Matching Items (26)
Filtering by
- All Subjects: deep learning
- All Subjects: Security
- Creators: Computer Science and Engineering Program
- Member of: Barrett, The Honors College Thesis/Creative Project Collection
Description
This paper presents work that was done to create a system capable of facial expression recognition (FER) using deep convolutional neural networks (CNNs) and test multiple configurations and methods. CNNs are able to extract powerful information about an image using multiple layers of generic feature detectors. The extracted information can be used to understand the image better through recognizing different features present within the image. Deep CNNs, however, require training sets that can be larger than a million pictures in order to fine tune their feature detectors. For the case of facial expression datasets, none of these large datasets are available. Due to this limited availability of data required to train a new CNN, the idea of using naïve domain adaptation is explored. Instead of creating and using a new CNN trained specifically to extract features related to FER, a previously trained CNN originally trained for another computer vision task is used. Work for this research involved creating a system that can run a CNN, can extract feature vectors from the CNN, and can classify these extracted features. Once this system was built, different aspects of the system were tested and tuned. These aspects include the pre-trained CNN that was used, the layer from which features were extracted, normalization used on input images, and training data for the classifier. Once properly tuned, the created system returned results more accurate than previous attempts on facial expression recognition. Based on these positive results, naïve domain adaptation is shown to successfully leverage advantages of deep CNNs for facial expression recognition.
ContributorsEusebio, Jose Miguel Ang (Author) / Panchanathan, Sethuraman (Thesis director) / McDaniel, Troy (Committee member) / Venkateswara, Hemanth (Committee member) / Computer Science and Engineering Program (Contributor) / Barrett, The Honors College (Contributor)
Created2016-05
Description
Modern computer processors contain an embedded firmware known as microcode that controls decode and execution of x86 instructions. Although proprietary and relatively obscure, this microcode can be modified using updates released by hardware manufacturers to correct processor logic flaws (errata). At the same time, a malicious microcode update could compromise a processor by implementing new malicious instructions or altering the functionality of existing instructions, including processor-accelerated virtualization or cryptographic primitives. Not only is this attack vector capable of subverting all software-enforced security policies and access controls, but it also leaves behind no postmortem forensic evidence since the write-only patch memory is cleared upon system reset. Although supervisor privileges (ring zero) are required to update processor microcode, this attack cannot be easily mitigated due to the implementation of microcode update functionality within processor silicon. In this paper, we reveal the microarchitecture and mechanism of microcode updates, present a security analysis of this attack vector, and provide some mitigation suggestions.
ContributorsChen, Daming Dominic (Author) / Ahn, Gail-Joon (Thesis director) / Lee, Joohyung (Committee member) / Barrett, The Honors College (Contributor) / Computer Science and Engineering Program (Contributor) / School of International Letters and Cultures (Contributor) / School of Mathematical and Statistical Sciences (Contributor)
Created2014-05
Description
We discuss processes involved in user-centric security design, including the synthesis of goals based on security and usability tasks. We suggest the usage of implicit security and the facilitation of secureuser actions. We propose a process for evaluating usability flaws by treating them as security threats and adapting traditional HCI methods. We discuss how to correct these flaws once they are discovered. Finally, we discuss the Usable Security Development Model for developing usable secure systems.
ContributorsJorgensen, Jan Drake (Author) / Ahn, Gail-Joon (Thesis director) / VanLehn, Kurt (Committee member) / Wilkerson, Kelly (Committee member) / Barrett, The Honors College (Contributor) / Computer Science and Engineering Program (Contributor)
Created2013-05
Description
Radio Frequency Identification (RFID) technology allows objects to be identified electronically by way of a small electronic tag. RFID is quickly becoming quite popular, and there are many security hurdles for this technology to overcome. The iCLASS line of RFID, produced by HID Global, is one such technology that is widely used for secure access control and applications where a contactless authentication element is desirable. Unfortunately, iCLASS has been shown to have security issues. Nevertheless customers continue to use it because of the great cost that would be required to completely replace it. This Honors Thesis will address attacks against iCLASS and means for countering them that do not require such an overhaul.
ContributorsMellott, Matthew John (Author) / Ahn, Gail-Joon (Thesis director) / Thorstenson, Tina (Committee member) / Barrett, The Honors College (Contributor) / Computer Science and Engineering Program (Contributor)
Created2014-05
Description
IoT Media broadcast devices, such as the Roku stick, Amazon Fire, and Chromecast have been emerging onto the market recently as a portable and inexpensive alternative to cable and disk players, allowing easy integration between home and business Wi-Fi networks and television systems capable of supporting HDMI inputs without the additional overhead of setting up a heavy or complicated player or computer. The rapid expansion of these products as a mechanism to provide for TV Everywhere services for entertainment as well as cheap office appliances brings yet another node in the rapidly expanding network of IoT that surrounds us today. However, the security implications of these devices are nearly unexplored, despite their prevalence. In this thesis, I will go over the structure and mechanisms of Chromecast, and explore some of the potential exploits and consequences of the device. The thesis contains an overview of the inner workings of Chromecast, goes over the segregation and limited control and fundamental design choices of the Android based OS. It then identifies the objectives of security, four different potential methods of exploit to compromise those objectives on a Chromecast and/or its attached network, including rogue applications, traffic sniffing, evil access points and the most effective one: deauthentication attack. Tests or relevant analysis were carried out for each of these methods, and conclusions were drawn on their effectiveness. There is then a conclusion revolving around the consequences, mitigation and the future implications of security issues on Chromecast and the larger IoT landscape.
ContributorsHuang, Kaiyi (Author) / Zhao, Ziming (Thesis director) / Ahn, Gail-Joon (Committee member) / W. P. Carey School of Business (Contributor) / Computer Science and Engineering Program (Contributor) / Barrett, The Honors College (Contributor)
Created2016-12
Description
Classical planning is a field of Artificial Intelligence concerned with allowing autonomous agents to make reasonable decisions in complex environments. This work investigates
the application of deep learning and planning techniques, with the aim of constructing generalized plans capable of solving multiple problem instances. We construct a Deep Neural Network that, given an abstract problem state, predicts both (i) the best action to be taken from that state and (ii) the generalized “role” of the object being manipulated. The neural network was tested on two classical planning domains: the blocks world domain and the logistic domain. Results indicate that neural networks are capable of making such
predictions with high accuracy, indicating a promising new framework for approaching generalized planning problems.
the application of deep learning and planning techniques, with the aim of constructing generalized plans capable of solving multiple problem instances. We construct a Deep Neural Network that, given an abstract problem state, predicts both (i) the best action to be taken from that state and (ii) the generalized “role” of the object being manipulated. The neural network was tested on two classical planning domains: the blocks world domain and the logistic domain. Results indicate that neural networks are capable of making such
predictions with high accuracy, indicating a promising new framework for approaching generalized planning problems.
ContributorsNakhleh, Julia Blair (Author) / Srivastava, Siddharth (Thesis director) / Fainekos, Georgios (Committee member) / Computer Science and Engineering Program (Contributor) / School of International Letters and Cultures (Contributor) / Barrett, The Honors College (Contributor)
Created2019-05
Description
Medical records are increasingly being recorded in the form of electronic health records (EHRs), with a significant amount of patient data recorded as unstructured natural language text. Consequently, being able to extract and utilize clinical data present within these records is an important step in furthering clinical care. One important aspect within these records is the presence of prescription information. Existing techniques for extracting prescription information — which includes medication names, dosages, frequencies, reasons for taking, and mode of administration — from unstructured text have focused on the application of rule- and classifier-based methods. While state-of-the-art systems can be effective in extracting many types of information, they require significant effort to develop hand-crafted rules and conduct effective feature engineering. This paper presents the use of a bidirectional LSTM with CRF tagging model initialized with precomputed word embeddings for extracting prescription information from sentences without requiring significant feature engineering. The experimental results, run on the i2b2 2009 dataset, achieve an F1 macro measure of 0.8562, and scores above 0.9449 on four of the six categories, indicating significant potential for this model.
ContributorsRawal, Samarth Chetan (Author) / Baral, Chitta (Thesis director) / Anwar, Saadat (Committee member) / Computer Science and Engineering Program (Contributor) / Barrett, The Honors College (Contributor)
Created2018-05
Description
Third-party mixers are used to heighten the anonymity of Bitcoin users. The mixing techniques implemented by these tools are often untraceable on the blockchain, making them appealing to money launderers. This research aims to analyze mixers currently available on the deep web. In addition, an in-depth case study is done on an open-source bitcoin mixer known as Penguin Mixer. A local version of Penguin Mixer was used to visualize mixer behavior under specific scenarios. This study could lead to the identification of vulnerabilities in mixing tools and detection of these tools on the blockchain.
ContributorsPakki, Jaswant (Author) / Doupe, Adam (Thesis director) / Shoshitaishvili, Yan (Committee member) / Computer Science and Engineering Program (Contributor, Contributor) / Barrett, The Honors College (Contributor)
Created2018-12
Description
Despite the more tightly controlled permissions and Java framework used by most programs in the Android operating system, an attacker can use the same classic vulnerabilities that exist for traditional Linux binaries on the programs in the Android operating system. Some classic vulnerabilities include stack overows, string formats, and heap meta-information corruption. Through the exploitation of these vulnerabilities an attacker can hijack the execution ow of an application. After hijacking the execution ow, an attacker can then violate the con_dentiality, integrity, or availability of the operating system. Over the years, the operating systems and compliers have implemented a number of protections to prevent the exploitation of vulnerable programs. The most widely implemented protections include Non-eXecutable stack (NX Stack), Address Space Layout Randomization (ASLR), and Stack Canaries (Canaries). NX Stack protections prevent the injection and execution of arbitrary code through the use of a permissions framework within a program. Whereas, ASLR and Canaries rely on obfuscation techniques to protect control ow, which requires su_cient entropy between each execution. Early in the implementation of these protections in Linux, researchers discovered that without su_cient entropy between executions, ASLR and Canaries were easily bypassed. For example, the obfuscation techniques were useless in programs that ran continuously because the programs did not change the canaries or re-randomize the address space. Similarly, aws in the implementation of ASLR and Canaries in Android only re-randomizes the values after rebooting, which means the address space locations and canary values remain constant across the executions of an Android program. As a result, an attacker can hijack the control ow Android binaries that contain control ow vulnerabilities. The purpose of this paper is to expose these aws and the methodology used to verify their existence in Android versions 4.1 (Jelly Bean) through 8.0 (Oreo).
ContributorsGibbs, Wil (Author) / Doupe, Adam (Thesis director) / Shoshitaishvili, Yan (Committee member) / Barrett, The Honors College (Contributor) / Computer Science and Engineering Program (Contributor)
Created2018-12
Description
Node.js is an extremely popular development framework for web applications. The appeal of its event-driven, asynchronous flow and the convenience of JavaScript as its programming language have driven its rapid growth, and it is currently deployed by leading companies in retail, finance, and other important sectors. However, the tools currently available for Node.js developers to secure their applications against malicious attackers are notably scarce. While there has been a substantial amount of security tools created for web applications in many other languages such as PHP and Java, very little exists for Node.js applications. This could compromise private information belonging to companies such as PayPal and WalMart. We propose a tool to statically analyze Node.js web applications for five popular vulnerabilites: cross-site scripting, SQL injection, server-side request forgery, command injection, and code injection. We base our tool off of JSAI, a platform created to parse client-side JavaScript for security risks. JSAI is novel because of its configuration capabilities, which allow a user to choose between various analysis options at runtime in order to select the most thorough analysis with the least amount of processing time. We contribute to the development of our tool by rigorously analyzing and documenting vulnerable functions and objects in Node.js that are relevant to the vulnerabilities we have selected. We intend to use this documentation to build a robust Node.js static analysis tool and we hope that other developers will also incorporate this analysis into their Node.js security projects.
ContributorsWasserman, Jonathan Kanter (Author) / Doupe, Adam (Thesis director) / Ahn, Gail-Joon (Committee member) / Zhao, Ziming (Committee member) / School of Historical, Philosophical and Religious Studies (Contributor) / Computer Science and Engineering Program (Contributor) / Barrett, The Honors College (Contributor)
Created2017-05