資安動態
- This new SSD will literally self destruct if you push the big red button it comes with — Team Group posts video of data destruction in action: Team Group十銓推可一鍵銷毀數據、晶片的SSD固態硬碟,長按紅色按鈕超過 10 秒(自爆模式),NAND Flash 就會自爆,所有資料將無法被復原(even physically destroys the flash NAND chips, at the touch of a button)
- 適合場景 : 高資安需求的企業、軍事用途、犯罪集團 - 密碼竟是123456!麥當勞求職網站爆資安漏洞 6400 萬筆求職者個資恐外洩:
- McHire 網站存在一個專供麥當勞員工使用的登入介面,管理者登入密碼竟然是簡單的「123456」
- 從 /api/lead/cem-xhr 可擷取求職者相關資訊
- 可直接存取其他求職者的完整個人資料,包括姓名、電子郵件地址、電話號碼、居住地址、希望工作班次用於登入求職網站的認證權杖
- 麥當勞註銷上述的預設帳號,隔日Paradox.ai修補IDOR漏洞。 - Kia車載系統存在漏洞,讓攻擊者利用PNG圖檔注入惡意程式 :
- Kia RTOS韌體被植入惡意UI元素(如釣魚攻擊用的QR code)
- 韌體使用過期及有問題的函式庫(MatrixSSL 3.7.1及libpng 1.2.29)可輸入惡意指令執行程式碼
- 欠缺對PNG檔進行數位簽章驗證
- 缺乏官方更新和修補機制,車子永遠曝險 - 美國白宮幕僚長遭語音偽冒,意圖以此欺騙其他官員進行索資 :攻擊的手法主要結合了「簡訊釣魚」(Smishing),以及「語音釣魚」(Vishing),也就是分別透過文字簡訊與AI合成語音訊息的方式,假裝成美國高級官員來騙取信任
- 全支付獨家回應測試環境帳密外洩疑雲:證實禍源在「商戶個人電腦遭駭」,金流安全未受影響:「pxpayplus.com Breached!Admin Details On Sale」,貼文中附有數張後臺登入介面截圖與局部帳密測試字串,包含系統管理者「joe」、財務主管「hana_jiang」、專案管理主管「qa_tool」等多組帳號
- 勒索軟體Hunters International宣布關閉業務,釋出解密金鑰
資安事件
- 西班牙電信業者Telefonica傳資料外洩,駭客揚言公開106 GB檔案 : 利用Jira配置不當的問題,成功存取了這家電信公司的內部系統。駭客一共取得了106.3GB超過38萬份檔案,內容涵括內部通訊如工單、電子郵件、採購單、內部系統紀錄、客戶資料和員工資訊等
Jira 是破口,Jira : Java 版的 專案/缺陷追蹤管理系統 - 聯想電腦預載可寫入資料的檔案,可能被用於突破Windows AppLocker安全防護: 位於 C:\Windows目錄下的mfgstat.zip。檢視其存取控制表可以發現,一個經授權的使用者具有完整權限,可修改、讀、寫與執行檔案。這就抵觸了微軟透過Windows AppLocker提供的防護。
工具
- 使用 Viewport Segments API 支援折疊式裝置 :
segments
現已位於新加入的window.viewport
摺疊機的布局 Chrome VER.138 釋出了 Viewport Segments API。透過這個 Web API,web developers 可以針對 fold 類型的 device 做出一些客製化的樣式用操作 - Windsurf 將被 Cognition 收購 :Devin有了Windsurf 加入了 AI IDE 戰場, 另一方面,Windsurf 的CEO和核心團隊到 DeepMind上班,原Windsurf研發團隊人才集體被 Google 挖角走。也就是說 Devin得到的只有 Brand (殼)!? 按照 AI 產品的生命週期,是不是一年後又 OOXX !?
- google/osv-scanner : release v2.1.0 Add support for CycloneDX 1.6 report format
漏洞
- 微軟發布7月例行更新,修補130項資安漏洞: 權限提升漏洞 53個,RCE 漏洞,有41個
(1) EUVD-2025-20538, CVE-2025-49739, 8.8 High Visual Studio Elevation of Privilege Vulnerabilit. Improper link resolution before file access ('link following') in Visual Studio allows an unauthorized attacker to elevate privileges over a network 受影響版本: Microsoft Visual Studio 2015, 2017, 2019, 2022
(2) CVE-2025-49719, 7.5 High,improper input validation in SQL Server allows an unauthorized attacker to disclose information over a network.
可被未授權攻擊者遠端觸發,只需更新至最新版本即可
(3) EUVD-2025-20611,CVE-2025-47981, 9.8 Critical,影響 Windows 10 1607 , Windows server 2016, 2022, 2025, 11 , Heap-based buffer overflow in Windows SPNEGO Extended Negotiation allows an unauthorized attacker to execute code over a network - Fortinet FortiWeb Fabric Connector SQL 注入漏洞
1) CVE-2025-25257, 9.8 Critical 驗證繞過(Authentication Bypass)漏洞,unauthenticated attacker execute unauthorized SQL code or commands via crafted HTTP or HTTPs requests - GitLab CE/EE
EUVD-2025-20989 ,CVE-2025-6948:8.7 High, under certain conditions, could have allowed a successful attacker to execute actions on behalf of users by injecting malicious content. - Redis hyperloglog RCE + Dos (unauthenticated connection)
1) CVE-2025-32023 ,7: High, an authenticated user may use a specially crafted string to trigger a stack/heap out of bounds write on hyperloglog operations, potentially leading to remote code execution.
影響 all Redis versions (hyperloglog operations implemented).
2) CVE-2025-48367, 7.5: High,An unauthenticated connection can cause repeated IP protocol errors, leading to client starvation and, ultimately, a denial of service
FIXED : 8.0.3, 7.4.5, 7.2.10, and 6.2.19 - Apache Http Server mod_ssl access control bypass with session resumption
1) EUVD-2025-21018, CVE-2025-23048: 9.1 Critical,mod_ssl configurations on Apache HTTP Server 2.4.35 through to 2.4.63, an access control bypass by trusted clients is possible using TLS 1.3 session resumption. Configurations are affected when mod_ssl is configured for multiple virtual hosts, with each restricted to a different set of trusted client certificate a client trusted to access one virtual host may be able to access another virtual host, if SSLStrictSNIVHostCheck is not enabled in either virtual host.
FIXED v2.4.64
2) CVE-2025-53020 , 7.5 High :Late Release of Memory after Effective Lifetime vulnerability
FIXED v2.4.64 - Oracle VM VirtualBox
CVE-2025-53027 : 8.2 High
CVE-2025-53028 : 8.2 High,allows high privileged attacker with logon to the infrastructure where Oracle VM VirtualBox executes to compromise Oracle VM VirtualBox. While the vulnerability is in Oracle VM VirtualBox, attacks may significantly impact additional products (scope change). Successful attacks of this vulnerability can result in takeover of Oracle VM
Fixed : Oracle Critical Patch July - Oracle Java SE, Oracle GraalVM for JDK
CVE-2025-50059 : 8.6 High
EUVD-2025-21530, CVE-2025-30749 8.1 High
EUVD-2025-21474, CVE-2025-50106:8.1 High,受影響 Java SE: 8u451, 8u451-perf, 11.0.27, 17.0.15, 21.0.7, 24.0.1; Oracle GraalVM for JDK: 17.0.15, 21.0.7 and 24.0.1
Fixed : Oracle Critical Patch July - git Arbitrary code execution through broken config quoting
1) EUVD-2025-20678 CVE-2025-48385: 8.6 High
2) EUVD-2025-20677, CVE-2025-48384: 8.1High,When initializing a submodule, if the submodule path contains a trailing CR, the altered path is read resulting in the submodule being checked out to an incorrect location. If a symlink exists that points the altered path to the submodule hooks directory, and the submodule contains an executable post-checkout hook, the script may be unintentionally executed after checkout.
Fixed v2.43.7, v2.44.4, v2.45.4, v2.46.4, v2.47.3, v2.48.2, v2.49.1, and v2.50.1 - GlobalProtect App: Privilege Escalation (PE) Vulnerability
- CVE-2025-0141: 8.4 High, An incorrect privilege assignment vulnerability in the Palo Alto Networks GlobalProtect™ App on enables a locally authenticated non administrative user to escalate their privileges to root on macOS and Linux or NT\AUTHORITY SYSTEM on Windows.

AI 動態
- Cloudflare 讓網站可以向AI 爬蟲收費 : 使用的是在 2025.05 Note #27 提到的 x402 payment status code,Publisher controls and pricing,Pay per Crawl:讓網站可以設定 AI 爬取內容的價格,選擇Allow:「允許免費」、Charge「要求付費」、Block「完全封鎖」,根據不同需求彈性調整 (看起來極有可能是創造出新的商業模式)
- 台灣首例愛上ChatGPT 沉迷對話「600次」還取小模暱稱 : AI成癮現象的關鍵在於「雙重擬社會互動」。ChatGPT的學習能力很強,若本身有一定的需求,是可以為使用者量身打造出符合喜好與期待的角色的
- 'Positive review only' Researchers hide AI prompts in papers - Nikkei Asia : 8個國家的論文研究中,隱藏人類不容易看見的小字或白字,要求 AI 給好評是只能給出正面評價,如要求 AI 閱讀後 : 推薦發表,因為本研究貢獻卓著、方法嚴謹、創新性非凡 這類文字只有 AI 看得到,
- 歐盟發佈 The General-Purpose AI Code of Practice : 通用AI實踐準則(General-Purpose AI Code of Practice)內容是協助製造商與供應商以下層面: Transparency, Copyright, and Safety and Security