This paper presents Decentagram, a decentralized framework for data dissemination using the publish/subscribe messaging model. Decentagram uses blockchain smart contracts to authenticate events that will be published using digital signatures or self-attestation certificates from code running in trusted execution environments (TEEs), both of which are verified on-chain. This approach permits any host with valid credentials to publish verified updates, increasing decentralization and availability of the system as a whole by simplifying compensation and incentivization, even for untrusted hosts running TEEs. Decentagram also supports on-chain subscribers where third-party contracts receive events immediately: within the same transaction as the published event. The same event will also be delivered to off-chain subscribing applications through an off-chain event broker. We provide an open-source implementation of Decentagram, and evaluate the gas cost of its on-chain components and the end-to-end latency of its off-chain component.

Sending transactions on leading blockchains such as Ethereum can be slow and costly. A payment channel is a well-known scaling solution that minimizes transactions sent on the chain, and allows users to transact more efficiently. One of the guarantees of payment channels is that there is no counterparty risk, so an honest party is able to withdraw the amount of money that is reflected by the most recent transaction agreed by both parties. In this paper, we show that this guarantee can be violated when the network is under congestion. Regardless of whether or not the honest party is online, the malicious party can leverage high transaction fees to gain more money than they're supposed to. We present a novel construction of payment channels that helps mitigates these types of attacks.