Twitter: Spammers vs. Users
1Monday, November 19, 2012 by Unknown
Information infrastructure involves, among other things, the people that support the creation, use, and dissemination of information in a given system. That being said, it also involves the people who would destroy that information, or tear down/subvert the system, and the tools, processes, etc., that allow them to do so.
After reading the article on "Detecting spam in a Twitter network", I became very interested in the ways in which spammers adapt to various information networks and infrastructures. The aspects of Twitter that make it so appealing to users - the aggregation of "trending" topics, the simple dissemination of links, and the ability to "follow" people without permission - are the same aspects that allow spammers to infiltrate it. As the study shows, there is not much difference between a user account and a spam account on Twitter, though there may be small ways to detect hazardous accounts.
The network analysis method employed by the authors seems to have been quite effective at separating out these small differences. It is interesting to see the implications that this type of analysis can have for better informed system design. When we know about the ways in which a network or system is abused by users, can't we adjust the design in order to thwart their attempts? On the one hand, there is the sense that spammers will always find a way to persist; they are invariably part of the infrastructure. On the other, this type of research can help a network or system to evolve, and to set up provisions.
My technical knowledge is limited, but in the case of Twitter, it seems as though bad link detection or "automatic follow" detection could be employed. This is along the lines of the means used by the authors to detect spammers. I'm not completely sure if I know what I'm talking about. I also realize that it is difficult to adjust the system or infrastructure too much, because you run the risk of discouraging real users as well.
After reading the article on "Detecting spam in a Twitter network", I became very interested in the ways in which spammers adapt to various information networks and infrastructures. The aspects of Twitter that make it so appealing to users - the aggregation of "trending" topics, the simple dissemination of links, and the ability to "follow" people without permission - are the same aspects that allow spammers to infiltrate it. As the study shows, there is not much difference between a user account and a spam account on Twitter, though there may be small ways to detect hazardous accounts.
The network analysis method employed by the authors seems to have been quite effective at separating out these small differences. It is interesting to see the implications that this type of analysis can have for better informed system design. When we know about the ways in which a network or system is abused by users, can't we adjust the design in order to thwart their attempts? On the one hand, there is the sense that spammers will always find a way to persist; they are invariably part of the infrastructure. On the other, this type of research can help a network or system to evolve, and to set up provisions.
My technical knowledge is limited, but in the case of Twitter, it seems as though bad link detection or "automatic follow" detection could be employed. This is along the lines of the means used by the authors to detect spammers. I'm not completely sure if I know what I'm talking about. I also realize that it is difficult to adjust the system or infrastructure too much, because you run the risk of discouraging real users as well.
Twitter's infrastructure is so interesting, and while I would never be able to conduct any sort of formal examination of it given my limited understanding, I think it's proven ability to disseminate information easily really makes it very attractive to spammers.
ReplyDeleteCompared to most other social networks, there is a cultural expectation of openness and transparency that seems fairly inherent to Twitter. I'd be hard-pressed to name many friends on Facebook with fully public accounts, but that is generally the norm on Twitter (this is of course a generalization, but I've observed this through anecdotal experience). The infrastructure of Twitter, and the cultural norms that have sprung up around it, promote openness - most accounts are public, public accounts can be followed by anyone with no need for reciprocation or approval, replies can be sent to those you follow (regardless of whether they follow you); the infrastructure promotes connectivity in a very unprecedented way. All of these, of course, makes trolls and spammers' lives much easier too.
It's a very fine line; as you say, any drastic change to the infrastructure to lessen spam accounts would fundamentally alter the culture of Twitter.