Takahiro Yoshimura
alterakey.bsky.social
Takahiro Yoshimura
@alterakey.bsky.social
#OWASP Saitama Local Chapter leader. #trueseeing / #clairvoyance core developer. Reverser and pentester. Loves #radare2. CISSP.

If you find my work useful, feel free to buy me a coffee: https://buymeacoffee.com/alterakey
先日の #OWASP Saitama MTG #29 トーク #1 についての資料を公開します。単なるふりかえりですが… ありがとうございました。今年はプライバシーやら #LLM やら扱えたかなという気はするものの、赤い話題が足りない…🤔 来年もがんばるぞ、と。よろしくお願いします🙇 #OWASP_Saitama /
speakerdeck.com/alterakey/re...
Reviewing 2025
Reviewing chapter activities in 2025. (OWASP Saitama MTG #29, talk #1)
speakerdeck.com
December 18, 2025 at 2:11 PM
本日です!僕も今年のふりかえりをやります。よろしくお願いします🙇 #OWASP #OWASP_Saitama
今年最後の開催となる #OWASP Saitama MTG #29 は本日12/16 1900〜、さいたま市RaiBoCホール集会室5で開催します!続けて、jetbeejpさんからCTF超超超入門としてお話をいただけることになりました!ハイブリッド開催ですので、オンラインでも参加可能ですよ!ただSkytalkと忘年会は現地限定です!お楽しみに。参加登録はこちらから! #OWASP_Saitama / owaspsaitama.connpass.com/event/376629/
OWASP Saitama MTG #29 (2025/12/16 19:00〜)
## OWASP Saitama MTGとは セキュリティ向上に貢献しようという地元有志がさいたま市あるいは春日部市に集まって開く、 OWASP Saitama チャプターの定例勉強会です。セキュリティに携わる幅広い分野の方をスピーカーとしてお呼びして開催します。職種・経験・年齢・性別・国籍などは一切問いません。どなたでもご参加ください。...
owaspsaitama.connpass.com
December 16, 2025 at 12:05 AM
Reposted by Takahiro Yoshimura
I decided to crack open #beepbox and compose the most epic chiptune song I've ever made.

COSMIC EXPEDITION
tinyurl.com/y3w5tjz9
November 29, 2025 at 8:47 PM
Reposted by Takahiro Yoshimura
I love everything about this. Also, "adversarial poetry" would be an excellent name for a rock album.

arxiv.org/abs/2511.15304
Adversarial Poetry as a Universal Single-Turn Jailbreak Mechanism in Large Language Models
We present evidence that adversarial poetry functions as a universal single-turn jailbreak technique for Large Language Models (LLMs). Across 25 frontier proprietary and open-weight models, curated po...
arxiv.org
November 22, 2025 at 3:24 AM
Arduinoプラットフォームが規約変更によりRE禁止などでもう終わりではないかという視点。ただこれArduino自体には特に影響ないような気がするところだけど... / archive.ph/05KK2
archive.ph
November 21, 2025 at 2:21 AM
LLMの自主規制の突破に詩的な表現が有効であるとのこと。面白いな.. / arxiv.org/abs/2511.15304
Adversarial Poetry as a Universal Single-Turn Jailbreak Mechanism in Large Language Models
We present evidence that adversarial poetry functions as a universal single-turn jailbreak technique for Large Language Models (LLMs). Across 25 frontier proprietary and open-weight models, curated po...
arxiv.org
November 21, 2025 at 1:57 AM
Googleがメッセージなどを読んでモデルのトレーニングに使っている可能性とその対策について。思った通り... / www.malwarebytes.com/blog/news/20...
Gmail can read your emails and attachments to train its AI, unless you opt out
A new Gmail update may allow Google to use your private messages and attachments for AI training. Here's how to turn it off.
www.malwarebytes.com
November 21, 2025 at 12:48 AM
LibreOfficeのJavaScriptマクロでwordleを書く例。元気だな.. / bojidar-bg.dev/blog/2025-11...
Wordle in LibreOffice
I implemented Wordle in LibreOffice with JavaScript!
bojidar-bg.dev
November 19, 2025 at 2:58 AM
タスクを最大限に分解しエラー訂正を行い、かつまずい判断についてフラグする方法論の提案。これでハノイの塔について1M stepもの間error freeで走れたと..面白いな🤔 / arxiv.org/abs/2511.09030
Solving a Million-Step LLM Task with Zero Errors
LLMs have achieved remarkable breakthroughs in reasoning, insights, and tool use, but chaining these abilities into extended processes at the scale of those routinely executed by humans, organizations...
arxiv.org
November 19, 2025 at 2:52 AM
Reposted by Takahiro Yoshimura
If I will no longer be able to use discord, bluesky, or youtube without sending my identity to the government and/or third parties, well, it was nice knowing ya.

Because I don't know what I'd do if they'll actually go through with it. Protest? We'll see.
November 12, 2025 at 11:54 PM
Firefoxのfingerprint protection phase 2とのこと。fontまわりの対策も入ったっぽいな… / www.bleepingcomputer.com/news/securit...
Mozilla Firefox gets new anti-fingerprinting defenses
Mozilla announced a major privacy upgrade in Firefox 145 that reduces even more the number of users vulnerable to digital fingerprinting.
www.bleepingcomputer.com
November 11, 2025 at 5:05 AM
macOSでTouch IDを使うためのPAMモジュールが用意されていたのは2017年ごろからなのか… / mostlymac.blog/2022/03/01/u...
November 10, 2025 at 4:15 AM
open-webuiのFLOSSなalternativeとして。ありがたい.. / github.com/ServiceStack...
GitHub - ServiceStack/llms: LLM Client, Server API and UI
LLM Client, Server API and UI. Contribute to ServiceStack/llms development by creating an account on GitHub.
github.com
November 3, 2025 at 11:11 PM
Reposted by Takahiro Yoshimura
So I wanted to make the NAND implementation in the iPod Touch QEMU fork using mmap, and it seems to work, until you realize, there is no Settings app! (but sometimes it does show up and I can't explain it)

(my fork of the fork is here: github.com/iProgramMC/q...)
October 30, 2025 at 10:51 PM
Reposted by Takahiro Yoshimura
It's Friday, meaning it's time to ship! (Right?!) 🚀

I've just released a minor update to my encrypted-shared-preferences fork

This release bumps to tink version v1.19.0

Check it out:
github.com/ed-george/en...

#AndroidDev
Release 1.1.1 · ed-george/encrypted-shared-preferences
What's Changed Update dependency com.google.android.material:material to v1.13.0 by @renovate[bot] in #43 Update actions/setup-java action to v5 by @renovate[bot] in #42 Update actions/checkout ac...
github.com
October 31, 2025 at 9:10 AM
当然なのかもしれないが #bluesky を見ていると往年のTwitterを思い出すな… なんか落ち着く🤔
October 30, 2025 at 8:02 AM
先日のOWASP Saitama MTG #28 トーク #1 についての資料を公開します。前回に引き続き、2023年の論文をもとに攻撃的なプロンプトの詳細に触れつつ新しいLLMへの有効性について考察したものです。ありがとうございました。 #OWASP
#OWASP_Saitama / speakerdeck.com/alterakey/re...
Repeat After Me #2
Deeper, detailed dive into the prompt injection attacks against LLMs, evaluating their effectiveness against recent ones. (OWASP Saitama MTG #28, talk #1)
speakerdeck.com
October 30, 2025 at 2:49 AM
(bridgeは立ててあるのでMastodonの方々もなにとぞ引き続き。)
October 30, 2025 at 2:07 AM
Xを閉じたあとしばらくMastodon (infosec.exchange) にいたのだけど… Blueskyに引っ越すことに。よろしくお願いしますmm
Infosec Exchange
A Mastodon instance for info/cyber security-minded people.
infosec.exchange
October 30, 2025 at 1:10 AM