Blog 10: Hackers and Smokers

The first article used social network analysis to understand the network structure of the hacker group “Shadowcrew.” Due to rapid advancements in information technology, the authors were attempting to assess how hackers were interacting and communicating within this deviant group to develop their existing network structure.

Image result for shadowcrew

Research question:  The study had 4 research questions:
1. What is the network centralization of a hacker network?
2. Are there members of a hacker network who stand out as critical leaders?

3. How strongly do leaders influence a hacker network?
4. What subgroups exist and interact in a hacker network?

How is the data collected: A search was conducted using LexisNexis Academia, using the exact keyword Shadowcrew.  LexisNexis Academia includes information from major newspapers, journals, and law/criminal proceedings.  The specific sources of data are listed below and the varying sources of information adds to the strength of the study.  The data was then mined using AutoMap.

What is the sample population:  Twenty three members of Shadowcrew, as identified using the method above, are the sample population.

What are the nodes:    Each node represents an individual hacker (figure 2).

What are the links:  The links represent communication or collaboration between hackers (figure 2).

Results:   There were pronounced differences in measures of centrality among hackers (table 4).

As it is evident in table 4, Andrew Mantovani consistently had  high values on measures of centrality.  Thus, he was identified as a leader.   To add further support, historically he is the co-founder and network administrator for Shadowcrew.  Below is a picture of Andrew.
Image result for andrew mantovani

Andrew had significant influence in Shadowcrew as evident in figure 3.  The people that he is connected to were key players in their group as evident by centrality measures.

Distinct subgroups existed in shadow crew as shown below.

An understanding of SNA allowed for the identification of the leader of Shadowcrew as well as his top lieutenants.   It conveys the power of measures of centrality in identifying salient nodes in a network.    SNA also allowed for the identification of cliques. These findings derived from SNA are important because it allows for law enforcement efforts to be more efficient, targeted towards key players.

 

The second article investigated smoking among adolescents. The authors believed that the position of an adolescent in their peer group would influence smoking.

Research question:  Do adolescents in various social peer group positions have different rates of  prevalence for current smoking?

How is the data collected: A survey was used to obtain smoking status, relationship and peer group information from ninth grade student in the North Carolina School System.

What is the sample population:  One thousand ninety two adolescents ninth grates that attended public school in North Carolina in 1990.

What are the nodes: Each adolescent surveyed is a node in the network.

What are the links:  The links represent a best friend relationship.  Students were asked to named their top three best friends, in decreasing order of priority.

What are the results:  Rates of smoking varied significantly across schools.  As table 2 demonstrates, social position did affect status as current smoker in 4 of the 5 schools.  Isolated adolescents were more likely to be smoker.  The results are mixed for clique members and  liaisons.  Liaisons interact with a clique but they are not members.

By understanding the social network, it became possible to determine which adolescents were isolated, liaisons and clique members.  Using this data the researchers were able to understand current smoking patterns in a social context.

References:

Ennett, S. T., & Bauman, K. E. (1993). Peer Group Structure and Adolescent Cigarette Smoking: A Social Network Analysis. Journal of Health and Social Behavior. https://doi.org/10.2307/2137204

Lu, Yong; Luo, Xin; Polgar, Michael; Cao, Y. (2010). SOCIAL NETWORK ANALYSIS OF A CRIMINAL HACKER COMMUNITY. The Journal of Computer Information Systems. https://doi.org/10.1108/13685201211238016

Blog 9: Social Network Analysis at the Community and Friend Levels

The first study investigated collaborations networks in the social work academic community using social network analysis. Their research question: have collaboration networks for scholars in the social work community changed overtime?  Their sole source of data was the Social Work Research Database which is a comprehensive collection of published social work articles.  The sample population includes scholars of social work that have published in the field (N=19,789) and these scholars compose the nodes of the network.  The links of the network represent collaboration where authors shared authorship on a research article.

Their primary conclusion is that levels of collaborations have increased significantly from 1990 to 2014, and findings were consistent with the Matthew effect.  The Matthew effect can be conceptualized using the adage, “the rich get richer and the poor get poorer.” It has three properties: proportional growth, cumulative advantage of a few nodes, and preferential attachment to central nodes.

As figure 2 demonstrates, there was proportional growth from 1990 to 2014.  The increase in network size provides additional nodes for potentially more edges to be formed.   Figure 2 also supports preferential attachment to central nodes because dark spots more central in the network increased in size and frequency.

Table 3 clearly demonstrates cumulative advantage.  Those authors with higher degrees to start increased from 17.2 to 126.7 degrees, while authors with smaller initial degrees remained small.

Satisfaction of the Matthew effect means that most of the new collaboration in figure 1 in the scholarly social work community occurred for those already advantaged in the social work community.

Quantitative and qualitative methods of social network analysis helped operationalize the data collected in a usable form.  Quantitative measures of degree clearly demonstrated levels of collaboration for each author.  Qualitative analysis of proportional growth (figure 2) allowed us to see that there was preferential attachment to central nodes.

The second articles investigated processes of assimilation and contagion in social networks of adolescent friends.  Their research question: does assimilation and contagion affect symptoms of depression among friend?  The study obtained data from a clinical trial on substance misuse which was conducted for 19 schools in England.  The sample population includes 9-year-old students from 7 of the 19 schools (N=1,114).  The nodes represent students and the links are contagion or assimilation processes among friends.  Consistent with contagion, those adolescents with friends having greater symptoms of depression also found their personal depression symptoms increase.  Assimilation was also supported.  Adolescents were likely to see their depression symptoms change to a level similar to their friends.  The use of contagion and assimilation as tools for social network analysis allowed us to determine the transmission of depression symptoms among adolescent friends.

Bibliography

Markus Eckl, Christian Ghanem, Heiko Löwenstein; The Evolution of Social Work from Disconnected Groups to a Scientific Community: A Social Network Analysis, The British Journal of Social Work, , bcy050, https://doi-org.proxy.library.vcu.edu/10.1093/bjsw/bcy050

Doucet, C., Lacourse, E., Vitaro, F., Stewart, S. H., & Conrod, P. (2016). 6.101 DO FRIENDS MATTER FOR ADOLESCENT DEPRESSION? A LONGITUDINAL SOCIAL NETWORK ANALYSIS. Journal of the American Academy of Child & Adolescent Psychiatry. https://doi.org/10.1016/j.jaac.2016.09.420

 

Blog 8: K-Core Decomposition and Density Mapping of the Cerebral Cortex.

Drawing on my neuroscience background, I chose a study that operationalized k-core decomposition and density to understand the structure of the human cerebral cortex.   Their research question: how can the cerebral cortex be mapped quantitatively to help us understand process and function.  Prior research has looked at connectivity within the cerebral cortex, but they were unable to map out the structural core, network modules and hubs objectively.  By applying k-core decomposition and density analysis neural pathways more at the periphery could be removed in a controlled fashion to provide a distillate of the structures involved in information processing.

The data was collected using diffusion spectrum imaging which allows for a detailed representation of pathways within the human brain.  As it is evident in the picture below, it allows for the obtainment of quantifiable data.

The sample size consisted of 5 participants whose diffusion spectrum images were averaged to increase generalizability.

The nodes represent neural structures involved in information processing and integration.

Image result for neural structure of the brain frontal parietal occipital

The links are axonal pathways that interconnect the brain.

The results indicate the presence of large structural networks within the brain.  Structural cores were revealed in the posterior medial and parietal cortex.  Network modules were present in the temporal and frontal cortex.   Structural cores had high degree and betweenness centrality values and served as connector hubs that linked significant neural structural modules.

While the images and the study may appear simple.  This represents research that could hardly be imagined during my undergraduate studies at Berkeley.  Back then, people were injecting dyes into each axon and following its path to understand interconnections (image below).  Now they are not only able to visualize these connections but obtain quantifiable data directly from the human brains.

Image result for axonal dye

This level of understanding of the human brain is startling and scary.  If the processing power of the brain is fully leveraged in technology,  it can have many unforeseen consequences that cannot be controlled.

It reminds me of what Facebook ran into when developing artificial intelligence (AI).  What they found was that their AI created its own programming language that could not be deciphered by leading computer scientists.  Once this happened the AI was shut down but under less controlled settings it could have resulted in disaster.

References

Hagmann, P., Cammoun, L., Gigandet, X., Meuli, R., Honey, C. J., Van Wedeen, J., & Sporns, O. (2008). Mapping the structural core of human cerebral cortex. PLoS Biology. https://doi.org/10.1371/journal.pbio.0060159

Omatu, S., Malluhi, Q. M., González, S. R., Bocewicz, G., Bucciarelli, E., Giulioni, G., & Iqbal, F. (2015). Distributed computing and artificial intelligence, 12th international conference. In Advances in Intelligent Systems and Computing. https://doi.org/10.1007/978-3-319-19638-1