Nancy Agarwal

Aakshita Chandiramani, Aaron Blakeman, Abdullahi Olaoye et al. · amazon-science, cmu

We describe the pre-training, post-training, and quantization of Nemotron 3 Super, a 120 billion (active 12 billion) parameter hybrid Mamba-Attention Mixture-of-Experts model. Nemotron 3 Super is the first model in the Nemotron 3 family to 1) be pre-trained in NVFP4, 2) leverage LatentMoE, a new Mixture-of-Experts architecture that optimizes for both accuracy per FLOP and accuracy per parameter, and 3) include MTP layers for inference acceleration through native speculative decoding. We pre-trained Nemotron 3 Super on 25 trillion tokens followed by post-training using supervised fine tuning (SFT) and reinforcement learning (RL). The final model supports up to 1M context length and achieves comparable accuracy on common benchmarks, while also achieving up to 2.2x and 7.5x higher inference throughput compared to GPT-OSS-120B and Qwen3.5-122B, respectively. Nemotron 3 Super datasets, along with the base, post-trained, and quantized checkpoints, are open-sourced on HuggingFace.

Nancy Agarwal, Syed Zeeshan Hussain

Signature-based Intrusion Detection System (SIDS) provides a promising solution to the problem of web application security. However, the performance of the system highly relies on the quality of the signatures designed to detect attacks. A weak signature set may considerably cause an increase in false alarm rate, making impractical to deploy the system. The objective of the paper is to identify the flaws in the signature structure which are responsible to reduce the efficiency of the detection system. The paper targets SQL injection signatures particularly. Initially, some essential concepts of the domain of the attack that should be focused by the developer in prior to designing the signatures have been discussed. Afterwards, we conducted a case study on the well known PHPIDS tool for analyzing the quality of its SQL signatures. Based on the analysis, we identify various flaws in the designing practice that yield inefficient signatures. We divide the weak signatures into six categories, namely incomplete, irrelevant, semi-relevant, susceptible, redundant and inconsistent signatures. Moreover, we quantify these weaknesses and define them mathematically in terms of set theory. To the best of our knowledge, we have identified some novel signature design issues. The paper will basically assist the signature developer to know what level of expertise is required for devising a quality signature set and how a little ignorance may lead to deterioration in the performance of the SIDS. Furthermore, a security expert may evaluate the detector against the identified flaws by conducting structural analysis on its signature set.

Nancy Agarwal, Syed Zeeshan Hussain

Intrusion Detection System (IDS) is one of the security measures being used as an additional defence mechanism to prevent the security breaches on web. It has been well known methodology for detecting network-based attacks but still immature in the domain of securing web application. The objective of the paper is to thoroughly understand the design methodology of the detection system in respect to web applications. In this paper, we discuss several specific aspects of a web application in detail that makes challenging for a developer to build an efficient web IDS. The paper also provides a comprehensive overview of the existing detection systems exclusively designed to observe web traffic. Furthermore, we identify various dimensions for comparing the IDS from different perspectives based on their design and functionalities. We also provide a conceptual framework of an IDS with prevention mechanism to offer a systematic guidance for the implementation of the system specific to the web applications. We compare its features with five existing detection systems, namely AppSensor, PHPIDS, ModSecurity, Shadow Daemon and AQTRONIX WebKnight. The paper will highly facilitate the interest groups with the cutting edge information to understand the stronger and weaker sections of the web IDS and provide a firm foundation for developing an intelligent and efficient system.

Mudasir Ahmad Wani, Nancy Agarwal, Suraiya Jabin et al.

Obtaining the desired dataset is still a prime challenge faced by researchers while analyzing Online Social Network (OSN) sites. Application Programming Interfaces (APIs) provided by OSN service providers for retrieving data impose several unavoidable restrictions which make it difficult to get a desirable dataset. In this paper, we present an iMacros technology-based data crawler called IMcrawler, capable of collecting every piece of information which is accessible through a browser from the Facebook website within the legal framework which permits access to publicly shared user content on OSNs. The proposed crawler addresses most of the challenges allied with web data extraction approaches and most of the APIs provided by OSN service providers. Two broad sections have been extracted from Facebook user profiles, namely, Personal Information and Wall Activities. The present work is the first attempt towards providing the detailed description of crawler design for the Facebook website.