Portal Overlapping Community Detection

According to Scanlon and Gerber (2014), Internet has facilitated the terrorist communities and groups to recruit violent extremists and expand their scale of activities and size of their groups. The potential of the online communities is more as they can use and exchange various online resources for the purpose of carrying out illegal activity.

This article said that in order to classify forum posts certain criteria needs to be fulfilled. The data must be collected from popular VE groups, the data should be within a time span like one decade and the data must be in English language or some language that can be translated in English. This article identified Dark Web Portal to be a storage space of messages from 28 online forums that focus on Islamic discussions and extremist religious (Ríos and Munoz 2012). The forum posts from Ansar1 have been collected and pre processed by manually annotating the collected information as ‘a’ if it is related to VE recruitment and as ‘b’ if it a non VE recruitment post. This article focused on an analytical approach to find out whether forum post is meant for VE recruitment or not. A probability model has been used.

Techniques Used in the Solution

Classification functions like naïve bayes, classification trees, boosting, logistic regression and support vector machines (SVM) were used for the purpose of classifying forum posts. The annotated information was utilized by randomly segmenting the data. The data was broken into ten folds (Scanlon and Gerber 2014). Cross validation have been applied. The classification methods have been evaluated by ROC curves. The ROC curves are known for showing trade-offs between FPR and TPR. An area under ROC curve called AUC has been employed for the purpose of comparing the performance of each method by making the use of single measure. The best performance was given by SVM classifier with 0.89 AUC. This article also mentioned the accuracy variation of all the methods. The range was 0.2 to 0.3 AUC. SVM showed the least variation in its performance. Boosting showed the highest variation. Turkey’s range test has been used for the purpose of determining the difference of AUC among the models. Classification trees method gave the worst AUC performance. The task of classification is best performed by logistic regression as well as SVM methods. The result of this article provides performance benchmark for comparing it with future methods. This article provided a clear evidence of the fact that the conflicts that took place in Somalia and Nigeria were the main topic of forum posts in Ansar1 data. Logit and SVM methods can detect any VE recruitment with accuracy where the mean AUC is greater than 0.85. The main aim of this research report was to use the methods of data collection and analytical efforts for developing supervised learning methods for the purpose of identifying the process of cyber recruitment that is carried out by violent extremists. This report suggested that the detection methods can be improved in the future by incorporating non English language in the detection as well as classification methods.

According to Jiongcong et al. (2016), cyber attack has the ability to cause critical hazards to the economic as well as secure operations and functioning of the power system. This research report has focused on the influence of FDIA on power systems. FDIA is a type of cyber attack where malicious data is injected into the meters. It results in fake estimations and it stops the detection of bad data (Liang et al. 2017).

This paper proposes two types of scenarios. One scenario is talks about a signal attack that is fake and secure. The other scenario talks about a fake signal attack that is insecure in nature. The first scenario is dangerous as it deceives the operator of the system and shows that the system is secure. The second scenario will initiate the operator to take corrective steps that are not necessary like load shedding and rescheduling (Jiongcong et al. 2016). This will affect the system performance and consume a lot of time. Security assessment is an essential need of the organizations that is used for monitoring and controlling the power system. This paper was intended to perform economy and security analysis of SSA. Two main assumptions under the methodologies used in this paper are: the attacker completely knows about system parameters etc, the attacker has the capability to falsify analog measurements. State estimator is known for providing state variables depending on the meter measurement combination. FDIA is analyzed under AC and DC models.

Techniques used in the Solution

In the first scenario, faulty condition of the open circuit is the base case. A mathematical model is proposed that can be solved by the attacker to convert insecure signal to secure signal. In the second scenario, unnecessary rescheduling is carried out. Here the main focus of the attacker is on online SSA. A mathematical model is formulated for the purpose of injecting malicious data. This scenario of fake insecure signal can lead to unnecessary load shedding as well. The attacker performs manipulation even after the rescheduling process. This paper uses differential evolution or DE method for the purpose of solving the fake signal problem. DE method is a heuristic algorithm. The FDIA problem has been simulated on SSA on modified benchmark system of IEEE-39.  Real power flows are compared in case of a fake secure signal. The fault condition of the open circuit is said to take place on 30h line of transmission that is present in the bus system of IEEE-39.  The overload of the real power system is of major concern. In case of a fake insecure signal problem or attack the attacker manipulates normal situations for the purpose of overloading situations by the method of solving equations that are based on corresponding measurements. The attacker inserts malicious codes into the original power flow to cause fake overload. The online SSA will then send insecure signal to carry out rescheduling. The value of load scheduling can be calculated by using the DE algorithm. This paper showed that unnecessary load shedding increases the cost by 3.71X10^6 dollar. This paper concludes that several researches on cyber attacks on power system would be helpful in enhancing the security of power system.

References

Jiongcong, C.H.E.N., Liang, G., Zexiang, C.A.I., Chunchao, H.U., Yan, X.U., Fengji, L.U.O. and Junhua, Z.H.A.O., 2016. Impact analysis of false data injection attacks on power system static security assessment. Journal of Modern Power Systems and Clean Energy, 4(3), pp.496-505.

Liang, G., Zhao, J., Luo, F., Weller, S. and Dong, Z.Y., 2017. A review of false data injection attacks against modern power systems. IEEE Transactions on Smart Grid.

Ríos, S.A. and Muñoz, R., 2012. Dark Web portal overlapping community detection based on topic models. In Proceedings of the ACM SIGKDD Workshop on Intelligence and Security Informatics.

Scanlon, J.R. and Gerber, M.S., 2014. Automatic detection of cyber-recruitment by violent extremists. Security Informatics, 3(1), p.5.