Mitigation of Rowhammer Attack on DDR4 Memory: A Novel Multi-Table Frequent Element Algorithm Based Approach

As Rowhammer attacks have become more and more of a threat to modern computer systems using DRAM as main memory, defenses against it have to improve to match. Rowhammer attacks cause unidentifiable bit flips within DRAM units by using flaws in DRAM's electrical characteristics, and the threat of Rowhammer increases as DRAM continues to scale down to have higher density. This has led to many methods of defense to be developed using many means to defend DRAM including Graphene [1], a space efficient, low cost algorithm to that identifies frequent elements from incoming data streams to prevent Rowhammer bit flips. We propose an improved version of the Graphene defense method, Multi-Table Graphene (MT-Graphene). MT-Graphene uses the same data stream frequent element identification that Graphene uses, but runs multiple of the algorithm throughout smaller sections of the DRAM unit, seeing a improvement of 74% less Rowhammer bit flips when compared to Graphene using two tables, and further increases in security when more tables are added.