Just as social sites start to gain or build their target audiences, spammers may rush in and ruin those initiatives.
Resource Shelf noted a recent report from researchers at Stanford University’s Computer Science Department, where they consider methods of combatting spam on places like Flickr, Wikipedia, and Delicious. A threat exists from spammers who wish to deceive or otherwise annoy users of these sites.
Social sites have minimal requirements to register as a user. In some cases an email address and a password choice provide all a spammer needs to join. Once they are in and engaged in spamming, the question revolves around how to mitigate the spammer as a pest.
The researchers summarized how much of a pest such spammers can be to operators of social sites:
Malicious users can mount several different attacks on a social system. If they know that there’s a limit on how much content they can submit, for instance, they can sign up using different identities. If they know that moderators will block users who post spam, they might try to disguise their attacks by contributing a fraction of “good” content. What malicious users do depends on their sophistication, their goals, and on whether they collude. Because malicious users are a moving target, it’s hard to know what they’ll do next.
The easiest method in controlling spammers relies on the manual approach, where other users identify something as spam. Once a certain threshold has been passed, like downvotes on Digg or Reddit, the spam content simply goes away.
Automated prevention, like the ubiquitous captchas seen on forms all over the Internet, help stop robotic spamming. Social websites could also use rank-based methods, similar to how search engines work, to drop spam deep within the bowels of social search results where it is unlikely to be viewed often.
The researchers also looked at tagging and its challenges. Since tagging can be a subjective medium, a spammer could place an inappropriate tag on an item, and when it is clicked, the tag brings up the spammer’s content.
Beating that means developing a spam model, which can be used to quickly identify unwanted spam content. This entails defining “good” tags for a given piece of content, and assessing other tags for their suitability.
The report’s creators noted how pre-existing solutions to spam, coupled with the detailed logs of user interaction maintained by social websites, make those solutions work even better on the social Web. As with anything in security, implementing them means finding a balance between user convenience and adequate spam protections.
