随着社交网络和移动设备的发展，照片分享已成为一种非常流行的社交活动。由于人们在日常生活中希望展现一种积极的印象，照片分享中的印象管理已变得越来越重要。然而，大多数现有的隐私保护工作都存在两个问题：①为了建立一个合适的印象，用户必须由自己决定是否将照片分享给其他用户；②用户在分享合影照片时，隐含在合影中的人际关系信息也存在泄漏风险，如通过人脸的相对位置可以推断用户的情侣关系或者性倾向。本文提出了一种人际关系印象管理方案（SRIM），通过为用户自动推荐照片分享策略保护用户的关系隐私信息。具体来说，本文设计了一种轻量级的人脸距离测量方法，利用照片的元数据和人脸检测结果计算用户在合影中的实际距离，再通过关系印象评估算法对合影中的人际关系进行评估和管理。基于该方案，本文实现了一个原型系统，并邀请了21 位志愿者对系统的功能进行了验证，评估结果验证了该方案的有效性和效率。

With the development of online social networks (OSNs) and modern smartphones, sharing photos with friends has become one of the most popular social activities. Since people usually prefer to give others a positive impression, impression management during photo sharing is becoming increasingly important. However, most of the existing privacy-aware solutions have two main drawbacks: ① Users must decide manually whether to share each photo with others or not, in order to build the desired impression; and ② users run a high risk of leaking sensitive relational information in group photos during photo sharing, such as their position as part of a couple, or their sexual identity. In this paper, we propose a social relation impression-management (SRIM) scheme to protect relational privacy and to automatically recommend an appropriate photo-sharing policy to users. To be more specific, we have designed a lightweight face-distance measurement that calculates the distances between users' faces within group photos by relying on photo metadata and face-detection results. These distances are then transformed into relations using proxemics. Furthermore, we propose a relation impression evaluation algorithm to evaluate and manage relational impressions. We developed a prototype and employed 21 volunteers to verify the functionalities of the SRIM scheme. The evaluation results show the effectiveness and efficiency of our proposed scheme.