DSS 2026 / Day 1 / Amber Hall
Security improvements and operational lessons from py-libp2p and Py-IPFS-lite
Sumanjeet
Source: YouTube automatic caption track · Exported 2026-11-16. Timestamps link to the source video. Cue end times were not supplied.
Transcript
0:06Hello everyone.
0:08My name is Sumanjit and I am an open
0:10source developer from India.
0:13Currently I am the maintainer of
0:14PylibP2P which is a Python
0:16implementation of libp2p and I am also a
0:19maintainer of unified testing where we
0:22do all kind of interrupt testing between
0:25different libp2p implementations.
0:28Recently I was auditing the security
0:30layer of Python libp2p and there I have
0:33found some of the security gaps and
0:36today I want to share about those
0:37security gaps with you. What was the
0:40gaps, how we have fixed it and what
0:43lessons we have learned from them.
0:48So for all the people new to libp2p
0:51libp2p is an open source project which
0:54is used to create peer-to-peer
0:56applications. Every peer has a unique
0:59identity called peer ID
1:01and the peer ID is derived from the
1:04public key of that peer.
1:06And libp2p allows these peers to
1:09discover each other and create
1:12connections with each other and along
1:14with that there are some security
1:16protocols like TLS using which these
1:19peers can create secure connections.
1:24So while auditing the security layer I
1:26have found that these are the three
1:28issues where I have found the same
1:31pattern. We are not correctly verifying
1:34some of the certificates due to which we
1:37are getting some of the security gaps.
1:39So the first security issue was related
1:42to quick module where we are not
1:45correctly verifying the remote peer's
1:48certificate. So that's why we can have a
1:51connection from remote peer even without
1:54that peer is correctly giving a
1:56certificate.
1:57The second issue was in TLS module.
2:00There also we are getting some inbound
2:02connections and even if they are not
2:04giving us certificate, then also we are
2:07accepting this connection.
2:10Third issue was in peer records where we
2:13are not verifying the signed peer
2:15identity. Like we are not verifying who
2:18has correctly signed that certificate.
2:22The peer records issue is more
2:25important. So what is a peer record?
2:28A peer record is an identity or better
2:31to say it is an signed identity of a
2:34peer. It also contains the address on
2:37which that peer can be connected. So it
2:39is a very important thing. So if we are
2:42not correctly verifying who has signed
2:45the identity or the peer record, then we
2:48are storing this peer record in our
2:51routing table. So if we store this peer
2:55record in our routing table without
2:56verifying, what things can go wrong?
2:59Okay. We can have wrong address of other
3:02peers inside our routing table. So our
3:05node can become isolated from the entire
3:07network. We might be connecting to
3:10different peers and we will think that
3:12okay, we have connected to the correct
3:14peer. So we always have to
3:18verify before storing any kind of peer
3:21records.
3:22And if we are not doing this, then we
3:25are isolating our our peer from rest of
3:29the networks.
3:33Okay. So there are some places or there
3:35are sometimes when we have one
3:37specifications and we have to create
3:40different implementations out of it.
3:42Talking about the libp2p, we have one
3:45specifications of libp2p and we have to
3:48create different implementations of it
3:50for different use cases. Like we have
3:53rust libp2p, Go libp2p and all other
3:57kind of implementations like Python
3:59libp2p.
4:00So where
4:01we have to verify that whether they are
4:04able to talk with each other
4:07each other or not.
4:08Even if we are able to create different
4:11implementations of libp2p, it will not
4:14verify that whether all the
4:16implementations are verifying the
4:20nodes identity or not. Just imagining a
4:24situation where we have two different
4:26implementations. One is Python libp2p
4:29and one is Go libp2p.
4:31So they may be interoperable. They may
4:33be able to communicate with each other,
4:35but then also it is not confirmed that
4:38whether both of them are doing the same
4:42security checks or not. So we have to do
4:45a deep analysis of each of the
4:47implementations that whether they are
4:50doing the right security checks or not.
4:55So
4:56these are the four PRs which we have
4:58merged for fixing all the security
5:01issues security issues. The first one
5:04was quick, then TLS, then peer store,
5:07and there was one more security issue
5:09which was related to circuit relay.
5:13So we have got some more issues like the
5:17four GitHub advisory we have got. We are
5:20working on it and we will fix it soon.
5:22For anyone working on
5:24decentralized systems or anyone creating
5:27some more applications, I just want to
5:30say that before negotiating on any of
5:32the application level protocols, you
5:35must verify whether the remote peers
5:38identity is correct or not.
5:40Okay. Thank you.
Source and accuracy details
- Names, organizations, technical terms, code and numbers may be misrecognized.
- Source wording and disfluencies are preserved; no editorial corrections were applied.
- Export contains whole-second cue starts only; cue ends and subsecond timing are unavailable.
- A full caption-track export does not establish that every spoken word is represented.
- Automatic YouTube captions; not human-verified against the audio.
Source export SHA256: 61e1e165ba08404e6bcac3f5e2468d5ad4e5edb42b02d70dcc16ad2cbf553458