Software To Read Your Lips For Secure Authentication And Access - Mu'tah University
Question asked by Ankita Katdare in #Coffee Room on Oct 5, 2014
Ankita Katdare · Oct 5, 2014
Rank A1 - PRO
With fingerprint sensors only now entering our mobile devices, it's time for other secure authentication techniques to arrive in the latest software on our computers. Keeping up with fast pace in the security techniques development, Ahmad Hassanat from the IT Department at Mu'tah University in Jordan has come up with a software that reads the user's lips to authenticate him for logging into the computer. By making the machine learn how to analyze the pattern in lip & mouth movements, Hassanat could train the system to identify words being uttered correctly about 80% of the time.
In a paper published on this research work, Hassanat shares that the system is a two stage process - setting the visual password stage and the verification stage. By speaking the word into the camera, the system captures the video of the user's face and a words-based VSR system processes these words to extract a sequence of feature vectors. Later, when the user is trying to log-in, these features are compared with the stored visual password and the access is granted.
The appearance of mouth when uttering the word 'Zero'.
Ahmad Hassanat evaluated his system with a video database of 10 males and 10 females and found that system's feasibility is great and has the potential for practical approach thanks to error rate in the range of 7.63% to 20.51% at the worst tested scenario. The author of the paper claims that this new form of visual password would be the newest way for biometric security as even the best of actors would not be able to exactly mimic someone else's lip movements.
What would be the disadvantages of such a system? Let us know your thoughts in comments.
Source: Research paper Posted in: #Coffee Room
In a paper published on this research work, Hassanat shares that the system is a two stage process - setting the visual password stage and the verification stage. By speaking the word into the camera, the system captures the video of the user's face and a words-based VSR system processes these words to extract a sequence of feature vectors. Later, when the user is trying to log-in, these features are compared with the stored visual password and the access is granted.
The appearance of mouth when uttering the word 'Zero'.
Ahmad Hassanat evaluated his system with a video database of 10 males and 10 females and found that system's feasibility is great and has the potential for practical approach thanks to error rate in the range of 7.63% to 20.51% at the worst tested scenario. The author of the paper claims that this new form of visual password would be the newest way for biometric security as even the best of actors would not be able to exactly mimic someone else's lip movements.
What would be the disadvantages of such a system? Let us know your thoughts in comments.
Source: Research paper Posted in: #Coffee Room
Rajni Jain · Oct 5, 2014
Rank B1 - LEADER
Awesome !
However, I am all puzzled if uttering the password would be a safe and secure way of authentication? Also, one can utter the password with different lip movement, how would software recognize that.
Meanwhile, it is a good start for another way of authentication.
However, I am all puzzled if uttering the password would be a safe and secure way of authentication? Also, one can utter the password with different lip movement, how would software recognize that.
Meanwhile, it is a good start for another way of authentication.
Ankita Katdare · Oct 5, 2014
Rank A1 - PRO
@Rajni Jain Yeah, doesn't look full-proof. It seems that any kind of movement (be it fluttering of eye-lids) that is unique to every individual and form a pattern is being tested for secure authentication mechanisms.
Mvraj · Oct 5, 2014
Rank E2 - BEGINNER
okay its good but others can easily hack that what she/he utter some words at that time so how well is your software please tell me madam?
Suraj Patil · Oct 5, 2014
Rank D3 - MASTER
Great! Now we'll speak our passwords, I guess hackers won't need keyloggers anymore.
Research should be focused on making good solutions to real problems, it shouldn't be to create another problem.
It is a good idea, wrong implementation.
May be lip reading can be used to give commands to your cellphone just like voice activated commands.
Research should be focused on making good solutions to real problems, it shouldn't be to create another problem.
It is a good idea, wrong implementation.
May be lip reading can be used to give commands to your cellphone just like voice activated commands.
hitoshianatomi · Oct 6, 2014
Rank E2 - BEGINNER
Whether static or behavioral or electromagnetic, biometrics cannot be claimed to be an alternative to passwords UNTIL it stops relying on a password for self-rescue against the false rejection altogether while retaining the near-zero false acceptance in the real outdoor environment. A dog which depends on a man cannot be an alternative to the man.
Biometrics can theoretically be operated together with passwords in two ways, (1) by AND/conjunction or (2) by OR/disjunction. I would appreciate to hear if someone knows of a biometric product operated by (1). The users of such products must have been notified that, when falsely rejected by the biometric sensor with the devices finally locked, they would have to see the device reset. It is the same with the biometrics operated without passwords altogether.
Biometric products are generally operated by (2) so that users can unlock the devices by passwords when falsely rejected by the biometric sensors. This means that the overall vulnerability of the product is the sum of the vulnerability of biometrics (x) and that of a password (y). The sum (x + y - xy) is necessarily larger than the vulnerability of a password (y), say, the devices with biometric sensors are less secure than the devices protected only by a password.
It is very worrying to see so many ICT people being indifferent to the difference between AND/conjunction and OR/disjunction when talking about “using two factors together”.
Biometrics can theoretically be operated together with passwords in two ways, (1) by AND/conjunction or (2) by OR/disjunction. I would appreciate to hear if someone knows of a biometric product operated by (1). The users of such products must have been notified that, when falsely rejected by the biometric sensor with the devices finally locked, they would have to see the device reset. It is the same with the biometrics operated without passwords altogether.
Biometric products are generally operated by (2) so that users can unlock the devices by passwords when falsely rejected by the biometric sensors. This means that the overall vulnerability of the product is the sum of the vulnerability of biometrics (x) and that of a password (y). The sum (x + y - xy) is necessarily larger than the vulnerability of a password (y), say, the devices with biometric sensors are less secure than the devices protected only by a password.
It is very worrying to see so many ICT people being indifferent to the difference between AND/conjunction and OR/disjunction when talking about “using two factors together”.
