RRC Signaling Storm Detection in O-RAN

Dang Kien Nguyen, Rim El Malki, Filippo Rebecchi · 2025

Read RRC Signaling Storm Detection in O-RAN by Dang Kien Nguyen, Rim El Malki, Filippo Rebecchi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-15738.txt.

Introduction

Dang Nguyen, Rim El Malki, and Filippo Rebecchi build a detector for signaling storms in open 5G networks, in a 2025 paper. In an attack, a flood of control messages from one or more devices drains a base station and cuts service. Open network design gives new data and tools to fight such floods.

The team made a reference attack on an open platform and measured its impact, then modeled wider loads in theory. Their detector uses control layer traits to tell attacks from honest heavy use. It runs as an add on app in the open system. Attacks were caught within ninety milliseconds, leaving a short window to act before the station failed.

Catalog record

CREATORDang Kien Nguyen, Rim El Malki, Filippo Rebecchi
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-15738

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

RRC Signaling Storm Detection in O-RAN Dang Kien Nguyen, Rim El Malki, Filippo Rebecchi Abstract The Open Radio Access Network (O-RAN) marks a significant shift in the mobile network industry. By transforming a traditionally vertically integrated architecture into an open, data-driven one, O-RAN promises to enhance operational flexibility and drive innovation. In this paper, we harness O-RAN’s openness to address one critical threat to 5G availability: signaling storms caused by abuse of the Radio Resource Control (RRC) protocol. Such attacks occur when a flood of RRC messages from one or multiple User Equipments (UEs) deplete resources at a 5G base station (gNB), leading to service degradation. We provide a reference implementation of an RRC signaling storm attack, using the OpenAirInterface (OAI) platform to evaluate its impact on a gNB. We supplement the experimental results with a theoretical model to extend the findings for different load conditions. To mitigate RRC signaling storms, we develop a threshold-based detection technique that relies on RRC layer features to distinguish between malicious activity and legitimate high network load conditions. Leveraging O-RAN capabilities, our detection method is deployed as an external Application (xApp). Performance evaluation shows attacks can be detected within 90ms, providing a mitigation window of 60ms before gNB unavailability, with an overhead of 1.2% and 0% CPU and memory consumption, respectively. I Introduction O-RAN (Open Radio Access Network) is a 5G architecture that emphasizes openness, flexibility, and interoperability by disaggregating traditional RAN components and using standardized interfaces [ 1 ] . One of its key capabilities is allowing applications to be plugged-in for various functions, including the detection and mitigation of security threats, offering a modular approach to optimizing network performance [ 2 ] . Leveraging cloud-native principles and AI/ML (Artificial Intelligence/Machine Learning) technologies, O-RAN enhances dynamic resource allocation and real-time network adaptation, making it a vital part of the next-generation telecommunication ecosystem [ 3 , 4 ] . In 5G networks, RRC signaling storms refer to a form of Denial-of-Service (DoS) attack (i.e., against the control plane) where a malicious User Equipment (UE) overwhelms the network by sending a surge of RRC connection requests, depleting available resources and potentially causing service degradation or outages (i.e., unavailability of the base station, hence preventing new UEs from reaching the RRC connected state). These storms exploit vulnerabilities in the signaling process, particularly in the RRC layer, which handles connections between the UE and gNodeB (gNB) [ 5 ] . Existing detection solutions focus on higher-layer features and do not adequately consider low-layer characteristics of the RRC protocol. Furthermore, they overlook specific causes, such as emergency or high-priority establishment requests, and fail to differentiate between genuine high-load scenarios and malicious attacks (i.e., since both may have similar traffic patterns), making them less effective in real-world conditions where such distinctions are critical [ 6 , 7 ] . In this paper, we have two main objectives: 1. to implement and test the impact of RRC signaling storms using OpenAirInterface (OAI); 2. to propose and evaluate a detection technique for these attacks in 5G networks based on the O-RAN architecture. In particular, we leverage on O-RAN’s capability to integrate applications (e.g., rApps, xApps) for detection. Our proposed solution requires to differentiate accurately between attack scenarios and legitimate high-load conditions using RRC layer features. We test the detection method in the most challenging cases, e.g., when UEs use a random identity to register and with emergency or high-priority establishment causes. In such cases, it is challenging to determine whether requests originate from a single Malicious UE (MUE) or multiple Benign UEs (BUEs). Additionally, the prioritization of these establishment causes allows a MUE to reserve RRC resources more effectively, complicating detection. To validate the experimental findings of the RRC signaling storm, we also propose a theoretical model that aligns with the observed attack behavior. The detection solution is evaluated for security and performance, proving its ability to distinguish between an attack and a high-load scenario before the gNB becomes unavailable (i.e., with an acceptable delay of \approx 90 ms), with minimal resource overhead (i.e., \approx 1.2\% CPU and \approx 0\% memory consumption). II Literature Review The work in [ 6 ] introduces an O-RAN xApp for detecting abnormal activities in Industrial Internet of Things (IIoT) devices during registration by using O-RAN interfaces to gather long-term network statistics. However, this approach relies on statistical threshold-based detection, which limits its ability to differentiate between attacks and high-load scenarios. In [ 7 ] , the authors combine time series prediction, adaptive thresholds, and anomaly detection to forecast signaling storms. While effective, this method lacks the inclusion of lower-layer features such as RRC, reducing its precision in detecting attacks targeting lower network layers. Similarly, [ 8 ] addresses 5G-RAN security challenges, focusing on two specific DoS attacks targeting RRC state transitions and fake system information requests. The proposed countermeasures include introducing randomness in system parameters to mitigate these threats. The randomization technique in [ 9 ] effectively reduces signaling volume in 5G RRC, demonstrating its efficacy in mitigating DoS signaling attacks. In [ 10 ] , the authors apply Dempster-Shafer theory to classify network operations as either attacks or normal activity. However, this method struggles with distinguishing high-load scenarios and is not applicable in cases involving random UE IDs. Meanwhile, the approach in [ 11 ] employs a supervised RNN-based (Recurrent Neural Network) model to detect excessive RRC signaling by monitoring packets at the network edge, though it does not analyze lower radio layers. The research in [ 12 ] uses thresholds on consecutive UE requests to identify low-volume attacks, but faces challenges with false positives and limited knowledge of UE identities. Lastly, [ 13 ] utilizes OpenRAN combined with machine learning for early attack detection in the RAN, focusing on a set of features, other than those at the RRC layer, which remain underexplored. The existing literature insufficiently analyzes RRC layer features for detecting signaling storms, often lacks testing under high-load conditions, and provides vague threat models in complex scenarios like random UE IDs and emergency access. Section VI addresses these gaps with a threshold-based detection solution. III Background This section provides the necessary background information on O-RAN architecture, the 5G connection establishment process, and the OAI platform which has been used for attack implementation and detection evaluation. III-A O-RAN Architecture The O-RAN architecture consists of [ 14 ] : • Disaggregated components: The architecture separates traditional base station functions into components such as the Central Unit (O-CU), Distributed Unit (O-DU), and Radio Unit (O-RU), which can be deployed flexibly. • RAN Intelligent Controller (RIC): It is comprised of a Near-Real-Time RIC (for decisions within milliseconds) and a Non-Real-Time RIC (for long-term tasks). • Service Management and Orchestration (SMO): It ensures efficient resource management across O-RAN components. • Open Interfaces: Open, standardized interfaces for communication between network elements (e.g., O-RU, O-DU, and O-CU), ensuring interoperability between different vendors’ equipment. III-B 5G Connection Establishment The 5G connection establishment process allows UEs to connect to the network for services such as data and voice communication. This involves authentication, authorization, and configuration of the UE. Figure 1 details its key steps, e.g., the downlink synchronization (step 1), the RACH (Random Access Channel) procedure (step 2), and the

Read the complete text