DRAM chips are vulnerable to read disturbance phenomena (e.g., RowHammer and RowPress), where repeatedly accessing or keeping open a DRAM row causes bitflips in nearby rows, due to DRAM density scaling. Attackers can leverage RowHammer bitflips in real systems to take over systems and leak data. Consequently, many prior works propose mitigations, including recent DDR specifications introducing new mitigation frameworks (e.g., PRAC and RFM). For robustness, it is timely and critical to analyze other security implications that widely-adopted RowHammer mitigations can introduce. Unfortunately, no prior work analyzes the timing channel vulnerabilities introduced by RowHammer mitigations. In this work, we present the first analysis and evaluation of timing channel vulnerabilities introduced by RowHammer mitigations. Our key observation is that RowHammer mitigations' preventive actions have two features that enable timing channels. First, preventive actions often reduce DRAM bandwidth availability because they block access to DRAM, thereby delaying regular memory requests and resulting in increased memory latencies. Second, preventive actions can be triggered on demand as they depend on memory access patterns. We systematically analyze two latest industry mitigations and introduce LeakyHammer, a new class of attacks that leverage the RowHammer mitigation-induced memory latency differences to establish communication channels between processes and leak secrets. First, we build two covert channel attacks exploiting two state-of-the-art RowHammer mitigations, providing 41.9 Kbps and 54.0 Kbps channel capacity. Second, we demonstrate a proof-of-concept website fingerprinting attack that can identify visited websites based on the RowHammer mitigation behavior. We discuss 3 mitigations against LeakyHammer and show that fundamentally mitigating LeakyHammer induces significant performance overheads.

翻译：暂无翻译