Kesav Kancherla

Kernel machines for malware classification and similarity analysis

In this paper we present a method of functionally classifying malicious code that might lead to automated attacks and intrusions using kernel machines. We study the performance of kernel methods in the context of robustness and generalization capabilities of malware classification.

Non Intrusive and Extremely Early Detection of Lung Cancer Using TCPP

In this paper, we introduce a method of functionally classifying lung cancer cells from normal cells by using Tetrakis Carboxy Phenyl Porphine (TCPP) and well-known computational intelligent techniques. Tetrakis Carboxy Phenyl Porphine (TCPP) is a porphyrin that is able to label cancer cells due to the increased numbers of low density lipoproteins coating the surface of cancer cells and the porous nature of the cancer cell membrane. Lung cancer is the leading cancer killer in the world. Novel early detection technologies are needed to maximize the chance for a potentially curable stage of lung cancer. When identified early (radiographic stage 1), non small cell lung carcinoma is routinely resected with survival rates of 40 to 85%. Unfortunately, most lung cancers present at an advanced stage resulting in a dismal overall 5 year survival of 15%. We study the performance of kernel methods in the context of classification accuracy on Biomoda cultured lung sputum dataset. We use a Library for Support Vector Machines (LIBSVM) for model selection. Through a variety of comparative experiments, it is found that SVMs perform the best for detecting lung cancer. Results show that all 79 features we use give the best accuracy to identify lung cancer cells. Our results, thus, demonstrate the potential of using learning machines in detecting and classifying lung cancer cells from normal cells.

Image visualization based malware detection

Malware detection is one of the challenging tasks in Cyber security. The advent of code obfuscation, metamorphic malware, packers and zero day attacks has made malware detection a challenging task. In this paper we present a visualization based approach for malware detection. First the executable is converted to a gray-scale image called byteplot. Later we extract low level features like intensity based and texture based features. We apply computationally intelligent techniques for malware detection using these features. In this work we used Support Vector Machines (SVMs) and obtained an accuracy of 95% on a dataset containing 25000 malware and 12000 benign samples.

Novel blind video forgery detection using markov models on motion residue

In this paper we present a novel blind video forgery detection method by applying Markov models to motion in videos. Motion is an important aspect of video forgery detection as it effects forgery detection in videos. Most of the current video forgery detection algorithms do not consider motion in their approach. Motion is usually captured from motion vectors and prediction error frame. However capturing motion for I-frame is computationally expensive, so in this paper we extract the motion information by applying collusion on successive frames. First a base frame is obtained by applying collusion on successive frames and the difference between actual and estimate gives information about motion. Then we apply Markov models on this motion residue and apply pattern recognition on this. We used Support Vector Machines (SVMs) in our experiment. We obtained an accuracy of 87% even for reduced feature set.

Video steganalysis using motion estimation

In this paper we present a novel video steganalysis method using neural networks and support vector machines to detect video steganograms with very limited a-prior knowledge about the steganogram embedding method.

Lung cancer detection using labeled sputum sample: multi spectrum approach

In this paper we demonstrate the use of multi-spectrum imaging and machine learning techniques for automated detection of lung cancer. The sputum samples from patients are first stained using Tetrakis Carboxy Phenyl Porphine (TCPP). Tetrakis Carboxy Phenyl Porphine (TCPP) is a porphyrin molecular marker which binds to cancer and pre cancerous cells, causing cancer cells to glow red under fluorescent microscope. After the sputum samples are stained, images are taken at multiple frequencies 650nm and 660nm. We extracted four different sets of features (shape based, intensity based, wavelet based and Gabor filter based features). Both wavelet based and Gabor based features capture the texture properties of cell. Using these features we built different machine learning models. We obtained an accuracy of 96% using initial set of 35 features (shape based, intensity based and wavelet based features). After adding Gabor based features to this initial set, we obtained accuracy of about 98%. Our experiments show the potential of using TCPP stain, machine learning techniques and Multi-spectrum imaging for early detection of lung cancer.

Video steganalysis using spatial and temporal redundancies

In this paper we present a novel video steganalysis method using neural networks and support vector machines to detect video steganograms with very limited a-prior knowledge about the steganogram embedding method.

Early lung cancer detection using nucleus segementation based features

In this study we propose an early lung cancer detection methodology using nucleus based features. First the sputum samples from patients are labeled with Tetrakis Carboxy Phenyl Porphine (TCPP) and fluorescent images of these samples are taken. TCPP is a porphyrin that is able to assist in labeling lung cancer cells by increasing numbers of low density lipoproteins coating on the surface of cancer. We study the performance of well know machine learning techniques in the context of lung cancer detection on Biomoda dataset. We obtained an accuracy of 81% using 71 features related to shape, intensity and color in our previous work. By adding the nucleus segmented features we improved the accuracy to 87%. Nucleus segmentation is performed by using Seeded region growing segmentation method. Our results demonstrate the potential of nucleus segmented features for detecting lung cancer.

Block Level Video Steganalysis Scheme

In this paper, we propose block level video steganalysis method. Current steganalysis methods detect steganograms at frame level only. In this paper, we present a new steganalysis method using correlation of pattern noise between consecutive frames as feature. First we extract the pattern noise from each frame and obtain difference between consecutive frames pattern noise. Later we divide the difference matrix into blocks and apply Discrete Cosine Transform (DCT). We use the 63 lowest frequency components of DCT coefficients as feature vector for the block. We used ten different videos in our experiments. Our results show the potential of our method in detecting video steganograms at block level.

Feature selection for lung cancer detection using SVM based recursive feature elimination method

Cancer is the uncontrolled growth of abnormal cells, which do not carry out the functions of normal cells. Lung cancer is the leading cause of death due to cancer in the world. The survival rate of cancer is about 15%. In order to improve the survival rate, we need an early detection method. In this study, we propose a new method for early detection of lung cancer using Tetrakis Carboxy Phenyl Porphine (TCPP) and well-known machine learning techniques. Tetrakis Carboxy Phenyl Porphine (TCPP) is a porphyrin that is able to label cancer cells due to the increased numbers of low density lipoproteins coating the surface of cancer cells and the porous nature of the cancer cell membrane. In our previous work we studied the performance of well know machine learning techniques in the context of classification accuracy on Biomoda internal study. We used 79 features related to shape, intensity, and texture. We obtained an accuracy of 80% using the current feature set. In order to improve the accuracy of our method, we performed feature selection on these 79 features. We used Support Vector Machine (SVM) based Recursive feature Elimination (RFE) method in our experiments. We obtained an accuracy of 87.5% using reduced 19 feature set.