Toward a Hardware-assisted Online Intrusion Detection System Based on Deep Learning Algorithms for Resource-Limited Embedded Systems
| Main Author: | |
|---|---|
| Language: | English |
| Published: |
University of Cincinnati / OhioLINK
2018
|
| Subjects: | |
| Online Access: | http://rave.ohiolink.edu/etdc/view?acc_num=ucin1535464571843315 |
| id |
ndltd-OhioLink-oai-etd.ohiolink.edu-ucin1535464571843315 |
|---|---|
| record_format |
oai_dc |
| spelling |
ndltd-OhioLink-oai-etd.ohiolink.edu-ucin15354645718433152021-08-03T07:08:29Z Toward a Hardware-assisted Online Intrusion Detection System Based on Deep Learning Algorithms for Resource-Limited Embedded Systems Al Rawashdeh, Khaled Computer Engineering Deep Learning Anomaly Detection Artificial Intelligence FPGA Intrusion Detection Embedded System Real-time designs of deep learning algorithms are challenged by two less frequently addressed issues. The first is data inefficiency, i.e., the model requires several epochs of trial and error to converge which makes it impractical to be applied to real-time applications. The second is the high precision computation load of the deep learning algorithms needed to achieve high accuracy during training and inference. To address the first issue, we propose a compressed training model for the contrastive divergence algorithm (CD) in the Deep Belief Network (DBN). The goal is to dynamically adjust the training vector according to the feedback from the free energy and the reconstruction error, which allows for better generalization. Furthermore, based on the previous compressed algorithm and to reduce the saturation of the Tanh and the Sigmoid activation functions, we propose a fast activation function, namely the Adaptive Linear Function (ALF). The ALF increases the convergence speed and accuracy of online training and inference using the Deep Belief Network (DBN). To address the second issue, we propose a Hybrid-Stochastic-Dynamic-Fixed-Point (HSDFP) method, which provides a training environment with high reduction in calculation, area, and power in FPGA. Cyber-Physical Systems (CPS) have become increasingly connected in recent years in what is known as the IoT (Internet of Things). As a result, the window for attacks available for hackers and adversaries has been greatly increased. The majority of the techniques currently available for detecting attacks use signature detection by checking against a database of known attacks. More work is needed to improve detection of zero-day attacks. It is not feasible to generate a profile for large systems such as large networks to detect misuse or anomalies. Exploring deep learning for security detection is a valid approach because deep learning algorithms can extract features from raw data. Deep learning has shown high detection rates in image recognition because it is able to extract new features from gray pixels. Features that are not already known can be learned by deep learning algorithms that mimic the learning mechanism of the human brain. The ability of the deep learning algorithms to learn unknown features from the incoming data will increase self-learning. We apply our model to the task of online anomaly detection using FPGA. Our framework enables the DBN structure to detect attacks online. Thus, the network can collect an efficient number of training samples and avoid overfitting. We show that our proposed method (1) converges faster than the state-of-the-art deep learning methods, (2) uses FPGA implementation to achieve accelerated inference speed of .008ms and a high power efficiency of 37 G-ops/s/W compared to CPU, GPU, and 16-bit fixed-point arithmetic (3) uses FPGA to also achieve minimal degradation in accuracy of 95\%, 95.4\%, and 97.9\% on the benchmark datasets MNIST, NSLKDD, and Kyoto, respectively. In addition, in order to reduce the cross-correlation in the stochastic computation, we propose a memory based cross-correlation reduction method for ransomware detection approach in hardware that achieves power efficiency of 39.9 G-ops/s/W. 2018-10-02 English text University of Cincinnati / OhioLINK http://rave.ohiolink.edu/etdc/view?acc_num=ucin1535464571843315 http://rave.ohiolink.edu/etdc/view?acc_num=ucin1535464571843315 unrestricted This thesis or dissertation is protected by copyright: some rights reserved. It is licensed for use under a Creative Commons license. Specific terms and permissions are available from this document's record in the OhioLINK ETD Center. |
| collection |
NDLTD |
| language |
English |
| sources |
NDLTD |
| topic |
Computer Engineering Deep Learning Anomaly Detection Artificial Intelligence FPGA Intrusion Detection Embedded System |
| spellingShingle |
Computer Engineering Deep Learning Anomaly Detection Artificial Intelligence FPGA Intrusion Detection Embedded System Al Rawashdeh, Khaled Toward a Hardware-assisted Online Intrusion Detection System Based on Deep Learning Algorithms for Resource-Limited Embedded Systems |
| author |
Al Rawashdeh, Khaled |
| author_facet |
Al Rawashdeh, Khaled |
| author_sort |
Al Rawashdeh, Khaled |
| title |
Toward a Hardware-assisted Online Intrusion Detection System Based on Deep Learning Algorithms for Resource-Limited Embedded Systems |
| title_short |
Toward a Hardware-assisted Online Intrusion Detection System Based on Deep Learning Algorithms for Resource-Limited Embedded Systems |
| title_full |
Toward a Hardware-assisted Online Intrusion Detection System Based on Deep Learning Algorithms for Resource-Limited Embedded Systems |
| title_fullStr |
Toward a Hardware-assisted Online Intrusion Detection System Based on Deep Learning Algorithms for Resource-Limited Embedded Systems |
| title_full_unstemmed |
Toward a Hardware-assisted Online Intrusion Detection System Based on Deep Learning Algorithms for Resource-Limited Embedded Systems |
| title_sort |
toward a hardware-assisted online intrusion detection system based on deep learning algorithms for resource-limited embedded systems |
| publisher |
University of Cincinnati / OhioLINK |
| publishDate |
2018 |
| url |
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1535464571843315 |
| work_keys_str_mv |
AT alrawashdehkhaled towardahardwareassistedonlineintrusiondetectionsystembasedondeeplearningalgorithmsforresourcelimitedembeddedsystems |
| _version_ |
1719454606490075136 |