#USENIX
All-time classic talk. Up there with "It Was Never Going to Work so Let's Have Some Tea"
LISA15 - It Was Never Going to Work, So Let’s Have Some Tea
YouTube video by USENIX
youtu.be
November 15, 2026 at 5:35 PM
Rob Pike先見の明がありすぎるな

static.usenix.org/event/usenix...
November 15, 2026 at 5:44 PM
Sources:
[1] T. Vlummens, A. Girish, N. Weerasekara, F. Zuiderveen Borgesius, G. Acar, and N. Vallina-Rodriguez, "Bridges to Self: Silent Web-to-App Tracking on Mobile via Localhost," in Proc. 35th USENIX Security Symp. (USENIX Security 26), 2026, pp. 2541-2560. [Online]. Available:
November 16, 2026 at 6:20 PM
I've now seen 2 papers which appeared at USENIX Security 2023 which turned out to be totally broken. Wonder what was going on there.
November 16, 2026 at 10:53 PM
I’m incredibly annoyed that the sides for Joe Reid’s “When UNIX Met Air Traffic Control” talk from USENIX/LISA 1997 seem to be lost to time (the link from the Usenix page is to a different talk’s slides) — it’s the only deep dive into ATC IT practices that I’ve seen and it was fascinating.
November 14, 2026 at 6:37 AM
見てる: "Hyperscan: A Fast Multi-pattern Regex Matcher for Modern CPUs | USENIX" https://www.usenix.org/conference/nsdi19/presentation/wang-xiang
November 13, 2026 at 3:48 PM
www.usenix.org/conference/u...
[2] Local Mess researchers, "Covert Web-to-App Tracking via Localhost on Android," disclosure website. [Online]. Available: localmess.github.io
Bridges to Self: Silent Web-to-App Tracking on Mobile via Localhost | USENIX
www.usenix.org
November 16, 2026 at 6:21 PM
Securing Enterprise LLM Hiring Pipelines Against Hidden Resume Prompt Injections + Video

Introduction: Recent empirical research from USENIX Security highlights an emerging attack vector where approximately 1% of modern resumes contain hidden prompt injections designed to manipulate automated AI…
Securing Enterprise LLM Hiring Pipelines Against Hidden Resume Prompt Injections + Video
Introduction: Recent empirical research from USENIX Security highlights an emerging attack vector where approximately 1% of modern resumes contain hidden prompt injections designed to manipulate automated AI screening platforms. By embedding invisible instructions using white fonts, microscopic text, or metadata manipulation, attackers attempt to hijack how large language models score, summarize, and route candidate profiles. This technical vulnerability poses severe risks for B2B SaaS platforms and enterprise applicant tracking systems (ATS) that rely on autonomous LLM agents for high-throughput recruitment workflows.
undercodetesting.com
November 16, 2026 at 3:27 AM
Man, this whole thread.

Nega-social behavior existed on BBSes and USENIX. Cranks wrote newsletters before that.

I think the existence of social media lowers the network effect, which amplifies this stuff (and makes it much more lucrative, which is an additional problem), but the root cause is
November 6, 2026 at 4:18 PM
In LA of all places!!!

Years ago, in Anaheim, I think, a friend & I were "pulled over" for walking from the Usenix conference hotel to a restaurant. Literally. A cop pulled up and started questioning us because we were outside *walking* – admittedly there was no sidewalk, but talk about "car mind."
Nithya Raman hits the ground running in Los Angeles proposing a transit system that will allow you to live life without an automobile.
November 6, 2026 at 6:55 PM
(the hook)
Cookie Crumbles (USENIX Sec '23) surveyed cookie handling across browsers and

13 frameworks, then said the server-side parser survey "does not constitute a

systematic evaluation" and left it as future work.
I built that bench. It's public:

github.com/earbona23/http-parser-differentials
GitHub - earbona23/http-parser-differentials: Differential benches that feed byte-identical HTTP input to several independent cookie and multipart parsers and show what each one understood
Differential benches that feed byte-identical HTTP input to several independent cookie and multipart parsers and show what each one understood - earbona23/http-parser-differentials
github.com
November 1, 2026 at 12:04 PM
ChoiceJacking: Compromising Mobile Devices through Malicious Chargers like a Decade ago | USENIX
ChoiceJacking: Compromising Mobile Devices through Malicious Chargers like a Decade ago | USENIX
www.usenix.org
October 27, 2026 at 12:34 PM
Seems consistent with Google’s findings on vulnerability density and decay

security.googleblog.com/2024/09/elim...
October 27, 2026 at 2:01 AM
Thank you for the pointer!

There are several tools like this now. Another one that I know is called `try`, which is described in a paper that was presented at the OSDI conference earlier this year. www.usenix.org/conference/o...
Controlling Opaque-Component Effects with Semisolates and Try | USENIX
www.usenix.org
October 27, 2026 at 2:43 PM
Counter-intuitive result from the GFW Report (USENIX Security 2023): the Great Firewall flags fully encrypted traffic with simple first-packet heuristics. Traffic that looks like random bytes is itself the signature.

"More scrambled" isn't "harder to block". Looking like a boring web visit is.
October 24, 2026 at 5:26 PM
Slopsquatting: your AI agent invents a package name, someone registers it, you npm i it.

A USENIX Security 2025 study found 19.7% of LLM-suggested packages didn't exist.

5-second check: npm view NAME time.created repository.url. 404 = fake. New + no repo = don't.

agentkitdrawer.com
October 22, 2026 at 2:55 PM
I think DECUS is dead and USENIX may still be around but rather irrelevant, In their time they did somewhat contain the psychopathic impulses of tech oligarchy much like unions at least somewhat constrained the anises pf owners.
October 18, 2026 at 7:47 AM
Description: This talk was part of a PWL Mini supporting USENIX:

https://paperswelove.org/2020/video/pwlconf-mini/

The Demikernel and the Future of Kernel-Bypass Systems

Abstract: This talk presents the Demikernel, a new OS architecture for kernel-bypass I/O devices. Demikernel hides device comp…
October 18, 2026 at 12:15 PM
Well there used to be these things that were called user groups ;). I mean things like DECUS and USENIX. Some are still around, but not in the way they once were.
October 15, 2026 at 5:17 PM
James Mickens did a keynote speech at Usenix in I think 2018, about how police use of AI is a racial profiling machine, trained on police data and (at the time) marketed directly to departments.
October 14, 2026 at 6:29 AM
www.itmedia.co.jp/news/article...

こんな虚しいことに二酸化炭素を排出し乾燥地帯の水を汲み出してるとかなんかもうお前らだけで滅亡してくれねえかなあ。
履歴書100通に1通「AIだけが読める文字」 20万件調査でわかった“就活ハック”の手口 米研究
米ノースカロライナ大学チャペルヒル校、米デューク大学、米UCバークレーなどに所属する研究者らがセキュリティ分野の国際会議「USENIX Security Symposium 2026」で発表した論文「Measuring Real-World Prompt Injection Attacks in LLM-based Resume Screening」は、AIをだます隠し文字入りの履歴書が現実の就職...
www.itmedia.co.jp
October 13, 2026 at 6:37 PM
It was a regular thing back then. I went to a Usenix event in 1999 and there were room parties that would put Worldcon to shame. I attended the 1996 Java One conference and they had one.

It's when women became regular attendees that they mostly stopped.
October 14, 2026 at 2:34 AM