Deep learning of multi-layer computational models allowed processing to recognize data representation at multiple levels of abstraction. These techniques have greatly improved the latest ear recognition technology. PNN is a type of radiative basis for classification problems and is based on the Bayes decision-making base, which reduces the expected error of classification. In this paper, strong features of images are used to give a good result, therefore, SIFT method using these features after adding improvements and developments. This method was one of the powerful algorithms in matching that needed to find energy pixels. This method gives stronger feature on features and gives a large number of a strong pixel, which is considered a center and neglected the remainder of it in our work.
Each pixel of which is constant for image translation, scaling, rotation, and embedded lighting changes in lighting or 3D projection. Therefore, the interpretation is developed by using a hierarchical cluster method; to assign a set of properties (find the approximation between pixels) were classified into one.

As a result of the tremendous development in the field of information technology and the internet and exposure to a flood of violations to circumvent and steal information organized and unorganized. The urgent need for the emergence of data protection technologies and data encryption techniques. As a result of the tremendous development in the field of information technology and the internet and exposure to a flood of violations to circumvent and steal information organized and unorganized. The urgent need for the emergence of data protection technologies and data encryption techniques one of these methods, The Salsa20 encryption stream and all of its reduced versions Salsa20 / 7 and Salsa20 / 12 are among the fastest stream ciphers today. In this paper, the Salsa20 method is therefore improved by adding a new variable by using chaos theory which can achieve faster propagation than the original Salsa20 and has been applied by encrypting medical images that need confidentiality because some patients do not want anyone to know about a disease so the patient’s medical data is encrypted and no one can access it. This method has been tested and measured with the original Salsa20 with a series of tests. Most tests show that the proposed messy salsa is faster than the original salsa

Face recognition is method of recognizing individuals by facial expressions. It has become essential for security and surveillance applications, including banks, organisations, workplaces, and social areas, and is needed everywhere. In face recognition, there are a variety of difficulties faced, including face shape, age, sex, lighting, and other variable factors. Another problem is that the scale of the servers for these apps is relatively limited. Education and acknowledgment, thus, are increasingly complicated. In recent years, many unchanged features have been proposed in the literature, in this paper approach the use of the fast algorithm as local descriptors, and as we shall see, it is not only fixed-size features, but also offers the advantage of being highly efficient. The proposed approach allows distinguishing the destination after converting the image to the HSV system, after which the force field features will be extracted using the fast algorithm and then classification by using the distance for three methods (Manhattan, Euclidean, and Cosine) through which a comparison is made to choose the best resolution, as it was found that the resulting accuracy of the two dataset (ORL and UFI) is 99.9%.