Voice over LTE (VoLTE) has revolutionized mobile communication by moving voice traffic into the packet-switched domain, offering superior audio quality and faster call setup. However, this transition exposes the Session Initiation Protocol (SIP)—the signaling backbone of VoLTE—to the same flooding-based Denial of Service (DoS) attacks that plague the broader internet. Traditional security measures often struggle with the stochastic nature of mobile traffic and the emergence of "stealthy" low-rate attacks. This paper presents PFilter, a novel two-tier security framework designed to identify and block SIP flooding with high precision and minimal resource overhead.
The Challenge of SIP Flooding in VoLTE
In a VoLTE environment, attackers can overwhelm the IP Multimedia Subsystem (IMS) or individual user devices by bombarding them with signaling messages like INVITE or BYE. These attacks are difficult to mitigate for three primary reasons:
- Traffic Volatility: Legitimate user behavior is highly variable, making it difficult to set static detection thresholds without causing false positives.
- Stealthy Low-Rate Attacks: Sophisticated adversaries use low-rate flooding that mimics normal traffic spikes to evade traditional detection.
- Carrier-Grade Performance Requirements: Security systems must process massive volumes of traffic in real-time without introducing latency that would degrade call quality.
The PFilter Framework
To address these challenges, the researchers developed PFilter, a two-tier design based on Counting Bloom Filters (CBF). This structure uses multiple hash functions (specifically MurmurHash3) to map SIP addresses to an array of counters, enabling high-speed processing with a negligible memory footprint.
Tier 1: Suspects Filter
The first tier acts as a preliminary screen. It employs a dynamic threshold calculated using an Exponentially Weighted Moving Average (EWMA). By tracking average transmission levels and variance over time, the system adapts to natural fluctuations in call volume (e.g., peak daytime hours vs. night). Messages that exceed this adaptive threshold are flagged as suspicious and passed to the second tier.
Tier 2: Attackers Finder and the "Collapse Strategy"
The second tier is responsible for pinpointing the actual attackers within the suspect pool. It utilizes a unique "Collapse Strategy"—analogous to the game Tetris. The system periodically scans the PFilter and deletes entries corresponding to known legitimate traffic patterns. As "good" traffic is cleared away, malicious messages—which occur at significantly higher frequencies—remain "stuck" in the filter, exposing the attackers' identities. This allows carriers to drop malicious traffic specifically while maintaining service for legitimate users.
Experimental Evaluation and Results
The researchers evaluated PFilter using a custom experimental testbed simulating normal users, an IMS server, and attackers. Normal traffic was randomized between 700 and 3,200 calls per second (cps), while attacks ranged from a stealthy 10 cps to a heavy 100 cps.
Key Performance Metrics
- High-Rate Precision: At flooding rates of 75 cps and above, the detection rate and filter rate converged at nearly 100%, demonstrating near-perfect mitigation for heavy attacks.
- Low-Rate Accuracy: PFilter significantly outperformed existing methods in stealthy scenarios. At a low attack rate of 15 cps, it achieved a 97.1% detection rate, compared to just 88% for previous state-of-the-art solutions.
- DDoS Resilience: In multi-source attack scenarios involving 80 simultaneous attackers, the system maintained a false detection rate of only 3.7%.
- Efficiency: The system processed complex multi-source attacks in approximately 34.7 milliseconds, proving its readiness for real-time deployment in carrier networks.
Conclusion
PFilter provides a robust, self-adaptive solution for securing VoLTE signaling. By focusing on identifying the specific sources of an attack rather than just detecting the presence of a surge, it allows for surgical mitigation that preserves user experience. Future work may extend this two-tier design to other protocols, such as Real-time Transport Protocol (RTP) media flooding, to provide comprehensive protection for the next generation of telecommunications infrastructure.
Summary updated