Coverage of Computer-Related Crime 1980-2010
One of the problems facing security personnel is that the number of published reports on vulnerabilities, exploits and successful attacks is growing so fast that it’s difficult to keep up.
Now, one of the instructions I give my students as they work on term papers is that assertions like the one I just wrote have to be backed up with evidence. I decided to do a little original research that would support or disprove the assertion about growth in reports.
In a quick survey to gather initial data, I used the Kreitzberg Library resources at Norwich University to look at the number of articles in a collection of electronic databases that included key words relating to computer crime over a period of years.
This initial query is not a rigorous research study: it’s just a first step to get a sense of the situation. The information garnered can lead to further research and also provide a sense of the scale of changes in coverage of computer crime over more than a decade. I also planned in advance to perform linear regressions of articles and of proportions vs. year.
First, I selected the following databases, which include some overlap in sources:
• ABI/INFORM GLOBAL (coverage 1971-present; 3,640 titles)
• ProQuest Computing (1998-present; 511 titles)
• ProQuest Criminal Justice (1981-present; 442 titles)
• ProQuest Newspapers (coverage varies; 54 titles)
• ProQuest Science Journals (1994-present; 1,596 titles)
After excluding duplicate journal titles, there were 5,609 titles in the aggregate database. A rough estimate of the total number of articles (by searching on string “the”) was 42,388,781. I searched all the articles using the following expression:
hacker
OR
“computer virus”
OR
((vulnerability OR exploit OR attack)
AND
(computer OR network))
and got 77,901 hits. I then recorded the numbers of articles by year and the total number of articles using “the” per year and prepared a table and a chart of the numbers of articles with the keywords and their proportion of the total articles. This table shows the data:
The next figure shows the graph of the number of hits per year:
And the final figure shows the percentage of the database articles with hits on the search string:
The regression analysis for hits per year generated with Excel 2010 Data | Data Analysis | Regression shows an extremely significant linear regression with p(H0)=1.68 x 10^(-9):
The regression equation has a slope showing an average increase of about 179 articles per year over 30 years; the 95% confidence limits for the slope are about 136 to 221. The coefficient of determination (“R Square”) suggests that about 72% of the total variation in hits per year can be explained using the linear regression.
The percentage of articles in the database with hits on the search string shows an extremely significant regression of percentage on year with p(H0)=1.56×10^(-6):
The regression equation suggests that the percentage of coverage has been rising about 0.007% per year (95% confidence limits are between roughly .005% and 0.010% per year). However, the coefficient of determination (about 55%) shows that the regression is not as effective an explanation of the growth in percentage as the regression was for growth in numbers of articles.
In my next article, I’ll discuss some useful resources for keeping up with the growing tide of articles about information security. Readers wanting to learn more about applied statistics can freely download a review article and a draft version of a short introductory textbook from my Website.
