Gouda Salama - Academia.edu (original) (raw)
Papers by Gouda Salama
Energies
In the smart grid, malicious consumers can hack their smart meters to report false power consumpt... more In the smart grid, malicious consumers can hack their smart meters to report false power consumption readings to steal electricity. Developing a machine-learning based detector for identifying these readings is a challenge due to the unavailability of malicious datasets. Most of the existing works in the literature assume attacks to compute malicious data. These detectors are trained to identify these attacks, but they cannot identify new attacks, which creates a vulnerability. Very few papers in the literature tried to address this problem by investigating anomaly detectors trained solely on benign data, but they suffer from these limitations: (1) low detection accuracy and high false alarm; (2) the need for knowledge on the malicious data to compute good detection thresholds; and (3) they cannot capture the temporal correlations of the readings and do not address the class overlapping issue caused by some deceptive attacks. To address these limitations, this paper presents a deep ...
The International Conference on Electrical Engineering (Print), May 1, 2008
International Conference on Aerospace Sciences & Aviation Technology, May 1, 2011
IOP conference series, Oct 11, 2019
International Conference on Aerospace Sciences & Aviation Technology, May 1, 2009
Astronomical images captured using optical telescopes usually suffer from severe noise effects wh... more Astronomical images captured using optical telescopes usually suffer from severe noise effects which makes the denoising step inevitable for image analysis. This paper proposes a denoising framework for astronomical images based on Convolutional Neural Network (Astro U-net). The modified Astro U-net model has been learned in four ways, the first method is using astronomical images from the Hubble Space Telescope data set with three types of noise (dark noise, read- out noise, shot noise) added, the second method is learned using the same data set with the dark noise (dn) added only, the third method is using the same data set with the read-out noise (ron) overlaid, the fourth method is using the same data set with the shot noise (sn) added. The proposed framework for denoising the astronomical images is based on a fusion of the image that was improved by the model learned in the first method with the image that was improved by the three models that were learned by the second, third and fourth methods sequentially. Experimentally, the proposed framework shows a significant improvement in both the peak signal-to-noise ratio (PSNR) and the structural similarity index (SSIM) as compared to the Astro U-net model on different exposure time ratios.
The International Conference on Electrical Engineering (Print), May 1, 2008
The International Conference on Electrical Engineering (Print), May 1, 2006
Lecture notes in networks and systems, 2023
Journal of physics, Dec 1, 2021
Skin cancer is becoming increasingly common. Fortunately, early discovery can greatly improve the... more Skin cancer is becoming increasingly common. Fortunately, early discovery can greatly improve the odds of a patient being healed. Many Artificial Intelligence based approaches to classify skin lesions have recently been proposed. but these approaches suffer from limited classification accuracy. Deep convolutional neural networks show potential for better classification of cancer lesions. This paper presents a fine-tuning on Xception pretrained model for classification of skin lesions by adding a group of layers after the basic ones of the Xception model and all model weights are set to be trained. The model is fine-tuned over HAM10,000 dataset seven classes by augmentation approach to mitigate the data imbalance effect and conducted a comparative study with the most up to date approaches. In comparison to prior models, the results indicate that the proposed model is both efficient and reliable.
International Conference on Aerospace Sciences & Aviation Technology, May 1, 2007
International journal of computer applications, Feb 15, 2013
International journal of intelligent systems and applications, Aug 8, 2020
Sentiment analysis has become an interesting field for both research and industrial domains. The ... more Sentiment analysis has become an interesting field for both research and industrial domains. The expression sentiment refers to the feelings or thought of the person across some certain issues. Furthermore, it is also considered a direct application for opinion mining. The huge amount of tweets jotted down daily makes Twitter a rich source of textual data and one of the most essential data volumes; therefore, this data has different aims, such as business, industrial or social aims according to the data requirement and needed processing. Actually, the amount of data, which is massive, grows rapidly per second and this is called big data which requires special processing techniques and high computational power in order to perform the required mining tasks. In this work, we perform a sentiment analysis with the help of Apache Spark framework, which is considered an open source distributed data processing platform which utilizes distributed memory abstraction. The goal of using Apache Spark’s Machine learning library (MLIB) is to handle an extraordinary amount of data effectively. We recommend some Preprocessing and Machine learning text feature extraction steps for getting greater results in Sentiment Analysis classification. The effectiveness of our proposed approach is proved against other approaches achieving better classification results when using Naïve Bayes, Logistic Regression and Decision trees classification algorithms. Finally, our solution estimates the performance of Apache Spark concerning its scalability.
Image matching is one of the most famous applications in computer vision and robotics. It is used... more Image matching is one of the most famous applications in computer vision and robotics. It is used for real time target detection and recognition systems. An ideal image matching technique should be robust to different image transformations such as scaling, illumination, noise and rotation. Different feature descriptors and detectors such as SIFT, SURF, BRISK, AKAZE and ORB have been introduced previously. However, each one of them has its own weak points in their matching performance. In this paper, we firstly perform a comprehensive comparison between such image matching techniques and their performance on different datasets of images. Then, we introduce a hybrid technique that combine between different feature descriptors. The experimental results showed that the proposed hybrid technique has improved the robustness of the image matching process. The conducted comparative analysis based on the execution time, number of keypoints detected and number of inliers (good matches after outliers’ rejection) has revealed the power of combined ORB and BRISK feature descriptors, outperforming the other feature descriptors combinations in enhancing the accuracy of matching and detection tasks.
Journal of Al-Azhar University Engineering Sector, Jul 1, 2018
International Journal of Information Security and Privacy, 2021
Many MANET research works are based on the popular informal definition that MANET is a wireless a... more Many MANET research works are based on the popular informal definition that MANET is a wireless ad-hoc network that has neither infrastructure nor backbone and every network node is autonomous and moves depending on its mobility. Unfortunately, this definition pays no attention to the network servers that are essential in core-based, mission-critical, and military MANETs. In core-based MANETs, external intrusion detection systems (IDS) cannot detect internal Byzantine attacks; in addition, internal Byzantine fault tolerant (BFT) systems are unqualified to detect typical external wireless attack. Therefore, there is a real need to combine both internal and external mobile ad-hoc network (MANET) ID systems. Here, CSMCSM is presented as a two-level client server model for comprehensive security in MANETs that integrates internal and external attack detectors in one device. The internal component is based on a BFT consensus algorithm while the external component employs decision tree to classify the MANET attacks.
Energies
In the smart grid, malicious consumers can hack their smart meters to report false power consumpt... more In the smart grid, malicious consumers can hack their smart meters to report false power consumption readings to steal electricity. Developing a machine-learning based detector for identifying these readings is a challenge due to the unavailability of malicious datasets. Most of the existing works in the literature assume attacks to compute malicious data. These detectors are trained to identify these attacks, but they cannot identify new attacks, which creates a vulnerability. Very few papers in the literature tried to address this problem by investigating anomaly detectors trained solely on benign data, but they suffer from these limitations: (1) low detection accuracy and high false alarm; (2) the need for knowledge on the malicious data to compute good detection thresholds; and (3) they cannot capture the temporal correlations of the readings and do not address the class overlapping issue caused by some deceptive attacks. To address these limitations, this paper presents a deep ...
The International Conference on Electrical Engineering (Print), May 1, 2008
International Conference on Aerospace Sciences & Aviation Technology, May 1, 2011
IOP conference series, Oct 11, 2019
International Conference on Aerospace Sciences & Aviation Technology, May 1, 2009
Astronomical images captured using optical telescopes usually suffer from severe noise effects wh... more Astronomical images captured using optical telescopes usually suffer from severe noise effects which makes the denoising step inevitable for image analysis. This paper proposes a denoising framework for astronomical images based on Convolutional Neural Network (Astro U-net). The modified Astro U-net model has been learned in four ways, the first method is using astronomical images from the Hubble Space Telescope data set with three types of noise (dark noise, read- out noise, shot noise) added, the second method is learned using the same data set with the dark noise (dn) added only, the third method is using the same data set with the read-out noise (ron) overlaid, the fourth method is using the same data set with the shot noise (sn) added. The proposed framework for denoising the astronomical images is based on a fusion of the image that was improved by the model learned in the first method with the image that was improved by the three models that were learned by the second, third and fourth methods sequentially. Experimentally, the proposed framework shows a significant improvement in both the peak signal-to-noise ratio (PSNR) and the structural similarity index (SSIM) as compared to the Astro U-net model on different exposure time ratios.
The International Conference on Electrical Engineering (Print), May 1, 2008
The International Conference on Electrical Engineering (Print), May 1, 2006
Lecture notes in networks and systems, 2023
Journal of physics, Dec 1, 2021
Skin cancer is becoming increasingly common. Fortunately, early discovery can greatly improve the... more Skin cancer is becoming increasingly common. Fortunately, early discovery can greatly improve the odds of a patient being healed. Many Artificial Intelligence based approaches to classify skin lesions have recently been proposed. but these approaches suffer from limited classification accuracy. Deep convolutional neural networks show potential for better classification of cancer lesions. This paper presents a fine-tuning on Xception pretrained model for classification of skin lesions by adding a group of layers after the basic ones of the Xception model and all model weights are set to be trained. The model is fine-tuned over HAM10,000 dataset seven classes by augmentation approach to mitigate the data imbalance effect and conducted a comparative study with the most up to date approaches. In comparison to prior models, the results indicate that the proposed model is both efficient and reliable.
International Conference on Aerospace Sciences & Aviation Technology, May 1, 2007
International journal of computer applications, Feb 15, 2013
International journal of intelligent systems and applications, Aug 8, 2020
Sentiment analysis has become an interesting field for both research and industrial domains. The ... more Sentiment analysis has become an interesting field for both research and industrial domains. The expression sentiment refers to the feelings or thought of the person across some certain issues. Furthermore, it is also considered a direct application for opinion mining. The huge amount of tweets jotted down daily makes Twitter a rich source of textual data and one of the most essential data volumes; therefore, this data has different aims, such as business, industrial or social aims according to the data requirement and needed processing. Actually, the amount of data, which is massive, grows rapidly per second and this is called big data which requires special processing techniques and high computational power in order to perform the required mining tasks. In this work, we perform a sentiment analysis with the help of Apache Spark framework, which is considered an open source distributed data processing platform which utilizes distributed memory abstraction. The goal of using Apache Spark’s Machine learning library (MLIB) is to handle an extraordinary amount of data effectively. We recommend some Preprocessing and Machine learning text feature extraction steps for getting greater results in Sentiment Analysis classification. The effectiveness of our proposed approach is proved against other approaches achieving better classification results when using Naïve Bayes, Logistic Regression and Decision trees classification algorithms. Finally, our solution estimates the performance of Apache Spark concerning its scalability.
Image matching is one of the most famous applications in computer vision and robotics. It is used... more Image matching is one of the most famous applications in computer vision and robotics. It is used for real time target detection and recognition systems. An ideal image matching technique should be robust to different image transformations such as scaling, illumination, noise and rotation. Different feature descriptors and detectors such as SIFT, SURF, BRISK, AKAZE and ORB have been introduced previously. However, each one of them has its own weak points in their matching performance. In this paper, we firstly perform a comprehensive comparison between such image matching techniques and their performance on different datasets of images. Then, we introduce a hybrid technique that combine between different feature descriptors. The experimental results showed that the proposed hybrid technique has improved the robustness of the image matching process. The conducted comparative analysis based on the execution time, number of keypoints detected and number of inliers (good matches after outliers’ rejection) has revealed the power of combined ORB and BRISK feature descriptors, outperforming the other feature descriptors combinations in enhancing the accuracy of matching and detection tasks.
Journal of Al-Azhar University Engineering Sector, Jul 1, 2018
International Journal of Information Security and Privacy, 2021
Many MANET research works are based on the popular informal definition that MANET is a wireless a... more Many MANET research works are based on the popular informal definition that MANET is a wireless ad-hoc network that has neither infrastructure nor backbone and every network node is autonomous and moves depending on its mobility. Unfortunately, this definition pays no attention to the network servers that are essential in core-based, mission-critical, and military MANETs. In core-based MANETs, external intrusion detection systems (IDS) cannot detect internal Byzantine attacks; in addition, internal Byzantine fault tolerant (BFT) systems are unqualified to detect typical external wireless attack. Therefore, there is a real need to combine both internal and external mobile ad-hoc network (MANET) ID systems. Here, CSMCSM is presented as a two-level client server model for comprehensive security in MANETs that integrates internal and external attack detectors in one device. The internal component is based on a BFT consensus algorithm while the external component employs decision tree to classify the MANET attacks.