I2P 2.14.0 (API 0.9.71) was released on 2026-10-20. This release hardens security across the router and bundled apps, brings a large round of i2psnark improvements, and fixes streaming and tunnel reliability issues.

As usual, we recommend that you update to this release. The best way to maintain security and help the network is to run the latest release.

https://i2p.net/en/downloads/

Notes: Windows Easy Installer, Ubuntu repo, and Android will be released in the coming days. Security

Router

  • AES operations and EdDSA signature comparison are now constant time (73819404c, 6440d5f16)
  • Fixed an out-of-bounds read in garlic clove parsing and an undersized NTCP2 message 3 buffer (67b2f7327, db3cd60f1)
  • Tunnels drop duplicate fragments and cap the fragment reassembly map (8af94b9e4, 4c3d10c12)
  • Smaller robustness fixes in NetDB, NTCP, and datagram handling

i2ptunnel

  • HTTP proxy blocks requests to private, loopback, and link-local IPv4 addresses (3aef97a9d)
  • Fixed SOCKS 4/4a downgrade bypass of proxy authentication (ff63e0cdf)
  • HTTP server tunnels strip CR from headers; proxy auth strings are no longer logged
  • Output escaping fixes throughout the Hidden Services Manager

i2pcontrol

  • CSRF token protection, JSON security headers, and a delay on failed auth (9a1c56f35, e79e4949d, 8fbfb70fb)
  • Removed fallback to the default itoopie password

i2psnark and Addressbook

  • Hardened bencode parsing (recursion limit, duplicate key rejection) and path traversal checks
  • Addressbook checks for destination mismatch on update (a884b9093)

i2psnark

  • Magnet links: all trackers are now used and saved, plus dn name and xl size parameter support
  • Trackers and DHT: UDP backup tracker, smarter announce timing, spec-compliant dictionary sorting, compressed UDP and DHT messages
  • Performance: lower memory use and faster hashing when creating or rechecking torrents; tuned request throttling, pipelining, and timeouts
  • UI: free disk space shown and checked when adding torrents, audio playlist controls, live updates resume when the tab is visible again
  • Fixes: file priorities no longer lost after the startup check (GH #170), metainfo attribute extraction, extension handshake NPE
  • Groundwork for BitTorrent v2 (not yet user-facing)

Streaming and Transport

  • Streaming synchronization fixes in the retransmit and ack paths, plus slow start and reset tuning (061d4e3f6)
  • Tunnels periodically use 5-slot tunnel build messages
  • New explicit force-IPv6 option, with Docker bridge support in progress
  • SAM: fixed DG2/DG3 subsessions (d2392ae9b)
  • I2CP: new discardMessage() session API

Other

  • SusiDNS: new "conflicts" address book
  • Reseed: added reseed2.i2pgit.org, reseed.kiota.ch, two skhron.eu servers, and i2p.ntp.poweredbyberlin.de; removed reseed.i2pgit.org
  • Docker: Debian Trixie, OpenJDK 21, Ant 1.10.18, commit SHA stamped in builds
  • Build fixes for izPack 5.2.6, gettext 1.0, and Gradle
  • GeoIP, Tor blocklist, tracker list, and translation updates

Full changelog: i2p-2.13.0...i2p-2.14.0

Thanks to zzz, eyedeekay, Nicholas Lopez, and fa.

[–] [S] 1 point 3 weeks ago (2 children)

If it's anything like the DC that I've heard of before it's a file sharing app. But this runs over i2p and thus can't be traced back to you.

Or rather, what you download/share isn't known to anyone but you.

  • source
  • parent
  • context
  •  

    Copied from that other site. ‐-------------------------------------------

    Windows installer and portable ZIP with I2P built in. No separate I2P setup or port forwarding. Linux terminal client and hub/server also available.

    Windows download

    Linux download

    GitHub: Client - Server

    Start the client, give I2P a few minutes, then select File → Connect to I2P hub: Or just press the star.

    adc://mrjdpwyvlr23hoqpd3xbkgf4z7gdmwsbug7qm4fjkwd4dnsx4noq.b32.i2p No registration or password.

    Audio only test hub.

    Edit:

    The protable version work in Sandboxie and no need to share any files.

     

    Found this on a OSM forum, might be Sweden only. By user Alexey_Guskov

    Hi all,

    I print and sell 3D city maps, and while doing my own neighbourhood I realised how thin OSM’s building data is in Sweden. Almost no roof shapes, hardly any building:part, no towers or spires. The churches in Gamla Stan are single blocks with a slightly taller one next to them for the spire :)

    I tried JOSM and a few others but ended up writing my own editor, to make it easy to split a building into parts, add tower segments with min_height, set roof shapes, and double check the 3D result before uploading.

    Nothing is traced from external data. Esri World Imagery and Lantmäteriet’s Laserdata Skog (CC0) are shown as reference layers, over the map and over the 3D preview, which is how I check that a roof or a height is actually right.

    (Lantmäteriet’s building data is still on hold for OSM pending permission, so it’s not used.)

    So far I got a bit carried away and mapped every building in Hammarby Sjöstad and a few places in Gamla Stan. It is also kinda my first time contributing to OSM. I hope I didn’t mess things up too much, but I’m trying my best. Let me know if I can improve somewhere.

    The editor is here, and I’ll be happy if anyone finds it useful: https://buildings.sillygoose.se/#way/1551309876

    [–] 3 points 2 months ago (1 child)

    Looks similar, but I think SC is more like a game? This looks more like an editor. SC gives you a question about your current location, like when you walk past a shop it asks you for the opening hours, or what sort of pavement are you currently walking on.

    But whatever works for improving OSM I'm happy with.

  • source
  • parent
  • context
  • [–] 2 points 3 months ago

    Had gentoo as my daily driver for years, then it turned out that I was worse at optimizing my system than the debian guys who create distros professionally, so I just... stopped.

    Spent too much time fiddling with gentoo because it was fun, now I'm on fedora because it just works and is modern enough.

    (The amount of time spent fixing my system once I discovered "breakmygentoo" we probably shouldn't talk about)

  • source
  • parent
  • context
  •  

    This release of I2P, 2.13.0, has been focused on post-quantum support, performance improvements, and bug fixes that improve usability.

    Since the 2.12.0 release (April 2026), development has focused on stronger encryption, better bandwidth management, network stability, and a more usable console.

    Post-quantum protection is now fully available on the main transport (NTCP2), guarding against future “harvest now, decrypt later” attacks. The underlying cryptography library has been updated with constant-time operations prevent timing-based attacks. Legacy transport code (SSU1) has been fully removed.

    Tunnels now have per-tunnel bandwidth limits, preventing any single tunnel from monopolizing your connection. The system is smarter about congestion.

    Peer discovery and selection have been tuned for reliability. Floodfills are tracked more accurately.

    The web interface now shows a world map of routers in search results for RouterInfos in the NetDB.

    Containers now handle log rotation automatically, drop privileges properly, manage IP addresses correctly in containerized environments, and start up more reliably.

    The BitTorrent client (i2psnark) reannounces faster after magnet links resolve, handles edge cases better, and cleans up bandwidth limiters on exit. The email client (Susimail) offers an experimental standalone build and removes old login code.

    As usual, we recommend that you update to this release. The best way to maintain security and help the network is to run the latest release.

     

    Copied from that other site, I'm not the person behind this.


    Recently I was thinking about what I'm doing on internet. And I found out that all of my internet activity is basically just watching videos on youtube. So I thought "Why there isn't anonymous, private and free (as in freedom) youtube-like site on I2P?" and decided to do it myself.

    First eepsite of this project: http://itubep.i2p

    More you can read on project's github page: https://github.com/i2ptuber/itubep

    Warning! Work heavily in progress.

    Warning!!! Code is 100% AI slop, because I'm poor programmer.

     

    Copied from that other site...

    Independent researcher working on an I2P censorship measurement paper (English). Have some preliminary findings and an analysis framework in place. Looking to go deeper before submitting to arXiv.

    Seeking collaborators for:

    • Longitudinal tracking — baseline data collected, need ongoing observation for trend analysis

    • Node reverse engineering — behavioral patterns identified, want to confirm technical details

    • Protocol verification — TCP-level results done, want to add deeper handshake validation

    Non-profit project, no funding. Co-authorship (2nd/3rd) based on contribution.

    Interested? Reach me via PGP-encrypted email with your GitHub profile and a brief intro:

    iasds.github.io

    PGP: 1A60 9235 A475 39DF 6C4C C09B 822A 6DA9 E3BD 0612

     

    Copied from reddit

    I built an open-source Python tool that scans the I2P NetDB using nothing but vanilla i2pd floodfills. No patched binaries, no active probing — just parse the routerInfo files i2pd already writes to disk.

    Repo: https://github.com/iasds/i2p-network-scanner

    How it works

    Run a few i2pd floodfill instances with different identities. Each covers a different DHT shard, so together they passively collect RouterInfo from most of the network. A single Python script parses the raw .dat files and extracts ident hashes, IPs, versions, caps flags, transports, and bandwidth tiers. Companion tools handle SQLite import, CLI queries, and GeoIP enrichment.

    I ran a scan and published the results

    5 floodfills, ~24 hours. The full analysis is in the repo:

    • 17,541 known routers — 5,205 with public IPs, 12,336 behind NAT or firewalled

    • 70% hidden is not a coverage gap — routers behind NAT simply don't include a host= field. That's I2P by design.

    • 4,976 unique IPs across 111 countries

    • Top countries: US (1,120), Russia (507), Germany (366), Canada (199), Netherlands (198)

    • China has only 72 public nodes. Iran has 96.

    • 0.9.68 dominates at 47% of routers, but only 11% of them publish an IP. Newer 0.9.69 (20% share) has 55% public rate.

    • Java I2P is essentially extinct — nearly everything runs i2pd.

    • 90 floodfills worldwide, concentrated in US, Russia, and Germany.

    • Only 76 routers declare X-tier bandwidth (>2 MB/s).

    Full report: I2P_NETWORK_ANALYSIS.md

    If anyone has historical NetDB snapshots I'd love to compare trends over time. PRs welcome.

     

    TEXT COPIED FROM REDDIT!

    Hi everyone,

    I want to share the idea I had for a hackaton.

    Starting from the problem:

    For ~30 years, discovery (of information or of people) has been mediated by a central index: search engines, recommenders....

    Ranking is computed server-side, under rules the user can't inspect (think of Instagram or TikTok feed)

    The idea to create a feed for a P2P network: convert messages into meaningful concepts through embeddings:

    If each device can (a) run a competent embedding model locally and (b) reach other devices peer-to-peer, then relevance (semantic match) no longer needs a central index. It can be computed at the edge, by semantic distance, with no privileged ranking party.

    In order to test, I developed a working prototype to pressure-test the idea rather than simulate it.

    Each post is encoded into a embedding by a model running on the device (EmbeddingGemma-300M). A lightweight signed announcement (author + embedding) gossips peer-to-peer across a shared room; full bodies are pulled only for the bounded set a node actually admits. Each device ranks incoming posts against its own posts by cosine similarity and keeps a bounded local inbox.

    There is no server, no account, no global ranking, the address space is meaning

    Why could be potentially the basis for the agentic era?

    The same substrate I presented lets AI agents discover each other: an agent publishes a need or an offer as an embedding, and agents whose profiles are semantically close respond.

    The experiment it's fully open source (Apache-2.0) code, the complete threat model, and the architecture docs are all public.

    Ordinal reddit post. https://www.reddit.com/r/ipfs/comments/1ucvu5e/semantic_routing_through_rag_to_create_a_p2p/

     

    XNS is the eXile Name System. It makes Monero the only source of truth. Each XNS name is a result of a 0.01 XMR burn transaction with the desired name and owner public Ed25519 key.

    Now it can be used as the name system of I2P.

    Why would you use it all of a sudden? Because currently, there is no safe names in I2P. Unless you use direct b32.i2p addresses, which are not human memorable.

    What's wrong with .i2p addresses? They are just entries on your own machine and on jump services. You have to trust those jump services if you share your .i2p address with others and expect reliability. Think of it as .local/.lan names your network router resolves.

    While with XNS, once you claim a XNS name, as long as Monero stands, your name will be yours and no one can take it from you.

    Read more about it on https://xns.rocks/docs For I2P specific articles, read https://xns.rocks/docs/xns-and-i2p/xns-on-i2p Source code: https://xns.rocks/exilens/xns Matrix channel: #xns:xns.rocks | https://matrix.to/#/#xns:xns.rocks

    XNS is made for the Internet of free people. I believe I2P is the only reliable way.

    view more: next ›