分享式Spam攻击的轻量级检测方案

Lightweight detection system of shared spam attacks

Spam攻击是针对社交网络最主要的攻击方式,分享式Spam攻击具有Spam内容的存储与传播分离的新特性,目前没有有效的检测方案。针对这一问题分析了其攻击过程和特征,利用分享式Spam攻击传播和存储的特征设计了轻量级迭代检测算法LIDA,通过目标筛选和内容检测2个步骤实现对分享式Spam的检测。同时,轻量级算法避免了传统算法对每个用户都做深度检测的问题,更具实用性。通过人人网的4次迭代实验,共检测到9 568个Spam账号、30 732个Spam相册以及2 626 780条Spam URL,表明所提的检测算法对于分享式Spam攻击是行之有效的。

Spam is one of the most serious attacks against online social networks（OSN）. Recently a new type of spam attack occurs, which named shared spam attack. The shared spam attack can separate the storage and dissemination of spam content, making the existing detection systems no longer effective. To address this problem, an empirical analysis of the process and properties of this new spam attack is performed. A novel lightweight iterative detection algorithm（LIDA） is proposed to detect spam accounts in OSN with these properties. LIDA contains two steps： target filter and content detection. It also noteworthy that LIDA is a lightweight algorithm to infer more spam accounts by exploiting spam accounts＇ sharing instead of scanning or analyzing all accounts. Experimental results in Ren Ren, which has successfully detected 9 568 spam accounts, 30 732 spam albums and 2 626 780 spam URL in four round iterations, indicate that LIDA is effective and efficiency in detecting shared spam accounts.