## Director cum Chief Forensic Scientist Directorate of Forensic Science Ministry of Home Affairs

### BibTeX

@MISC{Baragada_directorcum,

author = {Sambasiva Rao Baragada and S. Ramakrishna and M. S. Rao and S. Purushothaman and Sambasiva Rao Baragada and S. Ramakrishna and S. Purushothaman},

title = {Director cum Chief Forensic Scientist Directorate of Forensic Science Ministry of Home Affairs},

year = {}

}

### OpenURL

### Abstract

Steganographic tools and techniques are becoming more potential and widespread. Illegal use of steganography poses serious challenges to the law enforcement agencies. Limited work has been carried out on supervised steganalysis using neural network as a classifier. We present a combined method of identifying the presence of covert information in a carrier image using fisher’s linear discriminant (FLD) function followed by the radial basis function (RBF). Experiments show promising results when compared to the existing supervised steganalysis methods, but arranging the retrieved information is still a challenging problem.

### Citations

966 |
The use of multiple measurements in taxonomic problems
- Fisher
- 1936
(Show Context)
Citation Context ...erns of higher dimensions (more than two dimensions) can be seen on a two dimensional graph. The mapping operation can be linear or non-linear. R.A. Fisher developed a linear classification algorithm =-=[1]-=- and a method for constructing a classifier on the optimal discriminant plane, with minimum distance criterion for multi-class classification with small number of patterns [16]. The method of consider... |

302 | On The Limits of Steganography
- Anderson, Petitcolas
- 1998
(Show Context)
Citation Context ...is a type of hidden communication that literally means “covered writing”. The message is out in the open, often for all to see, but goes undetected because the very existence of the message is secret =-=[12, 20, 21]-=-. Steganalysis could be described as a method to prevent steganography. There are other attacks on steganography. Attacking the end hosts of the steganography algorithm by searching for security crede... |

245 |
Tukey, “A projection pursuit algorithm for exploratory data analysis
- Friedman, W
- 1974
(Show Context)
Citation Context ... error mapping [11] and projection International Journal of Computer Science and Security, Volume (2) : Issue (1) 13Sambasiva Rao Baragada, S. Ramakrishna, M.S. Rao, S. Purushothaman pursuit mapping =-=[6]-=-. In this work, the generalized declustering optimal discriminant plane is used. The mapping of the original pattern ‘X’ onto a new vector ‘Y’ on a plane is done by a matrix transformation, which is g... |

181 | Exploring Steganography: Seeing the Unseen
- Johnson, Jajodia
- 1998
(Show Context)
Citation Context ...is a type of hidden communication that literally means “covered writing”. The message is out in the open, often for all to see, but goes undetected because the very existence of the message is secret =-=[12, 20, 21]-=-. Steganalysis could be described as a method to prevent steganography. There are other attacks on steganography. Attacking the end hosts of the steganography algorithm by searching for security crede... |

85 | Detecting hidden messages using higher-order statistical models
- Farid
- 2002
(Show Context)
Citation Context ... steganography blindly, detect hidden data without knowing embedding methods, will be more useful in practical applications. A framework for steganalysis based on supervised learning has been done in =-=[34]-=-. The framework was further developed and tested. Limited work has been carried out on supervised steganalysis, using neural networks as a classifier [29, 30]. Fishers’ linear discriminant function (F... |

74 |
The Singular Value Decomposition: Its Computation and Some Applications
- Klema, Laub
- 1980
(Show Context)
Citation Context ...even for a more general discriminating analysis and multi-orthonormal vectors [7, 18, 19]. If the determinant of SW is zero, then singular value decomposition (SVD) on SW has to be done. On using SVD =-=[10, 13]-=-, SW is decomposed into three matrices U, W and V. The matrices U and W are unitary matrices, and V is a diagonal matrix with non-negative diagonal elements arranged in the decreasing order. A small v... |

64 | Detecting steganographic content on the internet
- Provos, Honeyman
(Show Context)
Citation Context ...rating covermessages from stego-messages. In this work, passive steganalysis is focused. Most of the present literature on steganalysis follows either a parametric model [28, 24, 26] or a blind model =-=[32, 27, 22, 23, 35, 33]-=-. A general steganalysis method that can attack steganography blindly, detect hidden data without knowing embedding methods, will be more useful in practical applications. A framework for steganalysis... |

55 |
An optimal set of discriminant vectors
- Foley, Sammon
- 1975
(Show Context)
Citation Context ...8) is the discriminant vector ϕ2. In Eq.(6) and Eq. (8), SW should be non-singular. The SW matrix should be non-singular, even for a more general discriminating analysis and multi-orthonormal vectors =-=[7, 18, 19]-=-. If the determinant of SW is zero, then singular value decomposition (SVD) on SW has to be done. On using SVD [10, 13], SW is decomposed into three matrices U, W and V. The matrices U and W are unita... |

52 | Steganalysis of additive noise modelable information hiding
- Harmsen, Pearlman
- 2003
(Show Context)
Citation Context ...rating covermessages from stego-messages. In this work, passive steganalysis is focused. Most of the present literature on steganalysis follows either a parametric model [28, 24, 26] or a blind model =-=[32, 27, 22, 23, 35, 33]-=-. A general steganalysis method that can attack steganography blindly, detect hidden data without knowing embedding methods, will be more useful in practical applications. A framework for steganalysis... |

39 | Steganalysis using color wavelet statistics and one-class support vector machines
- Lyu, Farid
(Show Context)
Citation Context ...y”. Steganalysis is about separating covermessages from stego-messages. In this work, passive steganalysis is focused. Most of the present literature on steganalysis follows either a parametric model =-=[28, 24, 26]-=- or a blind model [32, 27, 22, 23, 35, 33]. A general steganalysis method that can attack steganography blindly, detect hidden data without knowing embedding methods, will be more useful in practical ... |

33 |
An overview of mapping techniques for exploratory pattern analysis
- Siedlecki, Siedlecka, et al.
- 1988
(Show Context)
Citation Context ...trix transformation, which is given by where A = Y=AX (1) ⎡ϕ1 ⎤ ⎢ ⎥ ⎣ϕ 2 ⎦ (2) and ϕ1 and ϕ2 are the discriminant vectors (also called projection vectors). An overview of different mapping techniques =-=[14, 15]-=- is addressed earlier. The vectors ϕ1 and ϕ2 are obtained by optimizing a given criterion. The plane formed by the discriminant vectors is the optimal vectors which are the optimal discriminant planes... |

30 |
F Fogelman-Soulie F., “On The Relations Between Discriminant Analysis and Multilayer Perceptrons
- Gallinari, Badran
- 1991
(Show Context)
Citation Context ...ss classification with small number of patterns [16]. The method of considering the number of patterns and feature size [4], and the relations between discriminant analysis and multilayer perceptrons =-=[17]-=- has been addressed earlier. A linear mapping is used to map an n-dimensional vector space ℜ n onto a two dimensional space. Some of the linear mapping algorithms are principal component mapping [5], ... |

29 |
Optimal Discriminant Plane for a Small Number of
- Hong, Yang
- 1991
(Show Context)
Citation Context ...classification algorithm [1] and a method for constructing a classifier on the optimal discriminant plane, with minimum distance criterion for multi-class classification with small number of patterns =-=[16]-=-. The method of considering the number of patterns and feature size [4], and the relations between discriminant analysis and multilayer perceptrons [17] has been addressed earlier. A linear mapping is... |

27 |
Steganalysis of lsb encoding in color images
- Fridrich, Du, et al.
- 2000
(Show Context)
Citation Context ...y”. Steganalysis is about separating covermessages from stego-messages. In this work, passive steganalysis is focused. Most of the present literature on steganalysis follows either a parametric model =-=[28, 24, 26]-=- or a blind model [32, 27, 22, 23, 35, 33]. A general steganalysis method that can attack steganography blindly, detect hidden data without knowing embedding methods, will be more useful in practical ... |

24 |
Considerations of sample and feature size
- Foley
- 1972
(Show Context)
Citation Context ...on the optimal discriminant plane, with minimum distance criterion for multi-class classification with small number of patterns [16]. The method of considering the number of patterns and feature size =-=[4]-=-, and the relations between discriminant analysis and multilayer perceptrons [17] has been addressed earlier. A linear mapping is used to map an n-dimensional vector space ℜ n onto a two dimensional s... |

17 |
J.-Y.: Optimal Fisher discriminant analysis using the rank decomposition
- Cheng, Zhuang, et al.
- 1992
(Show Context)
Citation Context ...8) is the discriminant vector ϕ2. In Eq.(6) and Eq. (8), SW should be non-singular. The SW matrix should be non-singular, even for a more general discriminating analysis and multi-orthonormal vectors =-=[7, 18, 19]-=-. If the determinant of SW is zero, then singular value decomposition (SVD) on SW has to be done. On using SVD [10, 13], SW is decomposed into three matrices U, W and V. The matrices U and W are unita... |

17 | A mathematical framework for active steganalysis, Special issue on multimedia watermarking
- Chandramouli
(Show Context)
Citation Context ...y”. Steganalysis is about separating covermessages from stego-messages. In this work, passive steganalysis is focused. Most of the present literature on steganalysis follows either a parametric model =-=[28, 24, 26]-=- or a blind model [32, 27, 22, 23, 35, 33]. A general steganalysis method that can attack steganography blindly, detect hidden data without knowing embedding methods, will be more useful in practical ... |

14 | Secret Key Estimation in Sequential Steganography
- Trivedi, Chandramouli
- 2005
(Show Context)
Citation Context ...rating covermessages from stego-messages. In this work, passive steganalysis is focused. Most of the present literature on steganalysis follows either a parametric model [28, 24, 26] or a blind model =-=[32, 27, 22, 23, 35, 33]-=-. A general steganalysis method that can attack steganography blindly, detect hidden data without knowing embedding methods, will be more useful in practical applications. A framework for steganalysis... |

12 | Steganalysis based on moments of characteristic functions using wavelet decomposition, prediction-error image, and neural network
- Shi, Xuan, et al.
- 2005
(Show Context)
Citation Context ...based on supervised learning has been done in [34]. The framework was further developed and tested. Limited work has been carried out on supervised steganalysis, using neural networks as a classifier =-=[29, 30]-=-. Fishers’ linear discriminant function (FLD) as a classifier show impressive results in [31]. The present neural network based steganalytic work is implemented by combining the radial basis function ... |

9 |
Interactive Pattern Analysis and Classification
- Sammon
- 1970
(Show Context)
Citation Context ...near mapping is used to map an n-dimensional vector space ℜ n onto a two dimensional space. Some of the linear mapping algorithms are principal component mapping [5], generalized declustering mapping =-=[2, 3, 8, 9]-=-, least squared error mapping [11] and projection International Journal of Computer Science and Security, Volume (2) : Issue (1) 13Sambasiva Rao Baragada, S. Ramakrishna, M.S. Rao, S. Purushothaman p... |

9 |
On the Use of Singular Value Decomposition and Decimation in Discrete-Time Band-Limited Signal Extrapolation
- Sullivan
- 1984
(Show Context)
Citation Context ...even for a more general discriminating analysis and multi-orthonormal vectors [7, 18, 19]. If the determinant of SW is zero, then singular value decomposition (SVD) on SW has to be done. On using SVD =-=[10, 13]-=-, SW is decomposed into three matrices U, W and V. The matrices U and W are unitary matrices, and V is a diagonal matrix with non-negative diagonal elements arranged in the decreasing order. A small v... |

8 |
An optimal discriminant plane
- Sammon
- 1970
(Show Context)
Citation Context ...near mapping is used to map an n-dimensional vector space ℜ n onto a two dimensional space. Some of the linear mapping algorithms are principal component mapping [5], generalized declustering mapping =-=[2, 3, 8, 9]-=-, least squared error mapping [11] and projection International Journal of Computer Science and Security, Volume (2) : Issue (1) 13Sambasiva Rao Baragada, S. Ramakrishna, M.S. Rao, S. Purushothaman p... |

4 |
Attacks on steganographic systems,” Third Information Hiding Workshop
- Westfeld, Pfitzmann
- 1999
(Show Context)
Citation Context |

3 |
P.C.Young P.C, “Approach to Feature Selection Based on the Karhunen-Loeve Expansion
- Kittler
- 1973
(Show Context)
Citation Context ... [17] has been addressed earlier. A linear mapping is used to map an n-dimensional vector space ℜ n onto a two dimensional space. Some of the linear mapping algorithms are principal component mapping =-=[5]-=-, generalized declustering mapping [2, 3, 8, 9], least squared error mapping [11] and projection International Journal of Computer Science and Security, Volume (2) : Issue (1) 13Sambasiva Rao Baragad... |

3 |
A Declustering Criterion for Feature Extraction in Pattern Recognition
- Fehlauer, Eisenstein
- 1978
(Show Context)
Citation Context ...near mapping is used to map an n-dimensional vector space ℜ n onto a two dimensional space. Some of the linear mapping algorithms are principal component mapping [5], generalized declustering mapping =-=[2, 3, 8, 9]-=-, least squared error mapping [11] and projection International Journal of Computer Science and Security, Volume (2) : Issue (1) 13Sambasiva Rao Baragada, S. Ramakrishna, M.S. Rao, S. Purushothaman p... |

3 |
Mapping Algorithms in ISPAHAN,” pattern Recognition
- Gelsema
- 1980
(Show Context)
Citation Context |

3 |
Dimensionality Reduction Techniques Based on a Least Squared Error Criterion
- Mix, “A
- 1982
(Show Context)
Citation Context ...al vector space ℜ n onto a two dimensional space. Some of the linear mapping algorithms are principal component mapping [5], generalized declustering mapping [2, 3, 8, 9], least squared error mapping =-=[11]-=- and projection International Journal of Computer Science and Security, Volume (2) : Issue (1) 13Sambasiva Rao Baragada, S. Ramakrishna, M.S. Rao, S. Purushothaman pursuit mapping [6]. In this work, ... |

3 |
J.Y.Yang J.Y., “A Generalized Optimal Set of Discriminant Vectors
- Liu
- 1992
(Show Context)
Citation Context ...8) is the discriminant vector ϕ2. In Eq.(6) and Eq. (8), SW should be non-singular. The SW matrix should be non-singular, even for a more general discriminating analysis and multi-orthonormal vectors =-=[7, 18, 19]-=-. If the determinant of SW is zero, then singular value decomposition (SVD) on SW has to be done. On using SVD [10, 13], SW is decomposed into three matrices U, W and V. The matrices U and W are unita... |

3 |
Er,.Shiqian Wu, Juwei Lu and Hock Lye Toh, “Face Recognition with Radial BasisFunction(RBF) Neural Networks
- Joo
- 2002
(Show Context)
Citation Context ...es (or increases) monotonically with distance from a central point. The centre, the distance scale, and the precise shape of the radial function are parameters of the model, all fixed if it is linear =-=[25]-=-. A typical radial function is the Gaussian which, in the case of a scalar input, is h(x)=exp((-(x-c) 2 )/(r 2 )) (12) Its parameters are its centre c and its radius r. A Gaussian RBF monotonically de... |

3 |
Soft Computing Approach to Steganalysis of LSB Embedding
- Benton, Chu
- 2005
(Show Context)
Citation Context ...based on supervised learning has been done in [34]. The framework was further developed and tested. Limited work has been carried out on supervised steganalysis, using neural networks as a classifier =-=[29, 30]-=-. Fishers’ linear discriminant function (FLD) as a classifier show impressive results in [31]. The present neural network based steganalytic work is implemented by combining the radial basis function ... |

3 |
Ning Dai, Jian-Jing Shen, “Feature Selection Based on Bhattacharyya Distance: A Generalized Rough Set Method
- Sun, Han
- 2006
(Show Context)
Citation Context |

2 |
Xiaolin Wu, “Steganalysis of halftone images
- Jiang, Wong, et al.
- 2005
(Show Context)
Citation Context ...d. Limited work has been carried out on supervised steganalysis, using neural networks as a classifier [29, 30]. Fishers’ linear discriminant function (FLD) as a classifier show impressive results in =-=[31]-=-. The present neural network based steganalytic work is implemented by combining the radial basis function neural network with fishers’ linear discriminant function. 2. METHODOLOGY Machine learning th... |

1 |
Effective steganalysis based on statistical moments of differential characteristic function
- Liu, Pan, et al.
(Show Context)
Citation Context |