Talk for https://dl.acm.org/doi/10.1145/3744169.3744180
1# Steps Towards an Ecology of the Internet 2 3{#authors} 4[**Anil Madhavapeddy**](https://anil.recoil.org)¹, [Sam Reynolds](https://samreynolds.org/)¹, [Alec P. Christie](https://profiles.imperial.ac.uk/a.christie)², [David A. Coomes](https://coomeslab.org)¹, <br> [Michael W. Dales](https://mynameismwd.org)¹, [Patrick Ferris](https://patrick.sirref.org)¹, [Ryan Gibb](https://ryan.freumh.org)¹, [Hamed Haddadi](https://haddadi.github.io/)², [Sadiq Jaffer](https://toao.com)¹, <br> [Josh Millar](https://profiles.imperial.ac.uk/joshua.millar22)², [Cyrus Omar](https://web.eecs.umich.edu/~comar/)³, [William J. Sutherland](https://www.zoo.cam.ac.uk/directory/bill-sutherland)¹, [Jon Crowcroft](https://www.cst.cam.ac.uk/people/jac22 5 6{.university-logos} 7![University of Cambridge](unicam-logo.png){.uni-logo} 8![Imperial College London](imperial-logo.png){.uni-logo} 9![University of Michigan](umich-logo.png){.uni-logo} 10 11¹ *University of Cambridge* &nbsp; ² *Imperial College London* &nbsp; ³ *University of Michigan* 12 13![](ecology-qr.jpg){.qr-code} 14 15Paper: [10.1145/3744169.3744180](https://dl.acm.org/doi/10.1145/3744169.3744180) 16 17**The sixth decennial Aarhus conference: Computing X Crisis** 18*August 18–22, 2025 • Aarhus N, Denmark* 19 20{.remark title="Abstract"} 21The Internet has grown from a small set of protocols for end-to-end connectivity into a critical global system with no builtin "immune system". In the next decade it will likely grow to a trillion nodes and need protection from threats ranging from floods of fake generative data to AI-driven malware. Unfortunately, a breakdown of mutualism across the network means that surveillance capitalism is now the dominant business model. **What can we learn from ecology to restore diversity into the Internet fabric?** 22 23{pause center} 24 25{#coop} 26## The Internet's Cooperative Origins 27 28The Internet is a network-of-network of interconnected nodes, likely hitting 1 trillion this decade! 29It has been a deeply cooperative endeavour, based on the **[end-to-end principle](https://en.wikipedia.org/wiki/End-to-end_principle)**. 30 31> {.definition #e2e-detail title="End-to-end principle"} 32> The end-to-end principle dictates that application-specific functions should 33> reside in edge hosts rather than in the network itself. The network remains 34> "dumb" and just moves bits around. End nodes need to upgrade to evolve with user needs. 35 36{pause center} 37Internet communications protocols are openly documented in thousands of freely available **[Request for Comments](https://www.rfc-editor.org/) (RFCs)** so anyone can build an interoperable version. 38 39> {.definition #rfcs title="The RFC Series, see rfc-editor.org"} 40> The [RFC Series](https://www.rfc-editor.org) contains technical and organizational documents about the Internet, including the specifications and policy documents produced by five streams: the Internet Engineering Task Force (IETF), the Internet Research Task Force (IRTF), the Internet Architecture Board (IAB), Independent Submissions, and Editorial. 41 42> **Further reading:** 43> - [Saltzer, Reed & Clark](https://dl.acm.org/doi/10.1145/357401.357402), ACM Transactions on Computer Systems (TOCS) (1984) 44> - [RFC 3724 "Reflections on the Evolution of the Internet Architecture"](https://www.rfc-editor.org/rfc/rfc3724.html) (2004) 45 46{pause up} 47{#growth-crisis} 48## The Current Growth Crisis 49 50While the Internet's original openness originally made the tech giants 51possible, they have since hit efficiencies of scale that are crowding out the 52original network's diversity. 53 54{pause} 55![Big Internet tech companies](faang.jpg){.faang-overlap} 56 57{.example #os-mono title="System software monocultures dominate the world"} 58- A few operating systems dominate the landscape (Linux, Windows, macOS) 59- Sharing source reinforces this monoculture (Android from Linux, iOS from macOS) 60- Botnets can exploit millions of hosts in minutes due to this similarity 61 62{pause center} 63{.example #middlebox title="Protocol ossification from middleboxes hampers new experiments"} 64- Network middleboxes like home gateways have overly strict protocol interpretations 65- They only rarely update *(when did you last change your broadband router?)* 66- Old IoT devices become botnet targets due to lack of security updates 67 68{pause center} 69{.example #service-central title="Services tend to centralize for convenience and economies of scale"} 70- The fight against email spam resulted in Google becoming dominant for all email 71- Cloud hyperscalars can outcompete small hosting providers with power efficiency 72- Amazon crowded out smaller suppliers for deliveries with Prime 73 74{pause up} 75 76{#bio-analogy} 77## How Nature Fights Monocultures 78 79Nature abhors monocultures since they are brittle to external factors as the environment around them changes. 80A monoculture can collapse very quickly. 81 82![Evolution](evolution.jpg){.evolution-overlap} 83 84{.definition title="Darwin's Theory of Evolution"} 85The theory states that all species of organisms arise and develop through the 86natural selection of small, inherited variations that increase the individual's 87ability to compete, survive, and reproduce. 88 89{pause center} 90Biological processes exhibit both disciplined **self-organization** and **natural selection** for adaptation. 91Would similar concepts apply to Internet systems somehow? 92 93| **Biological** | **Internet** | 94|------------------------|----------------------| 95| DNA | Protocol specifications (RFCs) | 96| Genes | Software modules, reusable code | 97| Proteins | Operating system processes | 98| Virus | Malware, botnets, worms | 99| Cell | End nodes (IoT, servers) | 100| Tissues | Clusters (data centers) | 101| Organisms | Internet services (Google, Facebook) | 102| Ecosystems | Overall Internet resilience | 103 104{pause center} 105### Biological Defense Mechanisms 106 107Nature has evolved remarkably sophisticated defenses against pathogens over millions of years. Let's consider one familiar example... 108 109![Mycorrhizal Fungal Networks](fungi-roots.jpg){.fungi-overlap} 110 111{pause center} 112{.example #fungi title="The Mycorrhizal Fungal Network"} 113> - Fungi take sugars from plants in exchange for nutrients from soil 114> - When roots first appeared the fungi association was already [50 million years old](https://www.healsomerset.org.uk/post/before-roots-by-merlin-sheldrake) 115> - Responds to resource inequality by withholding/supplying nutrients to distant nodes 116> - Individual nodes specialize into cooperators, traders, selfish actors 117> - Soils packed with pathogenic fungi, yet plants have "handshakes" for friends 118> - **Trillions of daily "transactions", yet the protocol hasn't been "exploited"** 119 120{pause up} 121{.remark title="Towards an Ecology of the Internet"} 122How can we learn from these biological systems that have builtin "immune systems" at many levels to kickstart 123more diversity in edge software? 124 125{pause} 126### Artificial Intelligence is the problem, but also possibly the answer 127 128![](arch.png){.arch-small} 129 130AI-driven attacks are extraordinarily effective and getting harder and harder to 131spot due to the rapidity at which they are being deployed across the Internet. 132We need defenses, but **not the same defenses everywhere** to encourage 133diversity. 134 135{pause center} 136{.remark} 137Can we use local AI models to inject **tailored individual 138mutations** into our digital ecosystems? It needs to be safe, but not so safe 139that it takes a long time to deploy: there is protection in being slightly 140different from the herd when a digital pandemic strikes! 141 142*These are all currently thought experiments, and have not been deployed 143on live networks.* 144 145{pause center} 146{.theorem #antibotty title="Antibotty networks for community protection"} 147- What if every host acted as an "antibody" for their local network? 148- Local scanning beats global botnets: scan your friends and family 149- Hosts can protect their neighbors; we are pretty predictable in our behaviours 150- Need to isolate infected hosts, patch vulnerabilities, proxy traffic safely 151 152{pause center} 153![Modern software stack containment points](stack.png){.stack-overlap} 154 155{.theorem #mutations title="Protocol Mutatis Mutandis to jitter code"} 156- Use AI code models to inject targeted code diversity into our local software 157- Biological parallel is DNA mutation or horizontal gene transfer in bacteria 158- Internet protocol architecture separates machine language from machine thoughts 159- Also an opportunity to **individually tailor** software to exactly what we need 160 161{pause up} 162 163## The Centralization Problem 164The Internet has moved from mutualistic to parasitic relationships, with 165surveillance capitalism meaning a few giants extract value from billions. 166 167But from the lens of nature, this makes these giants less fit over to adapt 168to environmental change, as they get "addicted" to advertising revenue! 169 170{.remark #redecentralization title="The redecentralisation movement"} 171- New architectures are emerging to restore the original spirit of the Internet. 172- **Federated Protocols**: ActivityPub (Mastodon, Peertube), ATProto (Bluesky) 173- **User Agency**: Decentralized algorithms give users control over their feeds 174- **Self-Hosted Infrastructure**: Personal clouds, Databoxes, federated networks 175- **Historical Precedent**: Email and DNS show decentralization can work long-term 176 177{pause center} 178## "Rewilding" the Internet 179 180The Internet stands at a crossroads, where it will ossify under central control. 181Such a large network resists monocultures, but we need to nudge it 182in the right direction collectively. 183 184- **Current Choice**: Continue parasitic extraction models or evolve mutualistic cooperation through deliberate engineering. 185- **Our Role**: Foster islands of diversity, support redecentralization, build community borders around common-pool resources ([Ostrom 2015](https://www.cambridge.org/core/books/governing-the-commons/A8BB63BC4A1433A50A3FB92EDBBB97D5)) 186- **Managing Risks**: There are many, but that's life! The frontier is where we discover new ideas, and we may be able to leapfrog closer back to digital equity this time around. 187 188See ["We need to rewild the Internet"](https://www.noemamag.com/we-need-to-rewild-the-internet/), Maria Farrell, Robin Berjon (2024). 189 190{pause up} 191## Discuss these ideas with any of us! 192 193[**Anil Madhavapeddy**](https://anil.recoil.org)¹, [Sam Reynolds](https://samreynolds.org/)¹, [Alec P. Christie](https://profiles.imperial.ac.uk/a.christie)², [David A. Coomes](https://coomeslab.org)¹, <br> [Michael W. Dales](https://mynameismwd.org)¹, [Patrick Ferris](https://patrick.sirref.org)¹, [Ryan Gibb](https://ryan.freumh.org)¹, [Hamed Haddadi](https://haddadi.github.io/)², [Sadiq Jaffer](https://toao.com)¹, <br> [Josh Millar](https://profiles.imperial.ac.uk/joshua.millar22)², [Cyrus Omar](https://web.eecs.umich.edu/~comar/)³, [William J. Sutherland](https://www.zoo.cam.ac.uk/directory/bill-sutherland)¹, [Jon Crowcroft](https://www.cst.cam.ac.uk/people/jac22194 195{.university-logos} 196![University of Cambridge](unicam-logo.png){.uni-logo} 197![Imperial College London](imperial-logo.png){.uni-logo} 198![University of Michigan](umich-logo.png){.uni-logo} 199 200¹ *University of Cambridge* &nbsp; ² *Imperial College London* &nbsp; ³ *University of Michigan* 201 202![](ecology-qr.jpg){.qr-code} 203 204Paper: [10.1145/3744169.3744180](https://dl.acm.org/doi/10.1145/3744169.3744180) 205 206**The sixth decennial Aarhus conference: Computing X Crisis** 207*August 18–22, 2025 • Aarhus N, Denmark* 208 209<div class="credit-line slipshow-credit"> 210 211Built with [Slipshow](https://slipshow.readthedocs.io/en/stable/tutorial.html), thanks [@panglesd](https://github.com/panglesd/)! 212 213</div>