 | « Back to article | Print this article |
Terror outfits like the Lashkar-e-Tayiba thrive on one very major factor -- the element of surprise. Intelligence agencies across the globe have gone wrong on many occasions and the LeT has surprised many when they stage an attack.
Now what if one were able to predict what the Lashkar was up to? Knowing beforehand what they would do would surely go a long way in preventing an attack. Here is a programme known as the Computational Analysis of Terrorist Groups: Lashkar-e-Tayiba, developed by the researchers at the University of Maryland in US, which could act as a predictor of what this group is up to.
The project used algorithms to parse mined data on 770 variables from 20 years of Lashkar's activities. The variables were updated every month for a computational analysis and this gave an understanding of the factors that determined of how often the Lashkar attacks occurred.
During the study it was found that if between 5 and 24 Lashkar operatives were arrested and put on trial, then there was an 88 per cent chance of the group attacking local forces.
V S Subrahmanian, researcher at the Maryland University's Laboratory for Computational Cultural Dynamics, headed this project. He along with his co-author, Aaron James, tell Rediff.com's Vicky Nanjappa about the project and how it helps in acting as a predictor for attacks by the Lashkar.
Sir, for how long have you been working on the book Computational Analysis of Terrorist Groups: Lashkar-e-Tayiba?
For over five years.
Based on data mining and also data on variables over 20 years, it acts as a predictor for different types of terror strikes carried out by the Lashkar. Could you explain to us how this would work?
In our study, we identified 770+ variables describing different types of attacks carried out by the LeT, as well as different aspects of the environment in which LeT carried out these attacks. The values of these variables were identified from reliable open sources for each month, starting from January 1990 to December 2010.
We then learned conditions on the "environmental" variables which satisfied three conditions: (i) when the discovered condition was true in a given month, it was followed by certain types of attacks in the next 1-3 months with high probability (ii) when the discovered condition was false in a given month, it was followed by the same types of attacks in the next 1-3 months with very low probability, and (iii) the condition was true sufficiently often. Thus, the conditions discovered in this way are like the proverbial "canary in the coalmine" providing strong predictors of future attacks (of certain types).
In India there are various agencies which have been formed to study the pattern and mindset of the Lashkar and other terrorist organisations. Do you think they should base their study on the model that you and your team have sought to set up through your study?
Our framework is one of the most data-driven and systematic studies of LeT to date. However, given the seriousness of the LeT threat to India, our model should be combined with other methods that policy makers and analysts in these agencies have already found useful, so that their decisions are guided by the best methods discovered to date.
Security experts often speak about the manner in which a group such as the Lashkar changes its strategy. Is this correct or has the group become predictable in the manner in which it carries out an attack?
All terror groups (including the LeT) change their on-ground tactics in order to evade detection. However, terror groups, including LeT, have organisational goals and processes (as well as inertia) that enable their strategic goals to be more predictable in the short term (six months out, say).
That said, it is very difficult for anyone to accurately predict when and where LeT will carry out terrorist attacks. What our study does is to predict when certain types of attacks will be carried out -- but not exactly where.
How do you think this book would help in reducing or preventing future attacks by the Lashkar?
The conditions that we have identified that are predictive of certain kinds of LeT attacks give decision makers time to take appropriate protective measures against such attacks by beefing up security around the types of locations and situations LeT targets.
Moreover, the policies we propose in the book offer the promise of keeping LeT busy managing their own internal issues, reducing their time and ability to plan terror attacks.
What options would you suggest to reduce the threat of the Lashkar?
The book proposes three major methods. (i) Disrupt the internal cohesion of LeT by fostering internal dissent within the organisation, (ii) Identify carefully calibrated and publicly stated repercussions for the Pakistani government if LeT carries out another terror attack and (iii) Find ways to disrupt LeT press and publicity campaigns as well as LeT training camps.
How serious have you found policy makers especially in India and the United States of America in countering the threat of the Lashkar?
I have found both to be very receptive to data-driven approaches to countering the LeT threat.
What are the main factors which determine an attack by the Lashkar? In brief what have the monthly data on the 770 variables used by you suggested?
The main variables correlated (usually in some combination with each other) with LeT backed attacks are:
The book proposes three major methods: (i) Disrupt the internal cohesion of LeT by fostering internal dissent within the organisation; (ii) Identify carefully calibrated and publicly stated repercussions for the Pakistani government if LeT carries out another terror attack; and (iii) Find ways to disrupt LeT press and publicity campaigns as well as LeT training camps.
The bans, freezing of assets and also the arrests have had some amount of success in weakening the Lashkar. You have also said that fostering the dissent within the Lashkar commanders would help in destablising the outfit. Is there enough being done in this regard by the various agencies across the world?
We have not seen evidence of ongoing efforts to foster dissent amongst LeT's commanders.
Could you also tell us about the machine learning techniques applied by you to the data sets to counter terrorism?
We extended a data mining algorithm we developed and patented about 7-8 years ago to learn conditions of the kind described in response to your second question above.
Broadly the book suggests what lies in store while dealing with the Lashkar and its operatives. How accurate can the predictions be?
This remains to be seen. However, about 15 months ago, we made broad predictions of attack trends by LeT which were over 80 per cent accurate, so we hope the same holds for the future.
However, we are not aware of any prediction methods to date that can accurately pinpoint both when and where an attack will take place.
Image: A burning Taj Mahal Palace Hotel during the 26/11 terror attacks in Mumbai orchestrated by the LeT
Photograph: Arko Dutta/Reuters