Research paper in NDSS 2010: Improving Spam Blacklisting Through Dynamic Thresholding and Speculative Aggregation

Our blacklist paper titled:

Improving Spam Blacklisting Through Dynamic Thresholding and Speculative Aggregation

and authored by:

Sushant Sinha, Michael Bailey, and Farnam Jahanian
University of Michigan, Ann Arbor, MI - 48109.

is going to be presented in Network and Distributed System Security (NDSS) Symposium, 2010 from 28th Febuary to 3rd March in San Deigo, California.

Here is the abstract:

Abstract:

Unsolicited bulk e-mail (UBE) or spam constitutes a significant
fraction of all e-mail connection attempts and routinely frustrates
users, consumes resources, and serves as an infection vector for
malicious software. In an effort to scalably and effectively reduce
the impact of these e-mails, e-mail system designers have increasingly
turned to blacklisting. Blacklisting (blackholing, block listing) is a
form of course-grained, reputation-based, dynamic policy enforcement
in which real-time feeds of spam sending hosts are sent to networks so
that the e-mail from these hosts may be rejected. Unfortunately,
current spam blacklist services are highly inaccurate and exhibit
both false positives and significant false negatives. In this paper, we
explore the root causes of blacklist inaccuracy and show that the
trend toward stealthier spam exacerbates the existing tension between
false positives and false negatives when assigning spamming IP
reputation. We argue that to relieve this tension, global aggregation
and reputation assignment should be replaced with local aggregation
and reputation assignment, utilizing preexisting global spam
collection, with the addition of local usage, policy, and reachability
information. We propose two specific techniques based on this premise,
\emph{dynamic thresholding} and \emph{speculative aggregation}, whose
goal is to improve the accuracy of blacklist generation. We
evaluate the performance and accuracy of these solutions in the
context of our own deployment consisting of 2.5 million production
e-mails and 14 million e-mails from spamtraps deployed in 11 domains
over a month-long period. We show that the proposed approaches
significantly improve the false positive and false negative rates when
compared to existing approaches.

Merit and Justice -- Amartya Sen

Today population is divided into two portions: one who vehemently oppose selection by merit and believe it is constructed as another tool to advance control by privileged class and the other group who believes meritocracy as a whole is the only morally sound principal for advancing society. What has caused this significant difference in perception? Is the difference in perception because:
1. losers in a meritocratic system oppose it and the winners support it?
2. one group is observing the advantaged elite and the other group focusing on importance of society to progress?
Or, Is there more in the debate?

Off course this debate has stirred academia to identify the points of justification and the points of oppose. Amartya Sen identifies societal foundations of meritocracy and its limitations in this first chapter of the book "Meritocracy and Economic Inequality". Sen argues that merit system involves rewarding good actions in a society so that it could lead to good consequences in the society. For example, giving police job and good remuneration to a person who is identified to be knowledgeable in enactment of law will cause better enforcement of laws and will benefit the society as a whole (mine own). So "meritocracy is just an extension of a general system of rewarding merit and elements of such a system clearly have been present in one form or another throughout human history".

However, rewarding merit is conditional on what we consider good in the society. For example, in a totalitarian society or a country with a foreign rule may consider enforcement of arbitrary laws, and rewarding and recruiting best people for law enforcement to be bad for the society (mine own). Similarly Sen argues that if removing economic inequality is good for a society then the society may reward those positions and actions that have distributive outcomes.

A second point that Sen argues is that the actual value of reward is separated from the action for it to have any productive role. For example, paying something for his knowledge of law enactment in a specific case has no productive aspect (and just trade aspect) and only when we assign the job of enforcement of all laws to a person with some demonstateable actions will have any productive value. Similarly, choosing students for teaching engineering cannot involve testing them for engineering and has to be something different. Sen argues that people are skeptical of the system when the difference between reward and the measure is revoltingly high. I am actually not able to combine the above two arguments well: Isn't taking account of an unrelated objective like reducing economic disparity or providing social justice into the selection of a police officer actually widens the difference between the criteria of selection and the work required of a police officer. So can we put arbitrarily any objective into the rewarding system?

Some additional arguments that may creep into judging merit:

1. Personification and genetics: Rewarding of good action by a person changing into a label of 'meritorious' or personification of the person.

2. Desert and entitlement: Rewarding good action does not necessarily mean that a person "intrinsically deserves" to get more. It has to be judged by what is good in the society which is external to the person.

3. Distribution independence: A system of rewarding merits may well generate inequalities of well being and of other advantages. But much would depend on the nature of consequences that are sought on the basis of which merits are to be characterized. If the results desired have a strong distributive component, then in assessing merit concerns about distribution and inequality would enter the evaluation.