Mitigating the Effect of Free-Riders in BitTorrent using Trusted Agents

Alex Sherman, Angelos Stavrou, Jason Nieh, Clifford SteinColumbia 2008 · DOI

TL;DR

This paper tackles the significant performance degradation in BitTorrent caused by "free-riders" (users who download but don't upload). They propose a novel architecture using "Trusted Auditors" (TAs) to identify and isolate these free-riders, ensuring better download speeds for compliant users. Their solution proves more effective and resource-efficient than simply increasing server bandwidth.

What Problem Does This Paper Address?

Content providers often shy away from P2P systems like BitTorrent due to unpredictable performance, largely stemming from users who exploit the system by downloading content without contributing their own upload bandwidth. This "free-riding" behavior, even when implemented by a small number of users with multiple identities (Sybil attacks), can drastically slow down download times for legitimate users. While BitTorrent's "tit-for-tat" protocol aims to encourage fair sharing, it's not robust enough to prevent free-riding. This paper investigates the severity of this problem and proposes a solution to guarantee reliable performance for compliant users.

Methods

The core of the proposed solution is the introduction of "Trusted Auditors" (TAs). These are a small set of peers, managed by the content provider, that participate in the P2P network like regular clients. TAs observe the upload behavior of their neighbors (peers they are connected to) and build a reputation history for each client identity based on their bandwidth contribution. This reputation system, combined with strong client identities (e.g., password protection), allows the system to classify peers into different service pools: a compliant pool for good citizens and a free-rider pool for those who don't contribute. The system then prioritizes traffic and resource allocation for the compliant pool.

Datasets

The experiments were conducted on the PlanetLab testbed. The paper describes using a 64MB file for downloads, simulating a typical music video size. The experiments involved varying the number of seeds (1, 2, 3, 5, 10) and the percentage of free-riding clients (0%, 20%, 40%, 60%, 80%) within a 50-client system.

Key Insights and Ideas

  • Free-riding is a severe and pervasive problem: Even a small percentage of free-riders can cause disproportionately large performance degradation for compliant users.
  • BitTorrent's tit-for-tat is insufficient: Simple modifications to BitTorrent clients can easily circumvent its built-in incentives, leading to widespread free-riding.
  • Trusted Auditors offer a robust solution: By having a trusted, managed entity observe peer behavior, the system can effectively identify and isolate free-riders.
  • Separation into service pools is key: Isolating compliant users from free-riders ensures they receive preferential treatment and better download speeds.
  • TAs are more efficient than raw bandwidth: The proposed TA system provides better performance gains with significantly less resource investment (e.g., fewer seeds) compared to simply increasing server bandwidth.

Key Results

  • With 80% of free-riding identities in a 50-client system, compliant users experienced an average download time overhead of up to 430%.
  • Even a 20% free-rider presence resulted in a 25% increase in download time.
  • A single TA could reliably identify 75-80% of free-riders within 20 minutes in a system with 80% free-riders.
  • A system with 1 Seed + 1 TA showed under 6% overhead with 80% free-riders, significantly outperforming a system with 10 Seeds and no TAs, which had a much higher overhead.
  • The TA-based system with 1 Seed + 1 TA was projected to perform better than a 10-Seed system for file sizes of 200MB and larger.

Limitations and Future Work

  • Collusion concerns: While TAs are managed by the content provider to avoid peer collusion, the paper acknowledges that the system's reliability depends on the integrity of these TAs.
  • Dynamic behavior: The authors note that malicious peers might attempt to game the system by initially behaving compliantly to gain access to a good service pool and then switching to free-riding. Studying such dynamic strategies is left for future work.
  • Fine-grained classification: The current design focuses on separating compliant users from free-riders. Future work aims to further differentiate compliant users into multiple service pools based on their specific upload rates.
  • Full implementation: The paper presents a design sketch and measurement results. A full implementation of the system and its components is a direction for future research.

Summary updated

Abstract

Even though Peer-to-Peer (P2P) systems present a cost-effective and scalable solution to content distribution, most entertainment, media and software, content providers continue to rely on expensive, centralized solutions such as Content Delivery Networks. One of the main reasons is that the current P2P systems cannot guarantee reasonable performance as they depend on the willingness of users to contribute bandwidth. Moreover, even systems like BitTorrent, which employ a tit-for-tat protocol to encourage fair bandwidth exchange between users, are prone to free-riding (i.e. peers that do not upload). Our experiments on PlanetLab extend previous research (e.g. LargeViewExploit, BitTyrant) demonstrating that such selfish behavior can seriously degrade the performance of regular users in many more scenarios beyond simple free-riding: we observed an overhead of up to 430% for 80% of free-riding identities easily generated by a small set of selfish users. To mitigate the effects of selfish users, we propose a new P2P architecture that classifies peers with the help of a small number of {\em trusted nodes} that we call Trusted Auditors (TAs). TAs participate in P2P download like regular clients and detect free-riding identities by observing their neighbors' behavior. Using TAs, we can separate compliant users into a separate service pool resulting in better performance. Furthermore, we show that TAs are more effective ensuring the performance of the system than a mere increase in bandwidth capacity: for 80\% of free-riding identities a single-TA system has a 6\% download time overhead while without the TA and three times the bandwidth capacity we measure a 100\% overhead.
BibTeX
@inproceedings{sherman2008mitigating,
  author = {Alex Sherman and Angelos Stavrou and Jason Nieh and Clifford Stein},
  title = {{Mitigating the Effect of Free-Riders in BitTorrent using Trusted Agents}},
  booktitle = {Columbia},
  year = {2008},
  doi = {10.7916/d8m333m3},
}
Figure