard or server for further processing.
Let say,the email sent by the sender gets into the interface first where it strips out the message header and encrypts the email address, so when the email goes to the switchboard it is already modified, does not contain original sender's address but has the details about where it is to be sent.
Now when the receiver sends the reply then the email gets to the list moderator and after that it passes through the same interface which now decrypts it to the original address and forwards it after encrypting the sender's address. Stephen J. Turnbull writes:
Rashi Karanpuria writes:
Site admins (the people who install the software) have to have root privileges. Nothing can be hidden from them. The only way to protect your data from them is to send it encrypted with a key they don't have.
But it's hard to imagine why you would use an anonymous list, or any mailing list, hosted by people you don't trust enough to give them the keys, to send mail to people you know and trust well enough to give them those keys. The basic motivation for encrypted lists is avoiding the key distribution problem, either because it's annoying or because the point of the list is that the subscribers are not to know each other at all.
I think I might be missing a point here. What i understand is, we don't need to protect the data in the message from the admin, but the information about sender of the data from the list admin as well as other subscribers of that list (i.e. the email address and other details in header which includes traces of sender's identity). In an anonymous list we don't aim to build trust relationships or import the hierarchy maintained in an establishment where we put the decisive power and right to know everything in the hands of the admin/root. We aim at discussing things sharing the bitterest of views unbiased from personal identities and positions in an establishment.
A possible use case might be:
A suggestion and discussion list of an organisation where subscribers could post all problems and issues and slackenings in the organisation's structure and authorities without worrying about their names being involved. They could even use the list in decision making process, internal polling or maybe even reviewing the system.
Similarly a psychiatrist's list where patients can discuss there issues and arrive at mutual solutions with help from the doctor, while bonding with people of their kind. Some syndromes are not cured beacuse of social stigmas that could find a possible solution in such lists.
Rashi Karanpuria writes:
A possible use case might be:
If the president of the company doesn't have root, he can fire whoever does and replace them with somebody who will do what he says. Not going to work unless you trust the site admin *and* her boss. Note that they also have access to MTA and firewall logs, and so probably know who made connections when.
If you don't trust your psychiatrist and her office staff, you are in very big trouble. No need to protect the list from the list admin, but you might want to protect it from the site admin if the list is hosted by a third party.
Security (including privacy) of electronic communication is brain- breaking hard stuff. Note: I am *not* saying "don't do this project" -- past experience shows that the level of coding needed for proof of concept is about right for GSoC -- but you need to be very careful not to make overly optimistic claims that you're actually protecting anyone's privacy.
When I was mentoring Abhilash's project in 2013, I thought quite carefully about possible applications and came to the conclusion that the crypto was useless. One use-case you might find interesting was student evaluations of classes they took. Either the students trust the faculty not to peek, or they don't, and in the latter case there just isn't any sure way to protect the students' privacy *and* at the same time prevent ballot-box stuffing (ie, the students need to authenticate). The traditional method (check ids at the door and give one marksheet to each student to fill in, then have a student collect the marksheets and deliver them to the Faculty Development Office) is much closer to airtight.
OTOH, if you *do* have that level of trust in the list and site admins, encrypting traffic does make sense (otherwise mail is an open book to relaying hosts), and encrypting with your key is as good a signature as encrypting a hash of the message (the usual "digital signature").
Rashi Karanpuria writes:
Site admins (the people who install the software) have to have root privileges. Nothing can be hidden from them. The only way to protect your data from them is to send it encrypted with a key they don't have.
But it's hard to imagine why you would use an anonymous list, or any mailing list, hosted by people you don't trust enough to give them the keys, to send mail to people you know and trust well enough to give them those keys. The basic motivation for encrypted lists is avoiding the key distribution problem, either because it's annoying or because the point of the list is that the subscribers are not to know each other at all.
I think I might be missing a point here. What i understand is, we don't need to protect the data in the message from the admin, but the information about sender of the data from the list admin as well as other subscribers of that list (i.e. the email address and other details in header which includes traces of sender's identity). In an anonymous list we don't aim to build trust relationships or import the hierarchy maintained in an establishment where we put the decisive power and right to know everything in the hands of the admin/root. We aim at discussing things sharing the bitterest of views unbiased from personal identities and positions in an establishment.
A possible use case might be:
A suggestion and discussion list of an organisation where subscribers could post all problems and issues and slackenings in the organisation's structure and authorities without worrying about their names being involved. They could even use the list in decision making process, internal polling or maybe even reviewing the system.
Similarly a psychiatrist's list where patients can discuss there issues and arrive at mutual solutions with help from the doctor, while bonding with people of their kind. Some syndromes are not cured beacuse of social stigmas that could find a possible solution in such lists.
Rashi Karanpuria writes:
A possible use case might be:
If the president of the company doesn't have root, he can fire whoever does and replace them with somebody who will do what he says. Not going to work unless you trust the site admin *and* her boss. Note that they also have access to MTA and firewall logs, and so probably know who made connections when.
If you don't trust your psychiatrist and her office staff, you are in very big trouble. No need to protect the list from the list admin, but you might want to protect it from the site admin if the list is hosted by a third party.
Security (including privacy) of electronic communication is brain- breaking hard stuff. Note: I am *not* saying "don't do this project" -- past experience shows that the level of coding needed for proof of concept is about right for GSoC -- but you need to be very careful not to make overly optimistic claims that you're actually protecting anyone's privacy.
When I was mentoring Abhilash's project in 2013, I thought quite carefully about possible applications and came to the conclusion that the crypto was useless. One use-case you might find interesting was student evaluations of classes they took. Either the students trust the faculty not to peek, or they don't, and in the latter case there just isn't any sure way to protect the students' privacy *and* at the same time prevent ballot-box stuffing (ie, the students need to authenticate). The traditional method (check ids at the door and give one marksheet to each student to fill in, then have a student collect the marksheets and deliver them to the Faculty Development Office) is much closer to airtight.
OTOH, if you *do* have that level of trust in the list and site admins, encrypting traffic does make sense (otherwise mail is an open book to relaying hosts), and encrypting with your key is as good a signature as encrypting a hash of the message (the usual "digital signature").
participants (2)
-
Rashi Karanpuria -
Stephen J. Turnbull