arXiv subjects

Sankalp Mittal

Publications and source records attributed to Sankalp Mittal.

3 recordsLinked to original sources

arXiv2026

Text2Arch: A Dataset for Generating Scientific Architecture Diagrams from Natural Language Descriptions

Communicating complex system designs or scientific processes through text alone is inefficient and prone to ambiguity. A system that automatically generates scientific architecture diagrams from text with high semantic fidelity can be useful in multiple applications like enterprise architecture visualization, AI-driven software design, and educational content creation. Hence, in this paper, we focus on leveraging language models to perform semantic understanding of the input text description to generate intermediate code that can be processed to generate high-fidelity architecture diagrams. Unfortunately, no clean large-scale open-access dataset exists, implying lack of any effective open models for this task. Hence, we contribute a comprehensive dataset, \system, comprising scientific architecture images, their corresponding textual descriptions, and associated DOT code representations. Leveraging this resource, we fine-tune a suite of small language models, and also perform in-context learning using GPT-4o. Through extensive experimentation, we show that \system{} models significantly outperform existing baseline models like DiagramAgent and perform at par with in-context learning-based generations from GPT-4o. We make the code, data and models publicly available.

cs.CL↗

arXiv2024

CyberSentinel: Efficient Anomaly Detection in Programmable Switch using Knowledge Distillation

The increasing volume of traffic (especially from IoT devices) is posing a challenge to the current anomaly detection systems. Existing systems are forced to take the support of the control plane for a more thorough and accurate detection of malicious traffic (anomalies). This introduces latency in making decisions regarding fast incoming traffic and therefore, existing systems are unable to scale to such growing rates of traffic. In this paper, we propose CyberSentinel, a high throughput and accurate anomaly detection system deployed entirely in the programmable switch data plane; making it the first work to accurately detect anomalies at line speed. To detect unseen network attacks, CyberSentinel uses a novel knowledge distillation scheme that incorporates "learned" knowledge of deep unsupervised ML models (\textit{e.g.}, autoencoders) to develop an iForest model that is then installed in the data plane in the form of whitelist rules. We implement a prototype of CyberSentinel on a testbed with an Intel Tofino switch and evaluate it on various real-world use cases. CyberSentinel yields similar detection performance compared to the state-of-the-art control plane solutions but with an increase in packet-processing throughput by $66.47\%$ on a $40$ Gbps link, and a reduction in average per-packet latency by $50\%$.

cs.CR↗

arXiv2023

Rethinking Robustness of Model Attributions

For machine learning models to be reliable and trustworthy, their decisions must be interpretable. As these models find increasing use in safety-critical applications, it is important that not just the model predictions but also their explanations (as feature attributions) be robust to small human-imperceptible input perturbations. Recent works have shown that many attribution methods are fragile and have proposed improvements in either these methods or the model training. We observe two main causes for fragile attributions: first, the existing metrics of robustness (e.g., top-k intersection) over-penalize even reasonable local shifts in attribution, thereby making random perturbations to appear as a strong attack, and second, the attribution can be concentrated in a small region even when there are multiple important parts in an image. To rectify this, we propose simple ways to strengthen existing metrics and attribution methods that incorporate locality of pixels in robustness metrics and diversity of pixel locations in attributions. Towards the role of model training in attributional robustness, we empirically observe that adversarially trained models have more robust attributions on smaller datasets, however, this advantage disappears in larger datasets. Code is available at https://github.com/ksandeshk/LENS.

cs.LG↗