Did You Check Stack Overflow: An Empirical Study of What Programmers Ask When Learning a New Language @chrisparnin Nischal Basnet @NCStateUGR #SummerUGR

We won a prize at PackHacks 2019! 🙌 Devpost bit.ly/2IuWxIH

We are proud to present these recipients of the 2021 Marshall Scholarship! 🇬🇧🇺🇸 Our scholars this year will be continuing their education in the 🇬🇧 in programmes ranging from neuroscience to environmental studies! Congratulations to all! #MeetTheMarshalls

Colton Botta '19 is NC State's fourth Marshall Scholars and one of only 46 nationwide this year. 👏🐺 The NC State Engineering alum is interested in researching how to use technology to individualize education so that it resonates with more students. ncst.at/pBaY50COlv3 #ThinkAndDo

It's #Christmas Eve, so let's do this. RT this and on Dec. 26 we'll choose one RT at random. That person and his or her guest will get a round on No. 2 and #TheCradle on us for Christmas. Good luck.

Hearty Congratulations to N. Shrestha, C. Botta, T. Barik, C. Parnin 👏👏Their paper "Here We Go Again: Why Is It Difficult for Developers to Learn Another Programming Language?" (ICSE 2020) is in the list of ACM ACM SIGSOFT Research Highlights. Details at bit.ly/2O98cRH.

.NC State Engineering and NC State Sciences student Ana Sofia Uzsoy becomes NC State's third-ever Churchill Scholar. 👏🐺 With the prestigious award, Uzsoy will study machine learning and machine intelligence. ncst.at/cpMF50EevHu #ThinkAndDo

We know that LSTMs excel at syscall anomaly detection, but they have a notable drawback - doing it fast enough to matter. Most LSTM implementations process an average of ~15K syscalls/sec, but production systems generate 50-70K+/sec during peak loads. This is where Temporal

The biggest misconception in host-based intrusion detection: more data equals better security. Monitoring every syscall creates noise that hides real attacks. Many effective systems monitor just 12-15 critical syscalls and achieve 90%+ detection rates. Less is more when you know

The reason why most syscall-based intrusion detection systems (IDS) fail in production: they treat all system calls equally. But execve during business hours vs 3am tells completely different stories. Time-aware models that learn normal operational patterns dramatically

Most security teams obsess over false positives, but the real killer is alert delay. By the time your batch processing detects privilege escalation, the attacker has already moved laterally. Sub-second detection isn't a nice-to-have - it's the difference between containment and

AI agents are about to break every syscall-based detection system you have. Traditional anomaly detection assumes consistent program behavior, but LLM agents generate unpredictable execution paths based on context and reasoning. That git clone followed by curl and chmod +x could