Abstract

Online social networks rely on their valuable data stores to attract users and produce income. Their survival depends on the ability to protect users ’ profiles and disseminate it to other users through controlled channels. Given the sparse user adoption of privacy policies, however, there is increasing incentive and opportunity for malicious parties to extract these datasets for profit using automated “crawlers ” and “screen-scrapers. ” With the arrival of distributed botnets and low-cost hosted VMs, attackers can perform fast, distributed crawls that evade traditional detectors and rate limiters. We propose SpikeStrip, a server add-on that uses light-weight link encryption to isolate and rate limit crawlers. We experiment with real OSN data, and show that SpikeStrip successfully curtails sophisticated, distributed crawlers while imposing minimal server throughput overhead and inconvenience to end-users. 1

Citations