Sateesh Kumar Peddoju

Energy efficient computing- Green cloud computing

Moving towards Cloud Computing, high performance computing usage of huge data center (DC) and huge cluster is increasing day by day and energy consumption by these DC and energy dissipation in environment by these DC is also rising day by day. The large amount of CO2 dissipation in environment has generated the necessity of Green computing (saving energy by recycling it and reusing it over a period of time and minimizing the wastage in terms of usage of resources). More processor chips generates more heat, more heat requires more cooling and cooling again generates heats and thus we come to a stage where we want to balance the system by getting the same computing speed at decreased energy consumption. In this paper we proposed different ideas towards green cloud computing approach.

Malware Detection Using Network Traffic Analysis in Android Based Mobile Devices

Smart phones, particularly Android based, have attracted the users community for their feature rich apps to use with various applications like chatting, browsing, mailing, image editing and video processing. However the popularity of these devices attracted the malicious attackers as well. Statistics have shown that Android based smart phones are more vulnerable to malwares compared to other smart phones. None of the existing malware detection techniques have focused on the network traffic features for detection of malicious activity. To the best of our knowledge, almost no work is reported for the detection of Android malware using its network traffic analysis. This paper analyzes the network traffic features and builds a rule-based classifier for detection of Android malwares. Our experimental results suggest that the approach is remarkably accurate and it detects more than 90% of the traffic samples.

Handoff Strategy for Improving Energy Efficiency and Cloud Service Availability for Mobile Devices

The increase in capabilities of mobile devices to perform computation tasks has led to increase in energy consumption. While offloading the computation tasks helps in reducing the energy consumption, service availability is a cause of major concern. Thus, the main objective of this work is to reduce the energy consumption of mobile device, while maximising the service availability for users. The multi-criteria decision making (MCDM) TOPSIS method prioritises among the service providing resources such as Cloud, Cloudlet and peer mobile devices. The superior one is chosen for offloading. While availing service from a resource, the proposed fuzzy vertical handoff algorithm triggers handoff from a resource to another, when the energy consumption of the device increases or the connection time with the resource decreases. In addition, parallel execution of tasks is performed to conserve energy of the mobile device. The results of experimental setup with opennebula Cloud platform, Cloudlets and Android mobile devices on various network environments, suggest that handoff from one resource to another is by far more beneficial in terms of energy consumption and service availability for mobile users.

Energy Efficient Seamless Service Provisioning in Mobile Cloud Computing

Accessing Cloud via mobile device proves to be costly because of issues with wireless network. Due to communication overhead, offloading of application execution to Cloud consumes more energy than executing in the device itself. This paper proposes a novel framework in which application execution is offloaded to both Cloud and mobile ad hoc Cloud, in order to reduce this communication overhead. Distributed/Parallel execution of tasks is done to reduce the waiting time of mobile device, provided the cost of offloading is less, compared to cost of executing the application in device. Seamless service provisioning is achieved in this framework, by measuring the signal strength of the wireless medium. Once the signal strength threshold is reached, interim results are got from the device to which task is offloaded. This paper proposes the framework for energy efficient seamless service with features like, connecting heterogeneous mobile devices to form mobile ad hoc Cloud, service discovery in mobile ad hoc Cloud and offloading decisions.

Energy efficient mobile vision system for plant leaf disease identification

Close monitoring, proper control and management of plant diseases are essential in the efficient cultivation of crops. This paper presents a scheme that uses mobile phones for real-time on-field imaging of diseased plants followed by disease diagnosis via analysis of visual phenotypes. A threshold based offloading scheme is employed for judicious sharing of the computational load between the mobile device and a central server at the plant pathology laboratory, thereby offering a trade-off between the power consumption in the mobile device and the transmission cost. The part of the processing carried out in the mobile device includes leaf image segmentation and spotting of disease patch using improved k-means clustering. The algorithm is simple and hence suitable for Android based mobile devices. The segmented image is subsequently communicated to the central server. This ensures reduced transmission cost compared to that in transmitting full leaf image.

Response time based load balancing in cloud computing

Rapid growth of the users on Cloud services combined with the growth in the number of services provided to user increases the load on the Cloud servers multifold. This problem becomes more critical when some of the Cloud servers are under loaded and some overloaded. This necessitates an effective Load Balancing technique that can serve the purpose of not only properly utilizing the servers but also reducing the negative impact on the user services. The existing Load Balancing techniques suffer from various issues like (i) balancing the load after a server has been overloaded, (ii) constant querying the server about availability of its resources, hence, increasing computation costs and bandwidth consumption. This paper proposes an algorithm that takes a preventive approach of Load Balancing by considering only the response time of the each request. Based on the response time, the proposed method decides the allocation of next incoming request. The approach is not only dynamic in nature, but also reduces the communication and extra computation on each server. The algorithm is implemented and tested. Results prove the performance of proposed algorithm.

Behaviour analysis of machine learning algorithms for detecting P2P botnets

Botnets have emerged as a powerful threat on the Internet as it is being used to carry out cybercrimes. In this paper, we have analysed some machine learning techniques to detect peer to peer (P2P) botnets. As the detection of P2P botnets is widely unexplored area, we have focused on it. We experimented with different machine learning (ML) algorithms to compare their ability to classify the botnet traffic from the normal traffic by selecting distinguishing features of the network traffic. Experiments are performed on the dataset containing the traces of various P2P botnets. Results and tradeoffs obtained of different ML algorithms on different metrics are presented at the end of the paper.

Detection of Insider Attacks in Cloud Based e-Healthcare Environment

In recent years, Cloud computing has been receiving great attention from various business and research organizations as it promises to provide large storage facilities and highly managed remote services. Due to its characteristics like on-demand self service, rapid elasticity, ubiquitous network access and resource pooling, it shows high potential for providing e-Healthcare solutions. It can offer various financial and functional benefits to e-Healthcare which includes providing storage flexibility for the rapidly growing healthcare data, reduced cost, better accessibility, improved quality of care and enhancement in medical research. However at the same time, it faces many technical challenges like privacy, reliability, security etc. In the Cloud based ehealthcare environment where the patients data is transferred between entities, maintaining the security of data becomes a priority. Cryptographic techniques can only provide a secure channel of communication but it fails to provide security at end points. Security attacks may be accomplished by the malicious insider at the end points. A malicious insider may modify the patients data resulting in a false examination. The paper provides a detective approach for such attacks in the healthcare organizations. Our work is focused with the detection of insider attacks for preventing false examination of patients health records and assuring the accountability of data usage. Watermarking can be used for detection of modification by an insider attack but does not provide accountability of data usage. Hence our approach combines the functionalities of cryptographic techniques and watermarking together with an accountability framework for providing transparency of patients data usage.

Improved Detection of P2P Botnets through Network Behavior Analysis

Botnets are becoming powerful threats on the Internet because they launch targeted attacks towards organizations and the individuals. P2P botnets are resilient and more difficult to detect due to their nature of using different distributed approaches and encryption techniques. Classification based techniques proposed in the literature to detect P2P botnets, report high overall accuracy of the classifier but fail to recognize individual classes at the similar rates. Identification of non-bot traffic is equally important as that of bot classes for the reliability of the classifier. This paper proposes a model to distinguish P2P botnet command and control network traffic from normal traffic at higher rate of both the classes using ensemble of decision trees classifier named Random Forests. Further to optimize the performance, this model also addresses the problem of imbalanced nature of dataset using techniques like downsampling and cost sensitive learning. Performance analysis has been done on the proposed model and evaluation results show that true positive rate for both botnet and legitimate classes are more than 0.99 whereas false positive rate is 0.008.

Minimizing Network Traffic Features for Android Mobile Malware Detection

Smartphones have emerged as one of the dominant computing platforms in todays era where Android has been the first choice for users as well as app developers due to its open source nature and feature rich apps. Such popularity has come hand-in-hand with an equivalent increase in malware targeting Android. Since mobile devices allow easy-to-use, touch-sensitive, and anywhere-anytime access to its resources, the mobile-specific applications like SMS, MMS, Bluetooth, e-mail, and other services may pose serious threats and lead to financial losses and privacy leakages. In recent time, high attention is drawn by the researchers for detecting Android malware; very fewer community have considered network traffic analysis in their detection models. The majority of these models have considered the detection primarily on traffic features that distinguish malware traffic from normal traffic. In this paper, we have proposed an algorithm to prioritize network traffic features with an aim to minimize the number of features to be analyzed to give high detection accuracy along with reduced training and testing time. To this extent, we have used statistical tests to rank the features. Results demonstrate that using prioritized features for detection not only reduces the training and testing time, but also gives slightly higher detection accuracy than using all the features together by measuring Fmeasure, a widely used measure for detection accuracy. The training time of 300 applications is reduced from 11.7 seconds to 5.8 seconds and testing time of 230 applications is reduced from 25.1 seconds to 17.3 seconds, hence reduction of around 50% and 31% in training time and testing time respectively. We believe this time difference will have a larger impact if there are thousands of files to be tested.