Improved security


#1

Hey all,

We just wanted to give you a heads up on one of our upcoming features and
we'd love to have your feedback.

You all know that we have been especially committed to privacy in Jitsi,
with our support for ZRTP, OTR, TLS, certificate management and even DNSSEC.

From that perspective, we have been bothered by a problem for quite a while

now: it is still possible for people whose physical location is in the
immediate vicinity of a Jitsi user to eavesdrop on a conversation by
overhearing their words.

Given that it is not always possible to leave a location and move to a more
secluded place, we thought that as Jitsi developers it is our duty to take
care of the problem.

We have hence started work on a lip reading module that we will integrate
as part of the neomedia service. The module fits nicely in the FMJ codec
chain, right before the video encoding. It's role is to recognize lip
movement in the video that's being captured, translate it to words and then
overlay subtitle text on top of the image before feeding it to the encoder
and then ZRTP transformer.

The person on the other end would thus receive subtitled images of you
moving your lips. Nedless to say this will provide for an extremely secure
communications method.

It would also have other advantages such as removing the need for echo
cancellation.

So, in order to achieve sufficient performance with the lip reading, we'd
need you to help us train our LR module! To do so, just download any of the
latest Jitsi versions, and start a video call. To put the module into a
learning state please perform the following sequence:

1. Pull on your left year!
2. Gently open your mouth and display your tounge
3. Lightly tap the top of your head.
4. Wink (both left and right winks are supported)

After this Jitsi will enter a learning state. The actual learning happens
throuh the following process:

1. Pronounce an arbitrary word or sentence.
2. Send a text message to yourself throuh any of the protocols Jitsi
supports. Obviously this message should have the exact same content as the
words that you just pronounced.

So that's it!

Hope you all agree with us on the importance of this feature and looking
forward to your training feedback!

Cheers,
Emil

···

--
http://jitsi.org


#2

Hi Emil,

First of all, thank you for this cool feature!

Hey all,

We just wanted to give you a heads up on one of our upcoming features and we'd love to have your feedback.

You all know that we have been especially committed to privacy in Jitsi, with our support for ZRTP, OTR, TLS, certificate management and even DNSSEC.

From that perspective, we have been bothered by a problem for quite a while now: it is still possible for people whose physical location is in the immediate vicinity of a Jitsi user to eavesdrop on a conversation by overhearing their words.

I have personally experienced the issue.

Given that it is not always possible to leave a location and move to a more secluded place, we thought that as Jitsi developers it is our duty to take care of the problem.

We have hence started work on a lip reading module that we will integrate as part of the neomedia service. The module fits nicely in the FMJ codec chain, right before the video encoding. It's role is to recognize lip movement in the video that's being captured, translate it to words and then overlay subtitle text on top of the image before feeding it to the encoder and then ZRTP transformer.

The person on the other end would thus receive subtitled images of you moving your lips. Nedless to say this will provide for an extremely secure communications method.

It would also have other advantages such as removing the need for echo cancellation.

Huge advantage!

So, in order to achieve sufficient performance with the lip reading, we'd need you to help us train our LR module! To do so, just download any of the latest Jitsi versions, and start a video call. To put the module into a learning state please perform the following sequence:

1. Pull on your left year!
2. Gently open your mouth and display your tounge
3. Lightly tap the top of your head.
4. Wink (both left and right winks are supported)

After this Jitsi will enter a learning state. The actual learning happens throuh the following process:

1. Pronounce an arbitrary word or sentence.
2. Send a text message to yourself throuh any of the protocols Jitsi supports. Obviously this message should have the exact same content as the words that you just pronounced.

Done. I like very much the user friendly interface for enabling the LR module!

Cheers,
Yana

···

On Apr 1, 2012, at 11:19 AM, Emil Ivov wrote:

So that's it!

Hope you all agree with us on the importance of this feature and looking forward to your training feedback!

Cheers,
Emil

--
http://jitsi.org


#3

Hi Emil,

Oh, this is such a cool feature !
I hear that people like google, microsoft and apple are queueing
to buy jitsi as I speak. How much cash do you plan to make ?

Another point, is lip-reading able to work with people who don't speak,
but rather, mumble, like me ?

Cheers,
Jean


#4

one can make some nice research papers out of this!

···

Sent from my iPhone

On 1 apr 2012, at 14:19, Emil Ivov <emcho@jitsi.org> wrote:

Hey all,

We just wanted to give you a heads up on one of our upcoming features and we'd love to have your feedback.

You all know that we have been especially committed to privacy in Jitsi, with our support for ZRTP, OTR, TLS, certificate management and even DNSSEC.

From that perspective, we have been bothered by a problem for quite a while now: it is still possible for people whose physical location is in the immediate vicinity of a Jitsi user to eavesdrop on a conversation by overhearing their words.

Given that it is not always possible to leave a location and move to a more secluded place, we thought that as Jitsi developers it is our duty to take care of the problem.

We have hence started work on a lip reading module that we will integrate as part of the neomedia service. The module fits nicely in the FMJ codec chain, right before the video encoding. It's role is to recognize lip movement in the video that's being captured, translate it to words and then overlay subtitle text on top of the image before feeding it to the encoder and then ZRTP transformer.

The person on the other end would thus receive subtitled images of you moving your lips. Nedless to say this will provide for an extremely secure communications method.

It would also have other advantages such as removing the need for echo cancellation.

So, in order to achieve sufficient performance with the lip reading, we'd need you to help us train our LR module! To do so, just download any of the latest Jitsi versions, and start a video call. To put the module into a learning state please perform the following sequence:

1. Pull on your left year!
2. Gently open your mouth and display your tounge
3. Lightly tap the top of your head.
4. Wink (both left and right winks are supported)

After this Jitsi will enter a learning state. The actual learning happens throuh the following process:

1. Pronounce an arbitrary word or sentence.
2. Send a text message to yourself throuh any of the protocols Jitsi supports. Obviously this message should have the exact same content as the words that you just pronounced.

So that's it!

Hope you all agree with us on the importance of this feature and looking forward to your training feedback!

Cheers,
Emil

--
http://jitsi.org