In this article, real-time jamming detection against unmanned aerial vehicles (UAVs) is proposed via the integration of a software-defined radio (SDR) with an on-board Raspberry Pi processor. The SDR is utilized for capturing and forwarding the radio frequency signals to a receiver module hosted in the processor. This module extracts signal features characterized by orthogonal frequency division multiplexing (OFDM) parameters, energy parameters, and signal-to-noise ratio (SNR) parameters. Upon feature extraction, the aforementioned module exploits a machine learning (ML) classifier for detecting and classifying four jamming types; namely, barrage, single-tone, successive-pulse, and protocol-aware. The resulting configuration yielded in an overall detection rate (DR) of 93% and a false alarm rate (FAR) of 1.1%, which are in proximity to their counterparts obtained during the validation stage of the receiver module.
more »
« less
A Machine Learning Approach for Detecting and Classifying Jamming Attacks Against OFDM-based UAVs
In this paper, a machine learning (ML) approach is proposed to detect and classify jamming attacks on unmanned aerial vehicles (UAVs). Four attack types are implemented using software-defined radio (SDR); namely, barrage, single-tone, successive-pulse, and
protocol-aware jamming. Each type is launched against a drone that uses orthogonal frequency division multiplexing (OFDM) communication to qualitatively analyze its impacts considering jamming range, complexity, and severity. Then, an SDR is utilized in proximity to the drone and in systematic testing scenarios to record the radiometric parameters before and after each attack is launched. Signal-to-noise ratio (SNR), energy threshold, and several OFDM parameters are exploited as features and fed to six ML algorithms to explore and enable autonomous jamming detection/classification. The algorithms are quantitatively evaluated with metrics including detection and false alarm rates to evaluate the received signals and facilitate efficient decision-making for improved reception integrity
and reliability. The resulting ML approach detects and classifies jamming with an accuracy of 92.2% and a false-alarm rate of 1.35%.
more »
« less
- Award ID(s):
- 2006662
- NSF-PAR ID:
- 10273173
- Date Published:
- Journal Name:
- Proceedings of the 3rd ACM Workshop on Wireless Security and Machine Learning
- Page Range / eLocation ID:
- 1-6
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Small-scale unmanned aerial vehicles (UAVs) have become an increased presence in recent years due to their decreasing price and ease of use. Similarly, ways to detect drones through easily accessible programs like WireShark have raised more potential threats, including an increase in ease of jamming and spoofing drones utilizing commercially of the shelf (COTS) equipment like software defined radio (SDR). Given these advancements, an active area of research is drone security. Recent research has focused on using a HackRF SDR to perform eavesdropping or jamming attacks; however, most have failed to show a proposed remediation. Similarly, many research papers show post analysis of communications, but seem to lack a conclusive demonstration of command manipulation. Our security assessment shows clear steps in the manipulation of a WiFi drone using the aircrack-ng suite without the need for additional equipment like a SDR. This shows that anyone with access to a computer could potentially take down a drone. Alarmingly, we found that the COTS WiFi drone in our experiment still lacked the simple security measure of a password, and were very easily able to take over the drone in a deauthorization attack. We include a proposed remediation to mitigate the preformed attack and assess the entire process using the STRIDE and DREAD models. In doing so, we demonstrate a full attack process and provide a resolution to said attack.more » « less
-
Although consumer drones have been used in many attacks, besides specific methods such as jamming, very little research has been conducted on systematical methods to counter these drones. In this paper, we develop generic methods to compromise drone position control algorithms in order to make malicious drones deviate from their targets. Taking advantage of existing methods to remotely manipulate drone sensors through cyber or physical attacks (e.g., [1], [2]), we exploited the weaknesses of position estimation and autopilot controller algorithms on consumer drones in the proposed attacks. For compromising drone position control, we first designed two state estimation attacks: a maximum False Data Injection (FDI) attack and a generic FDI attack that compromised the Kalman-Filter-based position estimation (arguably the most popular method). Furthermore, based on the above attacks, we proposed two attacks on autopilot-based navigation, to compromise the actual position of a malicious drone. To the best of our knowledge, this is the first piece of work in this area. Our analysis and simulation results show that the proposed attacks can significantly affect the position estimation and the actual positions of drones. We also proposed potential countermeasures to address these attacks.more » « less
-
This work proposes the use of machine learning (ML) as a candidate for the detection of various types of message injection attacks against automatic dependent surveillance-broadcast (ADSB) messaging systems. Authentic ADS-B messages from a high-traffic area are collected from an open-source platform. These messages are combined with others imposing path modification, ghost aircraft injection, and velocity drift obtained from simulations. Then, ADS-B-related features are extracted from such messages and used to train different ML models for binary classification. For this purpose, authentic ADS-B data is considered as Class 1 (i.e., no attack), while the injection attacks are considered as Class 2 (i.e., presence of attack). The performance of the models is analyzed with metrics, including detection, misdetection, and false alarm rates, as well as validation accuracy, precision, recall, and Fl-score. The resulting models enable identifying the presence of injection attacks with a detection rate of 99.05%, and false alarm and misdetection rates of 0.76% and 1.10%, respectively.more » « less
-
GPS spoofing attacks are a severe threat to unmanned aerial vehicles. These attacks manipulate the true state of the unmanned aerial vehicles, potentially misleading the system without raising alarms. Several techniques, including machine learning, have been proposed to detect these attacks. Most of the studies applied machine learning models without identifying the best hyperparameters, using feature selection and importance techniques, and ensuring that the used dataset is unbiased and balanced. However, no current studies have discussed the impact of model parameters and dataset characteristics on the performance of machine learning models; therefore, this paper fills this gap by evaluating the impact of hyperparameters, regularization parameters, dataset size, correlated features, and imbalanced datasets on the performance of six most commonly known machine learning techniques. These models are Classification and Regression Decision Tree, Artificial Neural Network, Random Forest, Logistic Regression, Gaussian Naïve Bayes, and Support Vector Machine. Thirteen features extracted from legitimate and simulated GPS attack signals are used to perform this investigation. The evaluation was performed in terms of four metrics: accuracy, probability of misdetection, probability of false alarm, and probability of detection. The results indicate that hyperparameters, regularization parameters, correlated features, dataset size, and imbalanced datasets adversely affect a machine learning model’s performance. The results also show that the Classification and Regression Decision Tree classifier has an accuracy of 99.99%, a probability of detection of 99.98%, a probability of misdetection of 0.2%, and a probability of false alarm of 1.005%, after removing correlated features and using tuned parameters in a balanced dataset. Random Forest can achieve an accuracy of 99.94%, a probability of detection of 99.6%, a probability of misdetection of 0.4%, and a probability of false alarm of 1.01% in similar conditions.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1145/3468218.3469049
