We study the effects of the Internet, especially with respect to routing on public Blockchains, taking Bitcoin as our use case. To that end, we first uncover the impact that Internet routing attacks (such as BGP hijacks) and malicious Internet Service Providers (ISP) can have on Bitcoin (see paper). Next, we provide a concrete relay design that guarantees connectivity to the Bitcoin network even in the presence of a malicious ISP (see paper). Both our attacks and our relay design generalize to other public Blockchains.
Because of the extreme efficiency of Internet routing attacks and the centralization of the Bitcoin network in few networks worldwide, we show that the following two attacks are practically possible today:
Partition attack: Any ISP can partition the Bitcoin network by hijacking few IP prefixes.
Delay attack: Any ISP carrying traffic from and/or to a Bitcoin node can delay its block propagation by 20 minutes while staying completely under the radar.
The potential damage to Bitcoin is worrying. Among others, these attacks could reduce miner's revenue and render the network much more susceptible to double spending. These attacks could also prevent merchants, exchanges and other large entities that hold bitcoins from performing transactions.
To secure Bitcoin against the most effective attack, namely the partition attack we build SABRE. SABRE is a secure and scalable Bitcoin relay network which relays blocks worldwide through a set of connections that are resilient to routing attacks. SABRE runs alongside the existing peer-to-peer network and is easily deployable. As a critical system, SABRE design is highly resilient and can efficiently handle high bandwidth loads, including Denial of Service attacks. Our results demonstrate that SABRE is effective at securing Bitcoin against routing attacks, even with deployments of as few as 6 nodes.
We built SABRE by levaraging two key technical insights.
BGP Policies: We leverage fundamental properties of inter-domain routing (BGP) policies to host relay nodes: (i) in networks that are inherently protected against routing attacks; and (ii) on paths that are economically-preferred by the majority of Bitcoin clients. These properties are generic and can be used to protect other Blockchain-based systems.Although one can run a Bitcoin node anywhere on earth, the nodes that compose the network today are far from being spread uniformly around the globe.
Particularity, our results indicate that most of the Bitcoin nodes are hosted in few Internet Service providers (ISPs): 13 ISPs (0.026% of all ISPs) host 30% of the entire Bitcoin network (left graph).
Moreover, most of the traffic exchanged between Bitcoin nodes traverse few ISPs. Indeed, our results indicate that 60% of all possible Bitcoin connections cross 3 ISPs. In other words, 3 ISPs can see 60% of all Bitcoin traffic (right graph).
Together, these two characteristics make it relatively easy for a malicious ISP to intercept a lot of Bitcoin traffic.
A BGP hijack is a routing attack in which an ISP diverts Internet traffic by advertising fake announcements in the Internet routing system.
Such attacks are frequent. Actually, our results indicate that up to hundreds of thousands of hijacks happen each month. Some of those events also affect a huge number of Internet destination: up to 30,000 IP prefixes (left graph).
These attacks already affect the Bitcoin network, today. Indeed, we found that, each month, at least 100 Bitcoin nodes are the victims of BGP hijacks, while 447 distinct nodes (∼8% of the Bitcoin nodes) ended up hijacked in November 2015 (right graph).
An attacker can use routing attacks to partition the network into two (or more) disjoint components. By preventing nodes within a component to communicate with nodes outside of it, the attacker forces the creation of parallel blockchains. After the attack stops, all blocks mined within the smaller component will be discarded together with all included transactions and the miners revenue.
An attacker can use routing attacks to delay the delivery of a block to a victim node by 20 minutes while staying completely undetected. During this period the victim is unaware of the most recently mined block and the corresponding transactions. The impact of this attack varies depending on the victim. If the victim is a merchant, it is susceptible to double spending attacks. If it is a miner, the attack wastes its computational power. Finally, if the victim is a regular node, it is unable to contribute to the network by propagating the last version of the blockchain.
To avoid partitioning the SABRE network needs to protect relay-to-relay connections and relay-to-client connections. To protect relay-to-relay connections SABRE places nodes in ISPs that peer directly to each other, forming a densely connected graph of direct peering links and also in /24 prefixes. To protect relay-to-client connections SABRE places relay nodes such that most clients have for each potential attacker at least one path to SABRE that is more economically preferred than any path this attacker can advertise.
As a publicly-facing and transparent network, SABRE is an obvious target for attackers who could, among others, craft (D)DoS attacks against its publicly-known nodes to disrupt it. To protect SABRE deployments against such attacks, we leverage the fact that most of the relay operations are communication-heavy (propagating information around) as opposed to being computation-heavy. In addition to that, the content (block) that the relays need to propagate each time is predictable and small in size. These properties enable us to offload most SABRE operations to hardware, using programmable network devices. Thanks to this software- hardware co-design, SABRE relay nodes can sustain up to Tbps of load.