Introduction
Peer-to-Peer (P2P) streaming systems distribute video content by dividing files into small "chunks" shared among users. While efficient, this architecture is vulnerable to "pollution attacks," where malicious peers distribute corrupted data. Traditional integrity checks, such as those used in BitTorrent, require the entire file to be downloaded before verification, which is unsuitable for real-time streaming. Conversely, systems like eMule require trackers to store a hash for every chunk, leading to $O(n)$ storage overhead that fails to scale for large libraries or indefinite live streams. This paper proposes "checksum consolidation" to achieve chunk-level integrity with minimal server-side overhead.
Methodology
The proposed solution differentiates between Video-on-Demand (VoD) and Live streaming, utilizing distinct cryptographic structures for each.
Tree-Based Scheme for VoD
For static content, the authors employ a Merkle Hash Tree. The content source constructs a binary tree where leaf nodes are hashes of video chunks. Each internal node represents the hash of its children, culminating in a single Root Hash published to the tracker. To verify a specific chunk, a peer retrieves a "verification path"—the hashes of sibling nodes along the path to the root—from other peers.
As shown in Algorithm 2, the tree-path function identifies the necessary sibling hashes to reconstruct the root hash, allowing for $O(\log n)$ initial verification that amortizes toward $O(1)$ as more chunks are verified and internal nodes are cached locally.
Chain-Based Scheme for Live Streaming
Live streaming presents a challenge because the full content is not known in advance. The authors propose a Hash Chain where the root hash is updated periodically. Each new root hash incorporates the previous root and the latest set of chunks, maintaining a temporal chain of trust that allows for "partial publication" as content is generated.
Verification Strategies
Peers can adopt two distinct verification strategies: Eager, where the verification path is checked before downloading the video chunk, or Prudent, where chunks are downloaded in batches and verified together to minimize communication overhead.
Key Insights and Comparative Analysis
The effectiveness of these schemes depends on the specific requirements of the application type.
As illustrated in Figure 1, Live streaming necessitates real-time verification and partial publication, whereas VoD requires off-line verification to ensure content remains verifiable even if the original source disconnects.
The following table compares the proposed schemes (B and C) against the traditional Hash Vector approach (A):
| Metric | System A (Vector) | System B (Live Chain) | System C (VoD Tree) |
|---|---|---|---|
| Tracker Storage | $O(n)$ (High) | $O(1)$ (Low) | $O(1)$ (Low) |
| Publication Cost | $O(n)$ | $O(n/u)$ | $O(1)$ |
| Verification Cost | $O(1)$ | $O(1)$ (Average) | $O(1)$ (Average) |
By offloading the distribution of verification data to the P2P network, the central tracker's storage and communication costs are reduced from linear to constant ($O(1)$) for VoD.
Conclusion
This research demonstrates that robust integrity protection in P2P streaming is achievable without centralized bottlenecks. By treating verification data as a P2P resource, the system scales efficiently. Future work should address empirical performance on mobile devices and the impact of high peer churn on the availability of verification paths.
Summary updated