Privacy in blockchains is often known as more than just a cryptographic problem. If transaction values are hidden, sender and receiver identities are obfuscated, and proofs reveal nothing beyond validity, then privacy is secured here. In practice, this assumption breaks down, once transactions leave the wallet and enter the peer-to-peer network. Data in transit, even though protected by cryptographic security, can reveal information about the origin.
Even the strongest on-chain privacy mechanisms do not prevent trackers from watching your transaction progress. Timing, routing paths, and correlation form patterns that leak information about your transaction. This is the attack vector that Dandelion++ closes, and it is exactly why its integration into the
Beldex has default transaction anonymity through ring signatures, stealth addresses, and Bulletproofs++. Dandelion++ extends this beyond cryptography to network propagation, ensuring privacy persists from the moment a transaction is initiated.
The Network Layer: The Quiet Privacy Leak
Flooding
Most blockchains use a flooding-based propagation model. When a transaction is created, the originating node sends it to multiple connected nodes across the network simultaneously. Each of those nodes forwards it onward, allowing the transaction to spread across the entire network within seconds.
From a systems perspective, this approach is efficient. From a privacy perspective, it highlights why stronger, privacy-preserving network propagation protocols are essential. More on how privacy-network transactions may be deanonymized can be found in this research here: Exploring the Monero Peer-to-Peer Network.
Why Flooding is Inefficient
A tracker running multiple nodes (or a botnet) can monitor the order in which transactions arrive, measure propagation delays, and map connections between nodes. Over time, this enables attacks that probabilistically identify the origin of a transaction. Even if the transaction itself reveals nothing, the network behaviour around it often does. As the saying goes, metadata is more valuable than the data itself.
For privacy-focused blockchains, this highlights the gap between private transactions and public network behaviour. Users may believe that their transactions are anonymous, while their IP address or network location is systematically exposed during the transaction process.
Dandelion addresses this exact problem by changing how transactions move through the network. It is a resilient propagation mechanism tested in adversarial environments which ensures that even if a large number of nodes behave maliciously, network privacy is preserved.
How Dandelion Relays Transactions
Dandelion replaces this data exposure with a two-phase propagation model. Instead of exposing transactions, the network separates propagation into a stem phase and a fluff phase.
The Stem Phase
In the stem phase, the transaction is relayed quietly through a randomized path of nodes. It moves forward one hop at a time, resembling a thin stem rather than a branching tree. During this phase, the transaction’s origin is obscured.
The Fluff Phase
Only after the transaction has passed far, then the fluff phase begins. At that point, the transaction is passed across the network, the same as a regular transaction model, but with randomized timing to prevent correlation. This is called diffusion.
This staged approach to anonymity and network sharing is what makes Dandelion effective. By the time a transaction reaches the broader network, it is no longer directly linked with its origin.
Dandelion++: Dynamic Connectivity, Epochs, and A Fail Safe
Dynamic Connectivity and Epochs
The original Dandelion model introduced the stem-and-fluff idea but relied on a static path. If someone understood the network well enough, or had a large enough network of subnets or botnets, they would be able to guess transaction origins with increasing accuracy.
Dandelion++ removes this drawback through dynamic connectivity.
Nodes periodically reshuffle their node connections in short intervals called epochs. At the beginning of each epoch, a node randomly selects two new peers (relay paths). Whenever it receives a transaction, it relays it through one of these paths.
During each epoch, nodes roll a die to determine whether they're going to act as a relay passing on the transaction or initiate the full phase or diffusion phase. If a node acts as a relay, it forwards the transaction along one of the predetermined stem paths, while others become broadcasters, triggering the fluff phase. Because these roles change continuously during each epoch, the network never settles into a predictable structure.
From an attacker’s perspective, the graph keeps moving. There is no stable point from which to observe consistent patterns.
The Fail Safe
Dandelion++ also introduces a fail safe mechanism which the original Dandelion did not take into account. The original Dandelion propagation protocol assumed that every node on the network
- behaves honestly
- will run the Dandelion network propagation protocol
- will relay only one transaction at a time
- will always relay the transaction that it receives
But in real world scenarios, this is often not the case. A malicious node or a botnet within the network can,
- behave maliciously
- may not run the Dandelion network propagation protocol
- may relay several transactions at a time, some not verified or received during the stem phase
- may deny the service altogether without relaying or propagating the transaction
To ensure that transactions are propagated regardless of bad actors, each node that relays a transaction during the stem phase starts a timer. If the timer ends without the node receiving that transaction back via the fluff phase, it initiates its own fluff phase, propagating the transaction via delayed diffusion. This failsafe mechanism has two important functions:
- It thwarts deanonymization attempts based on timing analysis and
- It mitigates so-called black-hole attacks, where malicious nodes intentionally drop transactions during the stem phase instead of forwarding them.
Why This Matters in the Real World
Let us assume that a user is making repeated private payments from the same location. Even with encrypted transaction data, a surveillance adversary could monitor the network and notice that from a specific IP address the transactions are made frequently.
Over time, this forms a statistical fingerprint.
With Dandelion++, those transactions no longer originate publicly. Each transaction travels a different, randomized stem path before reaching the destination. To the tracker, transactions appear to emerge from unrelated points in the network, breaking the link between activity and location.
The same principle applies to merchants accepting BDX payments. Without network-layer privacy, transaction patterns could reveal business activity, operating hours, or geographic clustering. Dandelion++ ensures that accepting private payments does not quietly leak operational metadata.
Dandelion++ in the Context of Beldex
Beldex operates as a privacy-first blockchain and it is based on a Proof-of-Stake network that validates transactions while supporting a privacy ecosystem that includes
Dandelion++ integrates into the Beldex network to further increase anonymity.
When combined with
This is how privacy systems are built in adversarial environments.
Closing the Metadata Gap
As blockchain analytics evolve, attacks increasingly focus on metadata rather than transaction content. Network-layer information is easier to collect, harder to regulate, and often overlooked by users.
Dandelion++ over overcomes this concern by integration with the Beldex network. It does not rely on trust assumptions, centralized relays, or user configuration.
For Beldex, this represents more than a technical upgrade. It is a statement that privacy must be end-to-end, from transaction creation to final inclusion in a block. Cryptography obscures what you do. Dandelion++ hides where it starts. Both cryptographic and network level privacy complement each other to keep your transactions safe, private, and free from prying eyes.