One of the more extraordinary trends in internet use has been the rapid rise of video chat services such as Chatroulette. These services randomly link the webcams of people who visit the site.
But Chatroulette has a problem. The site is dominated by flashers who expose their genitals.
Some 6.3 million visitors used Chatroulette in July 2010, perhaps because of the sexual nature of its content.
But this poses a significant threat to minors. There is no easy way to police the age of people who visit websites and minors can gain access easily. According to Xinyu Xing at the University of Colorado at Boulder and a few pals, a significant number of Chatroulette users appear to be minors.
“Our observations on a typical Saturday night indicate that as many as 20-30% of Chatroulette users are minors,” they say in a paper published today on the arXiv.
Xing and co have a solution, however. This team has developed a “flasher detection” algorithm that spots the offenders, allowing them to be kicked out.
It turns out that catching flashers is harder than it might seem at first. One way is to employ a crowdsourcing mechanism in which users report offenders whose video feed can then be evaluated by trained individuals and stopped if necessary.
But that’s a time consuming and expensive task that is open to abuse. And with upwards of 20,000 users at any time, it’s unlikely to work for Chatroulette in the long run.
Another approach is to use existing algorithms designed to detect pornographic content. Exactly how these algorithms work isn’t entirely clear, but they appear to look for skin content in images.
Unfortunately, this type of software does not work well with video chat content, say Xing and co. That’s because the video images are often poorly lit making it hard to distinguish skin from yellowy-white walls in the background, for example.
So Xing and co have developed their own algorithm, called SafevChat which they’ve tested on some 20,000 stills taken from Chatroulette videos and supplied by the service’s founder Andrey Ternovskiy. Their paper gives a detailed insight into how it works.
The new approach is interesting because it analyses the images using several different criteria and then fuses the results before deciding whether the image is acceptable or not.
To get over the problem of skin-coloured walls and furniture, they combine skin detection with motion detection that compares sequential frames to see whether the”skin” is moving. And they use face, eye and nose detectors to distinguish facial from non-facial skin. The results are fused and the image is then classified as normal or offensive, having been trained on the initial dataset.
Xing and so say it works well and significantly better than a commercial pornographic image detector programme called PicBlock. They have a video showing how it works on their website.
In fact, SafeVChat works so well, that Chatroulette began using it on its website earlier this month.
Whether this makes Chatroulette more or less popular, we’ll have to wait and see. But with any luck, it will make the site safer for all concerned.
Ref: arxiv.org/abs/1101.3124: SafeVchat: Detecting Obscene Content and Misbehaving Users in Online Video Chat Services
Embracing CX in the metaverse
More than just meeting customers where they are, the metaverse offers opportunities to transform customer experience.
Identity protection is key to metaverse innovation
As immersive experiences in the metaverse become more sophisticated, so does the threat landscape.
The modern enterprise imaging and data value chain
For both patients and providers, intelligent, interoperable, and open workflow solutions will make all the difference.
Scientists have created synthetic mouse embryos with developed brains
The stem-cell-derived embryos could shed new light on the earliest stages of human pregnancy.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.