Along with @XiangZhuolun@SuperAluex and @socrates1024, we explore the use of batch amortization to bring the performance of DPSS in the asynchronous setting on par with prior work in the synchronous and partially synchronous settings
#SIGMOD22 paper: “Litmus: Towards A Cryptographically Verifiable Database System” https://t.co/PeEPoZ5omk
I’ll present our attempt on building a secure DBMS that provides verifiable transactional correctness and partial ACID properties.
Video https://t.co/t4ZhXG0W5g
Even when the server gets compromised, the hacker can at most mount a DoS attack, in which case the client would notice, or provide data hosting and transaction computation honestly.
How is this possible?
This JNDI lookup thing existed since the first log4j 2 release in 2014. It was even documented in Mar 2016, in the same year @pwntester 's famous BH talk was published. More than 5 years ago.
I thought it must be not exploitable until very recent, but NO!
Congrats to @rxin and @matei_zaharia on these performance results. It is impressive engineering.
My original quote in the article was supposed to be "Only old people care about official TPC results." The reporter did a good job cleaning it up.
A cute project from William Zhang, one of the high school students I have been mentoring. Since 2020, I have been a mentor in the MIT PRIMES project, which offers guided research projects to top high school students.
This is my other #VLDB2021 paper with @SuperAluex + @xiangyao_yu. It allows DBMS to maintain multiple WAL files to improve perf by tracking txn dependencies in logs. We implemented it in DBx1000 database system. Yu was able to saturate 8 NVMe drives. Faster than WAL w/ RAID
Check out our new work with @superaluex and Zack Newman (MIT) on authenticated dictionaries with cross-incremental proof (dis)aggregation from hidden-order groups, with application to stateless validation and transparency logging: https://t.co/iFxI08F6Zs