Physical Layer Defense and Attack Mechanisms in IoT

Project Overview

Internet-of-Things (IoT) devices that are limited in power and processing capabilities are susceptible to physical layer (PHY) spoofing (signal exploitation) attacks owing to their inability to implement a full-blown protocol stack for security. The overwhelming adoption of multicarrier techniques such as orthogonal frequency division multiplexing (OFDM) for the PHY layer makes IoT devices further vulnerable to PHY spoofing attacks. These attacks which aim at injecting bogus/spurious data into the receiver, involve inferring transmission parameters and finding PHY characteristics of the transmitted signals so as to spoof the received signal. Non-contiguous (NC) OFDM systems have been argued to have low probability of exploitation (LPE) characteristics against classic attacks based on cyclostationary analysis, and the corresponding PHY has been deemed to be secure. However, with the advent of machine learning (ML) algorithms, adversaries can devise data-driven attacks to compromise such systems. It is in this vein that we investigate the PHY spoofing performance of adversaries equipped with supervised and unsupervised ML tools in the multicarrier communication systems.

Project Objective

We investigate resilience of the multicarrier communication systems against sophisticated data-driven attacks based on the supervised and unsupervised machine learning/deep learning tools. The supervised ML approach is based on estimation/classification utilizing deep neural networks (DNN) while the unsupervised one employs variational autoencoders (VAEs) to infer physical layer characteristics such as frequency pattern and modulation scheme. Furthermore, we aim to devise novel data-driven algorithms as channel-based spoofing detectors to enhance the security of wireless communications from the physical layer. Toward this goal, we study the impact of practical considerations including paucity of available samples for training, and try to incorporate the available physics-based models as part of the final solution.

Our Methodology

In this project, we are integrating the data-driven models in the attack or defence side of the wireless communication links. On the adversary’s side, this methodology enables us to study the security performance of the communication system against state-of-the-art malicious attacks, and subsequently, take proper security measures against them. On the other hand, enhanced defence mechanisms can be devised by employing the power of data-driven models. However, these data-driven models tend to ignore the physics-based models and as a result require a large number of data samples for training. Also, various channel models are developed throughout the years in wireless communication literature, however, they usually suffer from the inability to tune the underlying parameters. We aim to alleviate these  difficulties by incorporating the physics-based models in the data-driven models in order to fuse their individual strengths towards fulfilling the security tasks.

Project Status

We have shown that VAEs are capable of learning representations from NC-OFDM signals related to their PHY characteristics such as frequency pattern and modulation scheme, which are useful for PHY spoofing. Simulation results demonstrate that the performance of the spoofing adversaries highly depends on the subcarriers’ allocation patterns used at the transmitter. Particularly, it is shown that utilizing a random subcarrier occupancy pattern precludes the adversary from spoofing and secures NC-OFDM systems against ML-based attacks. Furthermore, we have introduced a new classification framework, called HyPhyLearn, which enables us to incorporate physics-based information in the data-driven models and show its superiority against wide-range of existing statistical and learning-based classification methods.

Contact

Prof. Waheed U. Bajwa
Waheed.bajwa (AT) rutgers (DOT) edu

Prof. Narayan Mandayam
narayan (AT) winlab (DOT) rutgers (DOT) edu

References

Nooraiepour, W. U. Bajwa and N. B. Mandayam, “Learning-Aided Physical Layer Attacks Against Multicarrier Communications in IoT,” in IEEE Transactions on Cognitive Communications and Networking, vol. 7, no. 1, pp. 239-254, March 2021, doi: 10.1109/TCCN.2020.2990657.

A. Nooraiepour, K. Hamidouche, W. U. Bajwa and N. Mandayam, “How Secure are Multicarrier Communication Systems Against Signal Exploitation Attacks?,” MILCOM 2018 – 2018 IEEE Military Communications Conference (MILCOM), 2018, pp. 201-206, doi: 10.1109/MILCOM.2018.8599849.

A. Nooraiepour, W. U. Bajwa, and N. B. Mandayam, “A hybrid model-based and learning-based approach for classification with limited number of data samples,” ArXiv, vol. abs/2021., 2021.