S. Karthikeyan

Bio: S. Karthikeyan is an academic researcher from PSNA College of Engineering and Technology. The author has contributed to research in topics: Image segmentation & Video tracking. The author has an hindex of 14, co-authored 55 publications receiving 1316 citations. Previous affiliations of S. Karthikeyan include University of California, Santa Barbara & Institute of Medical Sciences, Banaras Hindu University.

Malware images: visualization and automatic classification

Abstract: We propose a simple yet effective method for visualizing and classifying malware using image processing techniques. Malware binaries are visualized as gray-scale images, with the observation that for many malware families, the images belonging to the same family appear very similar in layout and texture. Motivated by this visual similarity, a classification method using standard image features is proposed. Neither disassembly nor code execution is required for classification. Preliminary experimental results are quite promising with 98% classification accuracy on a malware database of 9,458 samples with 25 different malware families. Our technique also exhibits interesting resilience to popular obfuscation techniques such as section encryption.

Assessment of neurocognitive impairment after off-pump and on-pump techniques for coronary artery bypass graft surgery: prospective randomised controlled trial.

Abstract: Objective: To assess neurocognitive impairment after the off-pump and on-pump techniques for coronary artery bypass graft surgery in patients with triple vessel disease. Design:Randomised controlled trial. Setting: University Hospital of Wales, Cardiff. Participants: 60 patients undergoing coronary artery bypass graft surgery for triple vessel disease prospectively randomised to the off-pump or on-pump technique. Main outcome measures: Change in scores in nine standard neuropsychometric tests administered preoperatively and at 1 and 10 weeks postoperatively. Results: The on-pump group showed a significantly greater deterioration in scores for two and three tests at 1 week and 10 weeks postoperatively, respectively, than the off-pump group. The on-pump group also showed a significantly higher incidence of major deterioration in one of the tests both 1 week and 10 weeks postoperatively. The incidence of neurocognitive impairment at 1 week postoperatively was 27% (8 out of 30) in the off-pump group and 63% (19 out of 30) in the on-pump group (P=0.004); and at 10 weeks postoperatively was 10% (3 out of 30) in the off-pump group and 40% (12 out of 30) in the on-pump group (P=0.017). Conclusion: Off-pump coronary artery bypass graft surgery results in less neurocognitive impairment than the on-pump technique.

Weakly Supervised Localization Using Deep Feature Maps

Abstract: Object localization is an important computer vision problem with a variety of applications. The lack of large scale object-level annotations and the relative abundance of image-level labels makes a compelling case for weak supervision in the object localization task. Deep Convolutional Neural Networks are a class of state-of-the-art methods for the related problem of object recognition. In this paper, we describe a novel object localization algorithm which uses classification networks trained on only image labels. This weakly supervised method leverages local spatial and semantic patterns captured in the convolutional layers of classification networks. We propose an efficient beam search based approach to detect and localize multiple objects in images. The proposed method significantly outperforms the state-of-the-art in standard object localization data-sets.

Eye tracking assisted extraction of attentionally important objects from videos

Abstract: Visual attention is a crucial indicator of the relative importance of objects in visual scenes to human viewers. In this paper, we propose an algorithm to extract objects which attract visual attention from videos. As human attention is naturally biased towards high level semantic objects in visual scenes, this information can be valuable to extract salient objects. The proposed algorithm extracts dominant visual tracks using eye tracking data from multiple subjects on a video sequence by a combination of mean-shift clustering and Hungarian algorithm. These visual tracks guide a generic object search algorithm to get candidate object locations and extents in every frame. Further, we propose a novel multiple object extraction algorithm by constructing a spatio-temporal mixed graph over object candidates. Bounding box based object extraction inference is performed using binary linear integer programming on a cost function defined over the graph. Finally, the object boundaries are refined using grabcut segmentation. The proposed technique outperforms state-of-the-art video segmentation using eye tracking prior and obtains favorable object extraction over algorithms which do not utilize eye tracking data.

Investigating the Efficiency of Blowfish and Rejindael (AES) Algorithms

Abstract: The growth rate of the internet exceeds than any other technology which is measured by users and bandwidth. Internet has been growing at a rapid rate since its conception, on a curve geometric and sometimes exponential. Today, the Internet is moving exponentially in three different directions such as size, processing power, and software sophistication making it the fastest growing technology humankind has ever created. With the rapid growth of internet, there is need to protect the sensitive data from unauthorized access. Cryptography plays a vital role in the field of network security. Currently many encryption algorithms are available to secure the data but these algorithms consume lot of computing resources such as battery and CPU time. This paper mainly focuses on two commonly used symmetric encryption algorithms such as Blowfish and Rejindael. These algorithms are compared and performance is evaluated. Experimental results are given to demonstrate the performance of these algorithms.

On-Pump versus Off-Pump Coronary-Artery Bypass Surgery

Abstract: BACKGROUND Coronary-artery bypass grafting (CABG) has traditionally been performed with the use of cardiopulmonary bypass (on-pump CABG). CABG without cardiopulmonary bypass (off-pump CABG) might reduce the number of complications related to the heart–lung machine. METHODS We randomly assigned 2203 patients scheduled for urgent or elective CABG to either on-pump or off-pump procedures. The primary short-term end point was a composite of death or complications (reoperation, new mechanical support, cardiac arrest, coma, stroke, or renal failure) before discharge or within 30 days after surgery. The primary long-term end point was a composite of death from any cause, a repeat revascularization procedure, or a nonfatal myocardial infarction within 1 year after surgery. Secondary end points included the completeness of revascularization, graft patency at 1 year, neuropsychological outcomes, and the use of major resources. RESULTS There was no significant difference between off-pump and on-pump CABG in the rate of the 30-day composite outcome (7.0% and 5.6%, respectively; P = 0.19). The rate of the 1-year composite outcome was higher for off-pump than for on-pump CABG (9.9% vs. 7.4%, P = 0.04). The proportion of patients with fewer grafts completed than originally planned was higher with off-pump CABG than with on-pump CABG (17.8% vs. 11.1%, P<0.001). Follow-up angiograms in 1371 patients who underwent 4093 grafts revealed that the overall rate of graft patency was lower in the off-pump group than in the on-pump group (82.6% vs. 87.8%, P<0.01). There were no treatment-based differences in neuropsychological outcomes or short-term use of major resources. CONCLUSIONS At 1 year of follow-up, patients in the off-pump group had worse composite outcomes and poorer graft patency than did patients in the on-pump group. No significant differences between the techniques were found in neuropsychological outcomes or use of major resources. (ClinicalTrials.gov number, NCT00032630.)

Deep Learning in Mobile and Wireless Networking: A Survey

Abstract: The rapid uptake of mobile devices and the rising popularity of mobile applications and services pose unprecedented demands on mobile and wireless networking infrastructure. Upcoming 5G systems are evolving to support exploding mobile traffic volumes, real-time extraction of fine-grained analytics, and agile management of network resources, so as to maximize user experience. Fulfilling these tasks is challenging, as mobile environments are increasingly complex, heterogeneous, and evolving. One potential solution is to resort to advanced machine learning techniques, in order to help manage the rise in data volumes and algorithm-driven applications. The recent success of deep learning underpins new and powerful tools that tackle problems in this space. In this paper, we bridge the gap between deep learning and mobile and wireless networking research, by presenting a comprehensive survey of the crossovers between the two areas. We first briefly introduce essential background and state-of-the-art in deep learning techniques with potential applications to networking. We then discuss several techniques and platforms that facilitate the efficient deployment of deep learning onto mobile systems. Subsequently, we provide an encyclopedic review of mobile and wireless networking research based on deep learning, which we categorize by different domains. Drawing from our experience, we discuss how to tailor deep learning to mobile environments. We complete this survey by pinpointing current challenges and open future directions for research.

Seed, expand and constrain: Three principles for weakly-supervised image segmentation

Abstract: We introduce a new loss function for the weakly-supervised training of semantic image segmentation models based on three guiding principles: to seed with weak localization cues, to expand objects based on the information about which classes can occur in an image, and to constrain the segmentations to coincide with object boundaries. We show experimentally that training a deep convolutional neural network using the proposed loss function leads to substantially better segmentations than previous state-of-the-art methods on the challenging PASCAL VOC 2012 dataset. We furthermore give insight into the working mechanism of our method by a detailed experimental study that illustrates how the segmentation quality is affected by each term of the proposed loss function as well as their combinations.