Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Khalil Mustafa Ahmad Yousef has a total of 36 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Automated Vulnerability Exploitation Using Deep Reinforcement Learning

    by Anas AlMajali, Loiy Al-Abed, Khalil M. Ahmad Yousef, Bassam J. Mohd, Zaid Samamah, Anas Abu Shhadeh

    The main objective of this paper is to develop a reinforcement agent capable of effectively exploiting a specific vulnerability. Automating pentesting can reduce the cost and time of the operation. While there are existing tools like Metasploit Pro that offer automated exploitation capabilities, they often require significant execution times and resources due to their reliance on exhaustive payload testing. In this paper, we have created a deep reinforcement agent specifically configured to exploit a targeted vulnerability. Through a training phase, the agent learns and stores payloads along with their corresponding reward values in a neural network. When encountering a specific combination of a target operating system and vulnerability, the agent utilizes its neural network to determine the optimal exploitation options. The novelty of this work lies in employing Deep Reinforcement Learning in vulnerability exploitation analysis. To evaluate our proposed methodology, we conducted training and testing on the Metasploitable platform. The training phase of the reinforcement agent was conducted on two use cases: the first one has one vulnerability, and the second one has four vulnerabilities. Our approach successfully achieved the attacker’s primary objective of establishing a reverse shell with a maximum accuracy of 96.6% and 73.6% for use cases one and two, respectively.


  • Personal assistant robot using reinforcement learning: DARWIN-OP2 as a case study

    by Khalil M. Ahmad Yousef, Bassam J. Mohd, Omar Barham, Ahmad Al-Najjar, Mohammad Abu-Diab, Anas AlMajali


  • Open Access

    Quantization-Based Optimization Algorithm for Hardware Implementation of Convolution Neural Networks

    by Bassam J. Mohd, Khalil M. Ahmad Yousef, Anas AlMajali, Thaier Hayajneh

    Convolutional neural networks (CNNs) have demonstrated remarkable performance in many areas but require significant computation and storage resources. Quantization is an effective method to reduce CNN complexity and implementation. The main research objective is to develop a scalable quantization algorithm for CNN hardware design and model the performance metrics for the purpose of CNN implementation in resource-constrained devices (RCDs) and optimizing layers in deep neural networks (DNNs). The algorithm novelty is based on blending two quantization techniques to perform full model quantization with optimum accuracy, and without additional neurons. The algorithm is applied to a selected CNN model and implemented on an FPGA. Implementing CNN using broad data is not possible due to capacity issues. With the proposed quantization algorithm, we succeeded in implementing the model on the FPGA using 16-, 12-, and 8-bit quantization. Compared to the 16-bit design, the 8-bit design offers a 44% decrease in resource utilization, and achieves power and energy reductions of 41% and 42%, respectively. Models show that trading off one quantization bit yields savings of approximately 5.4K LUTs, 4% logic utilization, 46.9 mW power, and 147 μJ energy. The models were also used to estimate performance metrics for a sample DNN design.


  • Open Access

    Automatic License Plate Detection and Recognition for Jordanian Vehicles

    by Khalil Mustafa Ahmad Yousef, Bassam Jamil Mohd, Yusra Abd-Al-Haleem Al-Khalaileh, Ahlam Hani Al-Hmeadat, Bushra Ibrahim El-Ziq


  • Emotion Recognition Based on Decoupling the Spatial Context from the Temporal Dynamics of Facial Expressions

    by Rami Alazrai, Khalil M. Ahmad Yousef, Mohammad I. Daoud

    This paper presents an emotion recognition approach based on decoupling the spatial context from the temporal dynamics of facial expressions in video sequences. In particular, each emotional state is represented as a set of temporal phases, where each temporal phase exhibits different temporal dynamics such as the expressing speed and the variable length of each phase. In this work, we have developed an algorithm for automatically detecting the temporal phases of human facial expressions by employing the concept of mutual information to define a similarity measure among different video frames. Moreover, we have developed a two-layer framework for emotional state recognition. The first layer utilizes the spatial context to classify the frames in an input video into emotional-specific temporal phases using a support vector machine classifier. In the second layer, dynamic time warping is used to classify the sequence of labels associated with the video frames, which is generated in the first layer, into a specific emotional state. In order to validate the performance of the proposed approach, we have conducted extensive computer simulations and the results show an average classification accuracy of 93.53% using the extended Cohn-Kanade facial-expression database.


  • A Bispectrum-based Approach for Detecting Deception using EEG Signals

    by Rami Alazrai, Faisal Alqasem, Saqr Alaarag, Khalil M. Ahmad Yousef, Mohammad I. Daoud

    Deception is considered a ubiquitous social phenomenon that has a significant implications in clinical, moral, and law enforcement domains. In this study, we propose a novel electroencephalography (EEG)-based approach for detecting deception. The proposed approach utilizes higher-order spectra (HOS) analysis, namely the bispectrum analysis, to construct a representation of the EEG signals. The constructed bispectrum-based representation is utilized to compute a set of features that can be used to identify deception in EEG signals. Specifically, the extracted features are used to train a support vector machine (SVM) classifier to classify EEG signals into guilty or innocent classes. The performance of the proposed EEG-based deception detection approach was evaluated using EEG signals that were recorded for eleven subjects who participated in a guilty knowledge test (GKT). The results reported in the current study demonstrate the efficacy of the proposed in detecting deception using EEG signals. In particular, the average classification accuracy obtained in differentiating between the innocent and guilty classes is equal to 74%.


  • Hardware design and modeling of lightweight block ciphers for secure communications

    by Bassam Jamil Mohd, Thaier Hayajneh, Khalil M. Ahmad Yousef, Zaid Abu Khalaf, Zakirul Alam Bhuiyan


  • Open Access

    Analyzing Cyber-Physical Threats on Robotic Platforms

    by Khalil M. Ahmad Yousef, Anas AlMajali, Salah Abu Ghalyon, Waleed Dweik, Bassam J. Mohd

    Robots are increasingly involved in our daily lives. Fundamental to robots are the communication link (or stream) and the applications that connect the robots to their clients or users. Such communication link and applications are usually supported through client/server network connection. This networking system is amenable of being attacked and vulnerable to the security threats. Ensuring security and privacy for robotic platforms is thus critical, as failures and attacks could have devastating consequences. In this paper, we examine several cyber-physical security threats that are unique to the robotic platforms; specifically the communication link and the applications. Threats target integrity, availability and confidential security requirements of the robotic platforms, which use MobileEyes/arnlServer client/server applications. A robot attack tool (RAT) was developed to perform specific security attacks. An impact-oriented approach was adopted to analyze the assessment results of the attacks. Tests and experiments of attacks were conducted in simulation environment and physically on the robot. The simulation environment was based on MobileSim; a software tool for simulating, debugging and experimenting on MobileRobots/ActivMedia platforms and their environments. The robot platform PeopleBotTM was used for physical experiments. The analysis and testing results show that certain attacks were successful at breaching the robot security. Integrity attacks modified commands and manipulated the robot behavior. Availability attacks were able to cause Denial-of-Service (DoS) and the robot was not responsive to MobileEyes commands. Integrity and availability attacks caused sensitive information on the robot to be hijacked. To mitigate security threats, we provide possible mitigation techniques and suggestions to raise awareness of threats on the robotic platforms, especially when the robots are involved in critical missions or applications.


  • Security risk assessment of the PeopleBot mobile robot research platform

    by Khalil M. Ahmad Yousef, Anas AlMajali, Roa'A Hasan, Waleed Dweik, Bassam Mohd

    Nowadays, robots are widely ubiquitous and integral part in our daily lives, which can be seen almost everywhere in industry, hospitals, military, etc. To provide remote access and control, usually robots are connected to local network or to the Internet through WiFi or Ethernet. As such, it is of great importance and of a critical mission to maintain the safety and the security access of such robots. Security threats may result in completely preventing the access and control of the robot. The consequences of this may be catastrophic and may cause an immediate physical damage to the robot. This paper aims to present a security risk assessment of the well-known PeopleBot; a mobile robot platform from Adept MobileRobots Company. Initially, we thoroughly examined security threats related to remote accessing the PeopleBot robot. We conducted an impact-oriented analysis approach on the wireless communication medium; the main method considered to remotely access the PeopleBot robot. Numerous experiments using SSH and server-client applications were conducted, and they demonstrated that certain attacks result in denying remote access service to the PeopleBot robot. Consequently and dangerously the robot becomes unavailable. Finally, we suggested one possible mitigation and provided useful conclusions to raise awareness of possible security threats on the robotic systems; especially when the robots are involved in critical missions or applications.