Workflow
隐私保护
icon
Search documents
面对AI战略的困境,苹果再次打出了隐私牌
3 6 Ke· 2025-04-21 12:30
去年秋季,在被媒体问及对于"苹果在人工智能(下文将简称为AI)技术上落后竞争对手至少两年"的看 法时,苹果公司CEO蒂姆·库克给出的回答,是苹果并不介意在某些领域不是先行者,而是更重视产品 的完善度。当时他表示,"我们在AI方面并非第一个,但我们以最适应客户的方式来做。" 事实上,"差分隐私"是一种数据收集策略,利用Hash算法、分段抽样和数学噪声注入等方式,实现尽可 能多地了解特定群体表现出的特质,同时尽量不对群体中的个体进行用户画像。简单来说,就是苹果不 会将数据与个体一一对应,而是将其与其他上亿份信息混合在一起。 比如Apple Intelligence的Genmoji表情符号功能,就允许iPhone用户使用AI来生成定制的、基于文本输入 的表情字符。在用户选择共享设备分析数据时,系统会通过随机噪声算法来收集高频指令,只有某条指 令被数百个设备同时提交后,系统才会识别,并且所有的数据都与设备ID完全脱绑。 与此同时,在训练AI模型需要的语料数据层面,苹果方面也选择了使用合成数据,而非直接采集用户 的数据。苹果在这一文章中承诺,即便用户启用设备分析计划,个人数据也会始终加密存储在本地,其 仅获取经过数学验 ...
新加坡媒体:“我不是机器人”的验证测试曾令人烦躁,如今却变得危险
Huan Qiu Shi Bao· 2025-04-10 22:45
Core Viewpoint - The article discusses the evolution and current issues surrounding CAPTCHA systems, highlighting their ineffectiveness against advanced bots and the potential security risks they pose to users [1][2][3]. Group 1: CAPTCHA's Original Purpose and Evolution - CAPTCHA, which stands for "Completely Automated Public Turing test to tell Computers and Humans Apart," was invented in the late 1990s to prevent bots from abusing website resources [2]. - Initially designed to thwart spam and fake account creation, CAPTCHA has evolved from simple text distortions to more complex image and audio tests due to advancements in bot technology [3]. Group 2: Current Challenges and Risks - Recent studies indicate that advanced bots can nearly achieve 100% accuracy in solving distorted text CAPTCHAs, while human users only manage a success rate of 50% to 84% [3]. - The increasing complexity of CAPTCHAs leads to user frustration, often resulting in users abandoning web pages altogether [3][4]. - Cybersecurity experts warn that CAPTCHA is now being exploited by hackers as a tool for scams, particularly in investment fraud schemes [2]. Group 3: Alternatives and Future Directions - New technologies are being developed to identify human users without explicit CAPTCHA tests, such as monitoring mouse movements and typing patterns, but these raise privacy concerns [3][4]. - Biometric identification and government ID scanning are considered more reliable alternatives, yet their implementation faces challenges and public resistance [4]. - The current landscape suggests that while CAPTCHA is outdated and potentially harmful, effective and secure alternatives have yet to become widely adopted [4].
心智观察所:百度少女“开盒”事件,背后的水比你想的要深
Guan Cha Zhe Wang· 2025-03-24 06:37
心智观察所:百度少女"开盒"事件,背后的水比你 想的要深 【文/观察者网专栏作者 心智观察所】 按照孩子他爸谢广军的说法,他未成年的女儿"情绪失控下,把海外社群网站上的他人隐私信息发布在 自己的账号上,进而导致自己的个人信息也被曝光。"后据调查,开盒信息来自海外的社工库——通过 非法手段收集个人隐私信息的数据库。相关调查过程已取证,并得到公证机关公证。 综合各家媒体渠道,基本确认这个所谓的社工库就是在全球颇有影响力却屡屡陷入舆论漩涡的 Telegram。比开盒事件本身更值得关注的是,这个13岁的未成年少女能玩Telegram,而且知道从上面花 钱买隐私数据,这是一个危险的苗头,值得国内社会各界警惕。 Telegram的"黑魔法" Telegram由俄罗斯企业家帕维尔·杜罗夫(Pavel Durov)于2013年创立。杜罗夫此前曾创建俄罗斯最大 社交平台VKontakte(VK)。Telegram创立之初,就是以加密聊天和隐私保护为卖点,吸引了注重安全 的用户群体。2014年用户量达1500万,2024年突破9.5亿,2025年3月月活跃用户已经飙升超过10亿,成 为全球第二大通讯应用,仅次于WhatsApp, ...