Monthly Discussion Banner

What are your thoughts on facial "un-recognition" technologies and what will be needed to ensure the DoD's continued ability to identify potential terrorists using facial recognition technology?

Discussion Leader: 
Dr. Thirimachos Bourlai
Posted Date: 8/24/2015

This week’s thought-provoking question is based on a popular HDIAC Spotlight about Japan's new "privacy visor" uses light reflection to thwart facial recognition technology. It is not the first or only technology aimed at bypassing facial recognition software, and as these technologies advance, new research and measures will be needed for continued biometric identifications. 

Japan’s “privacy visor” uses light reflection to thwart facial recognition technology.

Disclaimer:

The appearance of external hyperlinks in this publication does not constitute endorsement by the HDIAC or the United States Department of Defense (DoD) of the linked sites, nor the information, products, or services contained therein. The HDIAC is a DOD sponsored Information Analysis Center (IAC), with policy oversight provided by the Assistant Secretary of Defense for Research and Engineering (ASD(R&E)), and administratively managed by the Defense Technical Information Center (DTIC). Reference herein to any specific commercial products, process, or services by trade name, trademark, manufacturer, or otherwise, does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or the HDIAC. Any views or opinions expressed in the Question of the Week do not represent those of HDIAC, DTIC or the DoD.

Comments

The method in which it works is by throwing light around the face in such a way that it is pushed outside the parameters of the current facial recognition algorithms. While this might be effective for current automatic tagging software used in Facebook and the like, the ability to hinder DoD based facial technology is unimpressive. For starters, a new algorithm that uses the predictability of the glasses' effects on facial highlighting could be used as a marker or flag to indicate that the glasses are being used and the photo needs further analysis. This, at the very least, makes it possible to flag the image as having a face, but one that is not currently identifiable. At that point, considering these glasses are semi-transparent, reconstructing and comparing should not be that difficult. Considering the glasses are only effective at fooling the current algorithm, simply expanding the parameters, or writing a new algorithm to include the predictable effects of the glasses, should be enough to deem them useless.

As things are now, these un-recognition glasses do is nothing that a traditional mask could not do, though it is more fashionable and less likely to attract as much unwanted attention on a busy street than the latter. I see more use for these glasses as means for celebrities and politicians who might not want photos taken of them to be as quickly identifiable. Using them to commit a crime seems a stretch at this time, as they do not make you completely unidentifiable just yet.

thirimachos's picture

Dear all, before I introduce myself, I would like to thank Mrs. Jessica Hill for giving me the opportunity to take part in the discussions related to this topic. My name is Thirimachos Bourlai and I am an Assistant Professor at LCSEE, WVU. My main research fields are biometrics and face recognition, including the pitfalls of the technology when it is used under controlled and uncontrolled conditions.

I have recently red about the "privacy visor" and I agree with the above comments (subject "Pitfalls"). It is very possible to design and develop algorithms that can efficiently detect faces in the crowd and, then, also detect glasses like that on faces wearing them. However, as it was also mentioned above, this will also mean that, now, in surveillance applications, the subject wearing the glasses can be flagged as non-identifiable. As a result, the subject is likely to attract more attention.

In desktop or mobile applications where somebody wants to verify their identity, face recognition is one of the biometrics that can be used. In the same applications, if someone does not want to unveil their personal information to others and to protect their internet privacy, they can simply cover their laptop, cell phone, or desktop computer camera lens rather than spend 240 dollars for such glasses.

TB

Mr. Bourlai, after seeing that you have expertise within this field of research, I definitely was interested in your response regarding this technology.

It is definitely thought-provoking to see that this type of technology is being brought to the surface. While this privacy visor is viewed as an obvious and unattractive way to hide their identify, do you think this could lead to the development of smaller, more unrecognizable applications, such as having contact lenses that are able to reflect, refract and absorb light? That way, if one was able to not be identified by facial recognition technologies, the operators would lean towards thinking the fault was within the technology, not the person.

I do agree that right now, these lenses are viewed as nothing short of ridiculous and will undoubtedly attract a large amount of attention. I will definitely be interested to see if any research continues to develop regarding this technology.

I am curious to see how these technologies will develop. As it's been said, these glasses are pretty obvious, but what happens when they look like ordinary, every day frames? Also, I am curious about Amsterdam's facial un-recognition makeup concept (discussed: http://innovation.avg.com/2015/03/02/protecting-your-visual-identity-in-...).

Right now it all seems like an exaggeration on wearing sunglassess and a hoodie to protect your identity, but if the technologies become more sleek and less identifiable, then perhaps research into biometric identifications other than facial recogntion will need to be advanced.

It's not surprising that people are already attempting to circumvent the relatively new use of facial recognition. I would think that this development should even further promote the use of multi-modality biometric devices for identification. While the glasses might be able to fool the facial or iris recognition component of the device, it would not be able to also fool identification by fingerprint, gait, heart beat and/or vein recognition.

Ultimately, it's curious how and why many people choose to selectively protect certain aspects of their lives and identities.  Anyone that uses the internet on a regular basis readily gives up their information to companies, including Google or Amazon, without thinking twice.  While these glasses provide another opportunity to potentially protect a single aspect of your identity they are not entirely effective and will still leave you vulnerable to other biometric collection methods.  As mentioned in the previous posts, there are other tools and modalities to collect biometric data.  I would also expect we will see similar capabilities emerge that are aimed at protecting or obscuring those individual biometric aspects as well.

As technology continues to advance towards a biometric approach, it provides new opportunities for society to move forward in security, while still retaining a reasonable level of convenience. Facial recognition technology, as well as other technologies based on physical and behavioral characteristics, are used in numerous commercial and consumer applications including protected access to computer/phone systems, social networking to identify friends, security systems for banking/credit accounts, tailoring products to fit individual users, target marketing/advertising, personalized customer services, etc. Various governments worldwide are using or considering using this technology to improve security and surveillance, including electronic passports, nationality ID cards, and identification databases. 

As biometric technologies progress, we are faced with tough decisions: on one hand, there are many opportunities that the technology can provide, but it is also important  to protect privacy of individuals.

Some applications of this technology offer better control of privacy, like joining a social media site or using identification feature options instead of passwords on smartphones or banking accounts; however many applications do not offer the public any control over their privacy, and many people are uncomfortable with the collection, usage,and storage of information that lends itself to reducing anonymity and increasing traceability. 

There are some potential federal laws that apply to facial recognition technology, but do not address all privacy concerns and do not expressly regulate the use of this technology (GAO, 2015).

Facial recognition technology offers many advantages, but the technology must be improved, and new technology continues to be developed, to protect privacy as well as improve security.