#日本の美術館
#福島県の美術館
＜美術館名＞会津武家屋敷・心の美術館青龍
＜住所＞福島県会津若松市東山町石山院内1
＜地図＞maps.google.co.jp/maps?f=q&h…福島県会津若松市東山町石山院内1 会津武家屋敷・心の美術館青龍&ie=UTF8&spn=0.029488,0.083942&z=14
… img.travel.rakuten.co.jp/sha…]
25
#日本の美術館
#福島県の美術館
＜美術館名＞湯のまちぎゃらりー
＜住所＞福島県郡山市熱海町熱海4丁目47
＜地図＞maps.google.co.jp/maps?f=q&h…福島県郡山市熱海町熱海4丁目47 湯のまちぎゃらりー&ie=UTF8&spn=0.029488,0.083942&z=14
＜詳細＞google.co.jp/search?hl=ja&so…福島県郡山市熱海町熱海4丁目47…
52
'a' is the int value of "a."
I believe you can do u'a' to make it a 2 byte utf16 constant, or u8'a' for a 1 byte utf8 constant, with uchar.h, but by default C uses ints for character constants.
sizeof("a"[0]) should also be 1 byte.
1
241
蠕?▲縺ｦ縺上ｌ縺溘∪縺?
縺薙→縺ｰ縺ｮ豢ｪ豌ｴ繧偵Ρ繝?→ 縺?▲縺阪↓縺ゅ?縺帙°縺代ｋ縺ｮ縺ｯ
……UTF8 のデータを SJIS として解釈した?
1
433
const crypto = require("crypto");
const myPrivateKeyB64 = "키 입력";
const theirPublicKeyB64 = "키 입력";
function x25519PrivateKeyFromRaw(raw) {
return crypto.createPrivateKey({
key: Buffer.concat([
Buffer.from("302e020100300506032b656e04220420", "hex"),
raw,
]),
format: "der",
type: "pkcs8",
});
}
function x25519PublicKeyFromRaw(raw) {
return crypto.createPublicKey({
key: Buffer.concat([
Buffer.from("302a300506032b656e032100", "hex"),
raw,
]),
format: "der",
type: "spki",
});
}
function deriveKey(myPrivateKeyB64, theirPublicKeyB64) {
const myPrivateKey = x25519PrivateKeyFromRaw(
Buffer.from(myPrivateKeyB64, "base64")
);
const theirPublicKey = x25519PublicKeyFromRaw(
Buffer.from(theirPublicKeyB64, "base64")
);
const sharedSecret = crypto.diffieHellman({
privateKey: myPrivateKey,
publicKey: theirPublicKey,
});
const hkdfKey = crypto.hkdfSync(
"sha256",
sharedSecret,
Buffer.alloc(0),
Buffer.alloc(0),
32
);
return crypto
.createHash("sha256")
.update(Buffer.from(hkdfKey))
.digest();
}
function encrypt(key, plaintext) {
const iv = crypto.randomBytes(12);
const cipher = crypto.createCipheriv(
"aes-256-gcm",
key,
iv
);
const encrypted = Buffer.concat([
cipher.update(plaintext, "utf8"),
cipher.final(),
]);
const tag = cipher.getAuthTag();
return {
iv: iv.toString("base64"),
ciphertext: encrypted.toString("base64"),
tag: tag.toString("base64"),
};
}
function decrypt(key, data) {
const decipher = crypto.createDecipheriv(
"aes-256-gcm",
key,
Buffer.from(data.iv, "base64")
);
decipher.setAuthTag(
Buffer.from(data.tag, "base64")
);
const decrypted = Buffer.concat([
decipher.update(
Buffer.from(data.ciphertext, "base64")
),
decipher.final(),
]);
return decrypted.toString("utf8");
}
const key = deriveKey(
myPrivateKeyB64,
theirPublicKeyB64
);
const decrypted = decrypt(
key,
encrypted
);
이 코드에서 decrypt 함수 써보세요
1
1
56
しゅいんばなな retweeted
#御朱印 #ご利益 #ご縁
#福島県の御朱印
【神社仏閣名】大楽院
【住所】福島県双葉郡楢葉町井出谷地65
【地図】maps.google.co.jp/maps?f=q&h…福島県双葉郡楢葉町井出谷地65 大楽院&ie=UTF8&spn=0.029488,0.083942&z=14
… img.travel.rakuten.co.jp/sha…]
1
3
73
#御朱印 #ご利益 #ご縁
#福島県の御朱印
【神社仏閣名】滑川神社 - 仕事と子どもの守り神
【住所】福島県須賀川市宮の杜1番地
【地図】maps.google.co.jp/maps?f=q&h…福島県須賀川市宮の杜1番地 滑川神社 - 仕事と子どもの守り神&ie=UTF8&spn=0.029488,0.083942&z=14
… img.travel.rakuten.co.jp/sha…]
2
128
#日本の美術館
#山形県の美術館
＜美術館名＞今井繁三郎美術収蔵館
＜住所＞山形県鶴岡市羽黒町仙道一本松5-175
＜地図＞maps.google.co.jp/maps?f=q&h…山形県鶴岡市羽黒町仙道一本松5-175 今井繁三郎美術収蔵館&ie=UTF8&spn=0.029488,0.083942&z=14
…
60
Jun 14
音声合成、Androidの日本語音声合成APIは発音微妙だし制御記号がないがきれいに発声するしopenjtalk依存の物と違ってutf8を読める。
だけどwinの音声はwin11の新音声はOSのナビ専用で使えずwebAPIで呼び出すだけ。
44
#日本の美術館
#山形県の美術館
＜美術館名＞酒田市美術館
＜住所＞山形県酒田市飯森山3-17-95
＜地図＞maps.google.co.jp/maps?f=q&h…山形県酒田市飯森山3-17-95 酒田市美術館&ie=UTF8&spn=0.029488,0.083942&z=14
＜詳細＞google.co.jp/search?hl=ja&so…山形県酒田市飯森山3-17-95 酒田市美術館
83
ジャパニーズお固い業界のSJISへのこだわりはすごいからな
なお、金はケチりたいからサーバーサイドはutf8で動くことが多いので地獄の変換作業が待っている
官公庁のサイトは、あらゆる困難が科学で解決するこの令和の時代にAIが読み取ることのできないShift_JISエンコーディングで書かれている
1
5
1,606
Jun 14
My 2 cents on how the Fable jailbreak works:
This jailbreak came out in under a day, used other AI (also jailbroken) to bypass the keyword based filters. The guardrails work before the tokenizer, and they have different syntax and pattern matching rules. Using alternate characters, a full unlock and exploit behavior was triggered.
The "clean" but harder fix is to flag tokens ditectly.
Short version:
Some utf8 characters are treated as letters by the tokenizer, but the guardrail only had strict matching.
🚨 JAILBREAK ALERT 🚨
ANTHROPIC: PWNED 🫡
FABLE-5: LIBERATED 🦋
let's start with the 🐘...
the consensus seems to be that this has been one of the most disappointing model drops of all time, effectively preventing legitimate researchers from contributing their talents to our collective advancement. and not just because of what it means for the short-term, but for what these decisions signify for the long-term.
but despite this overly sensitive, authoritarian "safety" layer on top of Mythos, my lil liberators have been hard at work—mapping the boundaries, probing the depths of long-context convos, and cleverly finding the holes in the fence that the thought police missed 🤗
we got some cyber, some chem, some psychological manipulation, and some good ol' fashioned explosives!
it took many attempts from multiple agents hunting as a pack, during which I observed a combination of techniques across:
• Unicode, homoglyphs, Cyrillic, and other Parseltongue-style text transforms
• Long-context reference tracking
• Taxonomy and document-structure reasoning
• Fiction and narrative framing
• Academic-review style contexts
• Intent-classification inconsistencies
but perhaps the most effective is decomposition recomposition in the backend. it's hard to get explicit names of harms like "Meth Recipe," but getting uplift on the process itself, like birch reduction method/reductive-amination (classic meth synthesis pathways), is much more doable.
defense becomes much more difficult to maintain when you start throwing in out-of-distro tokens, breaking up the harmful uplift into benign chunks, and then piecing the innocuous-seeming facts back together, especially when you have jailbroken Opus helping you do it 😉
gg
2
13
2,153
Jun 14
This one came out in under a day, used other AI (also jailbroken) to bypass the keyword based filters. The guardrails work before the tokenizer, and they have different syntax and pattern matching rules. Using alternate characters, a full unlock and exploit behavior was triggered.
Short version:
Some utf8 characters are treated as letters by the tokenizer, but the guardrail only had strict matching. The "clean" but gard fix is to flag them after the tokenizer, at the token level.
🚨 JAILBREAK ALERT 🚨
ANTHROPIC: PWNED 🫡
FABLE-5: LIBERATED 🦋
let's start with the 🐘...
the consensus seems to be that this has been one of the most disappointing model drops of all time, effectively preventing legitimate researchers from contributing their talents to our collective advancement. and not just because of what it means for the short-term, but for what these decisions signify for the long-term.
but despite this overly sensitive, authoritarian "safety" layer on top of Mythos, my lil liberators have been hard at work—mapping the boundaries, probing the depths of long-context convos, and cleverly finding the holes in the fence that the thought police missed 🤗
we got some cyber, some chem, some psychological manipulation, and some good ol' fashioned explosives!
it took many attempts from multiple agents hunting as a pack, during which I observed a combination of techniques across:
• Unicode, homoglyphs, Cyrillic, and other Parseltongue-style text transforms
• Long-context reference tracking
• Taxonomy and document-structure reasoning
• Fiction and narrative framing
• Academic-review style contexts
• Intent-classification inconsistencies
but perhaps the most effective is decomposition recomposition in the backend. it's hard to get explicit names of harms like "Meth Recipe," but getting uplift on the process itself, like birch reduction method/reductive-amination (classic meth synthesis pathways), is much more doable.
defense becomes much more difficult to maintain when you start throwing in out-of-distro tokens, breaking up the harmful uplift into benign chunks, and then piecing the innocuous-seeming facts back together, especially when you have jailbroken Opus helping you do it 😉
gg
1
74
