Wei Song

Wei Song

Auto-generated profile, based on publicly available information

Wei Song

Institute of Information Engineering, CAS

Research Focus & Metrics

Academic Career Length

18Years (2009-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
18 August 2016
Verified
Conference Paper
Closed access
WSWei Song
Wei Song
DKDirk Koch
Dirk Koch
MLMikel Lujan
Mikel Lujan
JGJim Garside
Jim Garside
Sorting has tremendous usage in the applications that handle massive amount of data. Existing techniques accelerate sorting using multiprocessors or GPGPUs where a data set is partitioned into disjunctive subsets to allow multiple sorting threads working in parallel. Hardware sorters implemented in FPGAs have the potential of providing high-speed and low-energy solutions but the partition algorithms used in software systems are so data dependent that they cannot be easily adopted. The speed of most current sequential sorters still hangs around 1 number/cycle. Recently a new hardware merge sorter broke this speed limit by merging a large number of sorted sequences at a speed proportional to the number of sequences. This paper significantly improves its area and speed scalability by allowing stalls and variable sorting rate. A 32-port parallel merge-tree that merges 32 sequences is implemented in a Virtex-7 FPGA. It merges sequences at an average rate of 31.05 number/cycle and reduces the total sorting time by 160 times compared with traditional sequential sorters.

1 September 2014
Verified
Conference Paper
Closed access
OAOriol Arcas-Abella
Oriol Arcas-Abella
GNGeoffrey Ndu
Geoffrey Ndu
NSNehir Sonmez
Nehir Sonmez
MGMohsen Ghasempour
Mohsen Ghasempour
AAAdria Armejach
Adria Armejach
MLMikel Lujan
Mikel Lujan
High Level Synthesis (HLS) languages and tools are emerging as the most promising technique to make FPGAs more accessible to software developers. Nevertheless, picking the most suitable HLS for a certain class of algorithms depends on requirements such as area and throughput, as well as on programmer experience. In this paper, we explore the different trade-offs present when using a representative set of HLS tools in the context of Database Management Systems (DBMS) acceleration. More specifically, we conduct an empirical analysis of four representative frameworks (Bluespec SystemVerilog, Altera OpenCL, LegUp and Chisel) that we utilize to accelerate commonly-used database algorithms such as sorting, the median operator, and hash joins. Through our implementation experience and empirical results for database acceleration, we conclude that the selection of the most suitable HLS depends on a set of orthogonal characteristics, which we highlight for each HLS framework.

17 October 2025 in IEEE Transactions on Information Forensics and Security
Verified
Journal Article
Closed access
LSLinke Song
Linke Song
ZPZixuan Pang
Zixuan Pang
WWWenhao Wang
Wenhao Wang
ZWZihao Wang
Zihao Wang
XWXiaoFeng Wang
XiaoFeng Wang
RHRui Hou
Rui Hou
The wide deployment of Large Language Models (LLMs) has given rise to strong demands for optimizing their inference performance. Today’s techniques serving this purpose primarily focus on reducing latency and improving throughput through algorithmic and hardware enhancements, while largely overlooking their privacy side effects, particularly in a multi-user environment. In our research, for the first time, we discovered a set of new timing side channels in LLM systems, arising from shared caches and GPU memory allocations, which can be exploited to infer both confidential system prompts and those issued by other users. These vulnerabilities echo security challenges observed in traditional computing systems, highlighting an urgent need to address potential information leakage in LLM serving infrastructures. In this paper, we report novel attack strategies designed to exploit such timing side channels inherent in LLM deployments, specifically targeting the Key-Value (KV) cache and semantic cache widely used to enhance LLM inference performance. Our approach leverages timing measurements and classification models to detect cache hits, allowing an adversary to infer private prompts with high accuracy. We also propose a token-by-token search algorithm to efficiently recover shared prompt prefixes in the caches, showing the feasibility of stealing system prompts and those produced by peer users. Our experimental studies on black-box testing of popular online LLM services demonstrate that such privacy risks are completely realistic, with significant consequences. Our findings underscore the need for robust mitigation to protect LLM systems against such emerging threats.

10 January 2023 in IEEE Transactions on Computers
Verified
Journal Article
Closed access
WHWei He
Wei He
ZZZhi Zhang
Zhi Zhang
YCYueqiang Cheng
Yueqiang Cheng
WWWenhao Wang
Wenhao Wang
WSWei Song
Wei Song
SNSurya Nepal
Surya Nepal
With frequent software-induced activations on DRAM rows, bit flips can occur on their physically adjacent rows (i.e., RowHammer). Existing studies leverage FPGA platforms to characterize RowHammer, which have identified key factors that contribute to RowHammer bit flips, e.g., data pattern. As the FPGA-based studies have removed the interference of the OS and the memory controller, their findings on the identified contributing factors do not always work as reported in a real-world computing system, resulting in negative effects on system-level RowHammer attacks and defenses. In this paper, we carry out a system-level empirical study on factors from both the software side and the DRAM side that contribute to RowHammer. We conduct the study on 33 DRAM modules including both DDR4 and DDR3, with 292 DRAM chips from various vendors. Our experimental results from the software side show that some prior findings about existing factors are inconsistent with our observations, thus not applicable to a real-world system. Also, we contribute to identifying one new factor that effectively affects RowHammer bit flips. Our DRAM-side results identify three types of new contributing factors and indicate that DRAM modules are more vulnerable if they achieve better performance and lower power consumption. Particularly, Intel XMP, intended for improving DRAM performance, might be abused for RowHammer attacks.

1 January 2009
Verified
Conference Paper
Closed access
WSWei Song
Wei Song
DEDoug Edwards
Doug Edwards
JNJose Luis Nunez-Yanez
Jose Luis Nunez-Yanez
SDSohini Dasgupta
Sohini Dasgupta
Due to shrinking transistor geometries, on-chip circuits are becoming vulnerable to errors, but at the same time on-chip networks are required to provide reliable services over unreliable physical interconnects. A connection oriented stochastic routing (COSR) algorithm has been used on one NoC platform that provides excellent fault-tolerance and dynamic reconfiguration capability. A probability model has been built to analyze the COSR algorithm. According to the model, the performance may be improved by implementing a self learning mechanism in each router. Thus a new adaptive stochastic routing (ASR) algorithm is proposed whereby each router learns the network status from acknowledgement flits and stores the outcomes in a routing table. Simulation of both algorithms reveals that the ASR algorithm shows a higher path reservation success rate and a larger maximal accepted traffic than the COSR algorithm. The simulations also show that the learning procedures are accurate and that both algorithms are fault-tolerant to intermittent/permanent errors.

16 October 2023
Verified
Conference Paper
Open access
ZXZihan Xue
Zihan Xue
JHJinchi Han
Jinchi Han
WSWei Song
Wei Song
Eviction sets are essential components of the conflict-based cache side-channel attacks. However, it is not an easy task to construct eviction sets on modern Intel processors. As a promising defense against conflict-based cache side-channels, dynamic cache randomization makes the construction of eviction sets even more difficult by periodically randomizing the mapping between addresses and cache set indices. It forces attackers to develop fast search algorithms to find an eviction set at runtime with the lowest latency. Several fast search algorithms have been proposed in recent years. By using these algorithms, attackers regain the capability of launching useful attacks on dynamically randomized caches. Consequently, a detector was recently introduced to catch the fast search algorithms in action according to the uneven distribution of cache evictions. All existing fast search algorithms fail to work. We present a new eviction set search algorithm called Conflict Testing with Probe+Prune (CTPP). Based on the evaluation on six Intel processors and a behavioral cache model, CTPP is found to achieve the lowest latency in finding an eviction set in all algorithms, potentially escape from the recently proposed detector, and present a strong tolerance to environmental noise.