| A | B | C | D | E | F | G | H | I | J | K | L | M | N | O | P | Q | R | S | T | U | V | W | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1 | The 2nd Annual Nelms Workshop on Women in IoT: Student Poster Presentation Directory | ||||||||||||||||||||||
2 | Poster Description | Presentation Time | Session Chair | Breakout Room # | YouTube Link | ||||||||||||||||||
3 | Accelerated Temperature Prediction in an IoT Enabled Scaled Home Estefania Sanchez, Caitlin Petro, Salih Safa Bacanli, Furkan Cimen, Ladislau Boloni and Damla Turgut Abstract: Testbed development for smart homes is a timely and expensive process. By using a scaled model home in an isolated environment as a testbed, we are able to perform accelerated experiments on smart homes under various scenarios. Temperature and humidity data were collected from our model home, which is able to conduct various realistic simulations under diverse climatic conditions. This data was used to train machine learning algorithms to predict changes in temperature and humidity within the home. These predictive models will be used during the development of an intelligent agent that will internally regulate humidity and temperature in a more financially and environmentally conscious manner by altering the state of the home. | 4:25PM - 4:35PM | My T Thai | 1 | https://youtu.be/2_cHebetfBs | ||||||||||||||||||
4 | A Privacy-Preserving Approach for Human Daily Activities Prediction Sharare Zehtabian, Siavash Khodadadeh, Ladislau Bölöni and Damla Turgut Abstract: The daily activities performed by a disabled or elderly person can be monitored by a smart environment, and the acquired data can be used to learn a predictive model of user behavior. To speed up the learning, several researchers designed collaborative learning systems that use data from multiple users. However, disclosing the daily activities of an elderly or disabled user raises privacy concerns. In this paper, we use state-of-the-art deep neural network-based techniques to learn predictive human activity models in the local, centralized, and federated learning settings. A novel aspect of our work is that we carefully track the temporal evolution of the data available to the learner and the data shared by the user. In contrast to previous work where users shared all their data with the centralized learner, we consider users that aim to preserve their privacy. Thus, they choose between approaches in order to achieve their goals of predictive accuracy while minimizing the shared data. To help users make decisions before disclosing any data, we use machine learning to predict the degree to which a user would benefit from collaborative learning. We validate our approaches on real-world data. | 4:25PM - 4:35PM | Damla Turgut | 2 | https://youtu.be/pqMTCOdoS7w | ||||||||||||||||||
5 | Towards an Automated Building Envelope Monitoring Naima Khan Abstract: Periodic or non-periodic temperature and humidity variations trigger various damages on the inside and outside surfaces of buildings, which eventually leads to poor insulation, additional energy consumption, and expensive repairing plan. Formal thermal inspection by professionals are expensive, often inconclusive and inconvenient for continuous or frequent monitoring. Monitoring thermal condition with sensors and thermal cameras is a potential non-intrusive way to supervise the structural well-being of buildings. Thermal condition monitoring with sensors and thermal images provides data-driven knowledge of thermal properties of built environments to residents and also helps in accelerating the process of thermal inspection by the professionals. We introduce sensor and thermal image based thermal condition monitoring framework which simultaneously learns spatial and temporal feature representations from inside and outside of built environment. We installed thermo-hygrometers and thermal cameras in three different homes. We collected temperature and humidity data for at least 40 days as well as captured thermal images from inside environment for 10 minutes in consecutive 4-5 hours on different days. Temporal clustering on the latent features of temperature and humidity time series data provides the pattern of indoor thermal conditions during different outside weather conditions. We demonstrated how indoor thermal variables respond to the outdoor thermal condition for each of the cluster patterns. Besides, we are working to have an automated scalable framework for analyzing spatial and temporal temperature variation over various building elements i.e., walls, windows, doors, etc. using longitudinal thermal images. We present the spatial and temporal relations among image regions from sequential thermal images by graphical representation. Our analysis on the spatial and temporal features of regions in the collected thermal images (from both day and night of different weather conditions) identifies the thermal variation and characterizes the spatio-temporal dynamics over different places in the built environment. | 4:25PM - 4:35PM | Tempestt Neal | 3 | https://youtu.be/6VeURep2D54 | ||||||||||||||||||
6 | Pass Gate Based Threshold Voltage Defined Logic Family with Resilience Against Hardware Attacks Morgan Thomas, Beom Soo Park and Nima Maghari Abstract: Electronic counterfeiting especially by reverse engineering is a longstanding problem with nontrivial impacts in many sectors. The sale of counterfeits results in substantial economic losses to the electronics industry, reportedly as high as $169B annually. In digital Integrated Circuits (ICs), the design, function and even identification tags be detected from the layout and routings via simple optical imaging. A remedy to this problem is to leverage different threshold voltage (VT) transistors provided in standard CMOS technologies. Therefore, a single circuit with the same layout can be used for different Boolean functions. These circuits are called threshold voltage defined (TVD) logic families. The different VT transistors are indistinguishable and thus the design is not susceptible to reverse engineering. Although conventional TVD logic families obfuscates reverse engineering from revealing the design, the advance inside channel attacks have become another issue to consider. Side channel attacks measure and utilize performance parameters such as time delay and current consumption to decode the functionality of the circuit. Thus, it has become essential to reduce the variation of the above parameters regarding different types of Boolean functions and input combinations. The proposed pass-gate based (PG-TVD) logic family when compared to conventional designs prevents side channel attacks from revealing any internal circuit information by having drastically less delay and current variation across different logic functions and input combinations. While the overall speed, power consumption, and area are comparable to the conventional TVD logic families, the variation is significantly reduced from 7% to 0.7%. This makes it far more secure and thus is a prominent candidate for obfuscation against both reverse engineering and side-channel attacks. | 4:25PM - 4:35PM | Yanning Shen | 4 | https://youtu.be/mrpmHdAt8Ws | ||||||||||||||||||
7 | Proof of Reverse Engineering Barrier: SEM Image Analysis on Covert Gates Tasnuva Farheen, Ulbert Botero, Nitin Varshey, Haoting Shen, Damon L. Woodard, Mark Tehranipoor and Domenic Forte Abstract: IoT is indispensable to our day-to-day life. It demands high confidentiality and integrity to ensure its security against malicious reverse engineering attacks. IC camouflaging has been proposed as a promising countermeasure against these attacks. Camouflaged gates contain multiple functional device structures but appear as one single layout under microscope imaging, thereby hiding the real circuit functionality from adversaries. The recent covert gate camouflaging design comes with a significantly reduced overhead cost, allowing numerous camouflaged gates in circuits and thus being resilient against various invasive and semi-invasive attacks. Dummy inputs are used in the design, but SEM imaging analysis was only performed on simplified dummy contact structures in prior work. Whether the e-beam during SEM imaging will charge differently on different contacts and further reveal the different structures or not requires extended research. In this study, we fabricated real and dummy contacts in various structures and performed a systematic SEM imaging analysis to investigate the possible charging and the consequent passive-voltage contrast on contacts. In addition, machine-learning-based pattern recognition was also employed to examine the possibility of differentiating real and dummy contacts. Based on our experimental results, we found that the difference between real and dummy contacts is insignificant in SEM imaging, which effectively prevents adversarial SEM-based reverse engineering. | 4:37PM - 4:47PM | My T Thai | 1 | https://youtu.be/6LDcrPR6WiA | ||||||||||||||||||
8 | Fairness-Aware Online Learning Over Networks Oyku Deniz Kose and Yanning Shen Abstract: Graphs see widespread use due to their power to represent complex networks. Online learning over networks has been deployed in settings in which a learning task is handled with sequentially available data, and the model parameters can be updated on the fly. Various Internet of things (IoT) applications benefit from learning over networks, and in particular online learning approaches on graphs, as the interconnection between devices can be readily modelled by graphs (power grids, sensors in a smart home, etc), and the applications collect real-time data. In many graph-based IoT studies, classifying the node/IoT device is of interest. While the benchmark studies for online learning over graphs have shown success for the node classification task, they do not utilize possibly available nodal features, which might be available in certain IoT applications, including e.g., sensor data obtained from patients in preventive health monitoring, properties of deployed interconnected sensors. Motivated by this, this study proposes an adaptation to incorporate nodal features in second-order online learning algorithms. Furthermore, the study examines performances of various online learning algorithms over graphs in terms of possible bias, and proposes pre-processing procedures to enhance fairness. This can be of great importance for numerous applications in various networked systems including but not limited to IoT. Overall, the presented scheme allows the utilization of nodal features in online learning algorithms over graphs, and assesses the fairness aspect of the algorithms. The performance over real datasets are presented to showcase the effectiveness of the proposed feature utilization scheme. | 4:37PM - 4:47PM | Yanning Shen | 4 | https://youtu.be/sDuLBaKKmNc | ||||||||||||||||||
9 | Kin-Wolf: Kinship Cues in Wolf Image Generation for Face Presentation Attack Pallabi Ghosh, Sumaiya Shomaji, Damon Woodard and Domenic Forte Abstract: Automated biometric authentication is a fast-growing technology in which human behavioral and physical traits like face, fingerprint, voice, and gait are used for verification and identification of a person by minimizing human supervision. Because of its ease of integration and rise in consumer acceptance, the role of biometric technology has a growing relevance on Internet of Things (IoT). IoT is best served by a set of secure data points, and it relies on the integrity of the data transferred. It shares vital information and make important connections that establish relationships and recommendations which often contain sensitive user data. This is where the security of biometrics becomes most important and serves as the key feature in strong security for connected devices. But the widespread application of automated biometric systems in IoT has raised concerns regarding the vulnerability of the data capture and authentication subsystems. Among the various types of attacks that can be performed, a presentation attack is one of the most challenging attacks as it is performed in the analogical domain independent of the digital processes of the biometric system. In a presentation attack, a face recognition system is subverted by presenting a facial biometric artifact or by using human-based presentation attack instruments(PAI) where the attacker modifies the input data from outside the system. In this work, we have exploited kinship cues as the artifact to generate a dictionary of input images, commonly called wolf images, in our proposed novel kinship-based wolf-presentation attack method (Kin-Wolf). We have used the kinship cues from the parents' images of the target in different proportions to generate a dictionary of wolf images which has properties of the target child. The main advantage of this work is, unlike the state-of-the-art approaches, for generating the wolf images, the feedback scores of the biometric system matcher are not utilized. This is very important, as the state-of-the-art biometric systems do not disclose the feedback score. The trade-off between kinship cues and randomization is also studied for the Kin-Wolf method. A considerably high verification accuracy is achieved for our proposed Kin-Wolf presentation attacks. | 4:37PM - 4:47PM | Tempestt Neal | 3 | https://youtu.be/QgYZCheT6XA | ||||||||||||||||||
10 | Firmware Analysis Based On Firmware Re-hosting and Program Modeling Yihang Bai and Tuba Yavuz Abstract: Embedded systems and IoT security have raised increasing concerns in recent decades. Due to the complexity of chip design and diversity of microcontroller families, it is unscalable to analyze firmware without interacting with the real device and various peripherals. Execution environment emulation tool such as QEMU, is a powerful tool to emulate execution process for many microcontroller families. Also by performing dynamic analysis on firmware, especially symbolic execution, one could explore the firmware in multiple execution paths. Thus we can gather tracing info, path constraints or find potential vulnerabilities. Fortunately, there is a QEMU based binary analysis framework, S2E, which combines emulator and symbolic execution to perform firmware analysis. Furthermore, S2E built off framework Avatar leverages the ARM port of S2E and provides us capability of analyzing ARM family based firmware. However, path explosion or state explosion, is a well-known limitation of symbolic execution. In this work, we modify Avatar as S2E-PERM to analyze real-world firmware by constraining state space and applying function modeling, to reduce path explosion and achieve higher code coverage without support of real devices. S2E-PERM improves coverage by up to 4X and it has found a potential vulnerability in TI-CC2640R2 SDK. | 4:37PM - 4:47PM | Damla Turgut | 2 | https://youtu.be/qBlULejuTWs | ||||||||||||||||||
11 | Short-Term Load Forecasting for Smart Home Appliances with Sequence to Sequence Learning Mina Razghandi, Hao Zhou, Melike Erol-Kantarci and Damla Turgut Abstract: Dynamic loads such as EVs can stress the distribution system and require fine-grained control of the micro-loads by the home energy management systems. Mid-term and short-term forecasting is essential for distribution system operation and control, while long-term prediction plays a crucial role in infrastructure planning. We propose an appliance-level short-term forecast using a sequence-to-sequence learning technique with LSTM cells. We use a real-world dataset collected from four residential buildings. We then compare our proposed scheme with three other methods, namely, Vector Auto Regression Moving Average (VARMA), Dilated One Dimensional Convolutional (Conv1D), and a Long Short-Term Memory (LSTM) Neural Network model. The results show that our model can distinguish the appliance type by its trend and learns the appliances' typical usage duration. The main contribution of this research is the improved accuracy of the predictions of the appliance-level consumption, through the proposed LSTM-based Seq2Seq learning algorithm, in comparison with the existing methods | 4:50PM - 5:00PM | My T Thai | 1 | https://youtu.be/TG4ajtK6swA | ||||||||||||||||||
12 | Enhanced IoT Network Communications Using Multi-PHY 6TiSCH Chloe Bae, Israat Haque, Michael Baddeley and Lareina Yang Abstract: Recently, single-radio, multi-protocol wireless chips are introduced to the market. These chips are equipped with multiple physical layer protocols and can switch between different physical layers (PHYs) on demand. Traditionally, IoT devices operate on a single physical layer protocol, IEEE 802.15.4, which have predetermined performance and limitations. Therefore, using the multi-protocol wireless chips could enable multi-PHY IoT wireless communications, especially in Low-Power and Lossy Networks (LLNs). A device can then switch between the PHYs to use the most suitable one to satisfy the application demand, e.g., offering a high throughput or a long-range. Multi-PHYs and 6TiSCH are two promising frameworks that can build a communication foundation in Industrial IoT. However, there is no detailed performance evaluations of these protocols together in their packet delivery ratio (PDR) and radio duty cycles and more. This study will evaluate the detailed performance and potential benefits of multi- PHYs on the 6TiSCH stack in Industrial IoT wireless network communications, and we present performance evaluation of multi-PHY in 6TiSCH stack over varying number of nodes in varying topology in industrial IoT network, so users can choose the most appropriate protocols to meet their application’s demands. | 4:50PM - 5:00PM | Damla Turgut | 2 | https://youtu.be/fHV84GAjYsQ | ||||||||||||||||||
13 | IoT-Enabled Smart Mobility Devices for Aging and Rehabilitation Laura Figueroa, Myat Win, Furkan Cimen, Salih Safa Bacanli, Nafisa Mostofa, Ladislau Bӧlӧni and Damla Turgut Abstract: Typical walkers are used by elderly with physical limitations to maintain stability as they walk. In addition, a number of these elderly also suffer from vision impairments due to aging which, makes it extremely hazardous to navigate in a complex environment. While the technological enhancements for mobility aids such as smart walkers have improved through the years, these smart walkers are focused either on visual or mobility impairment. Thus, there is a need to design a walker, which can provide support for both impairments. In this research work, we specifically targeted our walker for the needs of this population. We designed and implemented an IoT-enabled smart walker that was able to detect obstacles in a user’s path and provide feedback to the user about which direction to move in order to avoid those obstacles. Our approach for object avoidance is solely based on computer vision. We use a camera-live-stream and Google's Eficientdet-D5 object recognition. The camera is placed at the front of the rollator that captures a live stream of images to be continuously analyzed. Google’s EfficientDet object detection is paired with our developed distance-estimation machine learning regression model to warn users about obstacles in their path. We collected data by creating a realistic environment with everyday objects. We trained our collected data in the following five regression models: linear, polynomial, vector space, k-nearest neighbors, and random forest. We found that random forest performed best in predicting distance with given coordinates, class numbers, and confidence score. | 4:50PM - 5:00PM | Tempestt Neal | 3 | https://youtu.be/AMq_56nUfqo | ||||||||||||||||||
14 | |||||||||||||||||||||||
15 | |||||||||||||||||||||||
16 | |||||||||||||||||||||||
17 | |||||||||||||||||||||||
18 | |||||||||||||||||||||||
19 | |||||||||||||||||||||||
20 | |||||||||||||||||||||||
21 | |||||||||||||||||||||||
22 | |||||||||||||||||||||||
23 | |||||||||||||||||||||||
24 | |||||||||||||||||||||||
25 | |||||||||||||||||||||||
26 | |||||||||||||||||||||||
27 | |||||||||||||||||||||||
28 | |||||||||||||||||||||||
29 | |||||||||||||||||||||||
30 | |||||||||||||||||||||||
31 | |||||||||||||||||||||||
32 | |||||||||||||||||||||||
33 | |||||||||||||||||||||||
34 | |||||||||||||||||||||||
35 | |||||||||||||||||||||||
36 | |||||||||||||||||||||||
37 | |||||||||||||||||||||||
38 | |||||||||||||||||||||||
39 | |||||||||||||||||||||||
40 | |||||||||||||||||||||||
41 | |||||||||||||||||||||||
42 | |||||||||||||||||||||||
43 | |||||||||||||||||||||||
44 | |||||||||||||||||||||||
45 | |||||||||||||||||||||||
46 | |||||||||||||||||||||||
47 | |||||||||||||||||||||||
48 | |||||||||||||||||||||||
49 | |||||||||||||||||||||||
50 | |||||||||||||||||||||||
51 | |||||||||||||||||||||||
52 | |||||||||||||||||||||||
53 | |||||||||||||||||||||||
54 | |||||||||||||||||||||||
55 | |||||||||||||||||||||||
56 | |||||||||||||||||||||||
57 | |||||||||||||||||||||||
58 | |||||||||||||||||||||||
59 | |||||||||||||||||||||||
60 | |||||||||||||||||||||||
61 | |||||||||||||||||||||||
62 | |||||||||||||||||||||||
63 | |||||||||||||||||||||||
64 | |||||||||||||||||||||||
65 | |||||||||||||||||||||||
66 | |||||||||||||||||||||||
67 | |||||||||||||||||||||||
68 | |||||||||||||||||||||||
69 | |||||||||||||||||||||||
70 | |||||||||||||||||||||||
71 | |||||||||||||||||||||||
72 | |||||||||||||||||||||||
73 | |||||||||||||||||||||||
74 | |||||||||||||||||||||||
75 | |||||||||||||||||||||||
76 | |||||||||||||||||||||||
77 | |||||||||||||||||||||||
78 | |||||||||||||||||||||||
79 | |||||||||||||||||||||||
80 | |||||||||||||||||||||||
81 | |||||||||||||||||||||||
82 | |||||||||||||||||||||||
83 | |||||||||||||||||||||||
84 | |||||||||||||||||||||||
85 | |||||||||||||||||||||||
86 | |||||||||||||||||||||||
87 | |||||||||||||||||||||||
88 | |||||||||||||||||||||||
89 | |||||||||||||||||||||||
90 | |||||||||||||||||||||||
91 | |||||||||||||||||||||||
92 | |||||||||||||||||||||||
93 | |||||||||||||||||||||||
94 | |||||||||||||||||||||||
95 | |||||||||||||||||||||||
96 | |||||||||||||||||||||||
97 | |||||||||||||||||||||||
98 | |||||||||||||||||||||||
99 | |||||||||||||||||||||||
100 | |||||||||||||||||||||||