Thread by @Sec3dev

1/ How does @sec3dev's WatchTower `SmartThreatMonitor` learn which transactions are normal and which one's aren't to provide on demand real time alerts of abnormalities?

2/ First, it loads all historical transactions of a particular smart contract and analyzes them using a machine learning algorithm to construct an initial set of invariants (pattern of behaviour)

3/ It then tracks every new transaction and updates its database of invariants and sends out an alert if any transactions or interactions is out of the ordinary

4/ The more often you train the monitor, the smarter it gets. Here is an example of the dashboard for the `SmartThreatMonitor`, where in a normal transaction the 5th input accounts are to match the content in the table. When it doesn't, an alert is sent out through sms/slack/tg

5/ You can even go a step further, by pausing your smart contract as soon as a potential threat is detected

6/ What are some common invariants?
- Token transfer amts
- Contract dependencies
- Contract Cohorts
- Reentrancies
- Frequency
- Instructions
- Flash Loans

And much more. To test #WatchTower yourself, schedule a demo!

sec3.dev/watchtower