Com­puter Se­cu­rity & Cryptography

Se­cu­rity Mind­set and Or­di­nary Paranoia

On De­stroy­ing the World

Bet­ter Pass­word Peppering

Work on Se­cu­rity In­stead of Friendli­ness?

En­tropy isn’t suffi­cient to mea­sure pass­word strength

POC || GTFO cul­ture as par­tial an­ti­dote to al­ign­ment wordcelism

AI Safety is Drop­ping the Ball on Clown Attacks

The dan­ger of er­ror descriptions

Ho­mo­mor­phic en­cryp­tion and Bitcoin

What is Cryp­to­graph­i­cally Possible

Cryp­to­graphic Boxes for Un­friendly AI

Paper re­view: A Cryp­to­graphic Solu­tion to a Game The­o­retic Problem

Crypt­anal­y­sis as Episte­mol­ogy? (pag­ing cryp­ton­erds)

[Question] How good is se­cu­rity for LessWrong and the Align­ment Fo­rum?

Se­cure homes for digi­tal people

gamers be­ware: mod­ded Minecraft has new malware

Up­skil­ling, bridge-build­ing, re­search on se­cu­rity/​cryp­tog­ra­phy and AI safety

[Question] Is there a con­ve­nient way to make “sealed” pre­dic­tions?

The Un­bear­able Light­ness of Web Vulnerabilities

[Question] How much does cy­ber­se­cu­rity re­duce AI risk?

Crypto-fed Computation

Se­cu­rity Mind­set: Les­sons from 20+ years of Soft­ware Se­cu­rity Failures Rele­vant to AGI Alignment

Ac­cess to pow­er­ful AI might make com­puter se­cu­rity rad­i­cally easier

Uniker­nels: No Longer an Aca­demic Exercise

5 Rea­sons Why Govern­ments/​Mili­taries Already Want AI for In­for­ma­tion Warfare

[Question] What is good Cy­ber Se­cu­rity Ad­vice?

[Question] In soft­ware en­g­ineer­ing, what are the up­per limits of Lan­guage-Based Se­cu­rity?

Hacker-AI – Does it already ex­ist?

n of m ring signatures

Prevent­ing model exfil­tra­tion with up­load limits

Reli­a­bil­ity, Se­cu­rity, and AI risk: Notes from in­fosec text­book chap­ter 1

Not Get­ting Hacked

Pri­vacy Tradeoffs

How Likely is Los­ing a Google Ac­count?

Free­dom and Pri­vacy of Thought Architectures

AI Gover­nance & Strat­egy: Pri­ori­ties, tal­ent gaps, & opportunities

[Question] Hoard­ing Gmail-ac­counts in a post-CAPTCHA world?

AI se­cu­rity might be helpful for AI alignment

11 dice­ware words is enough

Fron­tier Model Security

AXRP Epi­sode 30 - AI Se­cu­rity with Jeffrey Ladish

The rise of AI in cybercrime

Biose­cu­rity Cul­ture, Com­puter Se­cu­rity Culture

[Question] Trans­fer­ring cre­dence with­out trans­fer­ring ev­i­dence?

Im­proved Se­cu­rity to Prevent Hacker-AI and Digi­tal Ghosts

Hacker-AI and Digi­tal Ghosts – Pre-AGI

Deep­fake(?) Phishing

Why do we post our AI safety plans on the In­ter­net?

Safe Devel­op­ment of Hacker-AI Coun­ter­mea­sures – What if we are too late?

Non-Tech­ni­cal Prepa­ra­tion for Hacker-AI and Cy­ber­war 2.0+

Hacker-AI and Cy­ber­war 2.0+

The 2/​3 rule for multi-fac­tor authentication

Re­vis­it­ing the Man­i­fold Hypothesis

Nav­i­gat­ing the Attackspace

Plan­ning to build a cryp­to­graphic box with perfect secrecy

Agent mem­branes and causal distance

Disprov­ing and par­tially fix­ing a fully ho­mo­mor­phic en­cryp­tion scheme with perfect secrecy

Pro­tect­ing agent boundaries

In­cor­po­rat­ing Mechanism De­sign Into De­ci­sion Theory

How im­por­tant is AI hack­ing as LLMs ad­vance?

PoMP and Cir­cum­stance: Introduction

End-to-end hack­ing with lan­guage models

Can Large Lan­guage Models effec­tively iden­tify cy­ber­se­cu­rity risks?

Cy­ber­se­cu­rity of Fron­tier AI Models: A Reg­u­la­tory Review

How to Fake Decryption

Can star­tups be im­pact­ful in AI safety?

Con­sider at­tend­ing the AI Se­cu­rity Fo­rum ’24, a 1-day pre-DEFCON event

Se­cret Col­lu­sion: Will We Know When to Un­plug AI?

Us­ing an LLM per­plex­ity filter to de­tect weight exfiltration

The Prag­matic Side of Cryp­to­graph­i­cally Box­ing AI

Join the $10K Au­toHack 2024 Tournament

GPT-4o Guardrails Gone: Data Poi­son­ing & Jailbreak-Tuning

AI in­fosec: first strikes, zero-day mar­kets, hard­ware sup­ply chains, adop­tion barriers

Boundaries-based se­cu­rity and AI safety approaches

The Virus—Short Story

Cryp­to­graphic and aux­iliary ap­proaches rele­vant for AI safety

Prevent­ing AI Mi­suse: State of the Art Re­search and its Flaws

[Question] How much should e-sig­na­tures have to cost a coun­try?

[Question] AI in­ter­pretabil­ity could be harm­ful?

Us­ing GPT-Eliezer against ChatGPT Jailbreaking

Microsoft and Google us­ing LLMs for Cybersecurity

Trans­former Ar­chi­tec­ture Choice for Re­sist­ing Prompt In­jec­tion and Jail-Break­ing Attacks

Scry­ing for out­comes where the prob­lem of deep­fakes has been solved

Es­ti­mat­ing the con­se­quences of de­vice de­tec­tion tech

On See­ing Through ‘On See­ing Through: A Unified The­ory’: A Unified Theory

Com­puter se­cu­rity story

Se­cu­rity Mind­set and the Lo­gis­tic Suc­cess Curve

Op­ti­mal User-End In­ter­net Se­cu­rity (Or, Ra­tional In­ter­net Brows­ing)

Proof of pos­te­ri­or­ity: a defense against AI-gen­er­ated misinformation

Pri­vacy in a Digi­tal World

Neu­ral pro­gram syn­the­sis is a dan­ger­ous technology

Ap­ply to the Con­cep­tual Boundaries Work­shop for AI Safety

Fix­ing In­sider Threats in the AI Sup­ply Chain

Lat­a­cora might be of in­ter­est to some AI Safety organizations