Description
The rise of Intent Based Networking (IBN) has paved the way for more efficient network and security management, reduced errors, and accelerated deployment times by leveraging AI processes capable of translating natural language intents into policies or configurations. Specialized neural networks could offer a promising solution at the core of translation operations. Still, they require dedicated, large-scale corpora for training generative models, which are not currently available due to the restricted and confidential nature of the information they convey. This talk fills this gap by proposing a novel methodology for creating a corpus specifically tailored to train a Large Language Model (LLM) for this translation task. Our approach leverages advancements in Natural Language Processing (NLP) to overcome the challenges posed by the limited training data, enabling accurate interpretation and translation of high-level performance and security directives from natural language into structured, actionable formats, represented by the intermediate Network Intent LanguagE (Nile). Our experimental evaluations, grounded by human experts, indicate that our corpus generation methodology yields promising results in terms of translation accuracy, language naturalness, and reliability. To further ensure the correctness of the produced translations, our evaluation framework compares and classifies LLM outputs, assessing their semantic fidelity. This process minimizes the risk of incorrect or unsafe intent translations by filtering out translations that fail to meet stringent accuracy criteria. As a result, we have been able to generate Nile-English Aligned Translations (NEAT), a corpus which is to date two orders of magnitude larger than currently available datasets, exhibits a wide coverage of the Nile syntax, and has much lower perplexity values than other generated corpora. NEAT has been made publicly available to facilitate further research and development in adapting generative models to network management and security.
Infos pratiques
Prochains exposés
-
Towards privacy-preserving and fairness-aware federated learning framework
Orateur : Nesrine Kaaniche - Télécom SudParis
Federated Learning (FL) enables the distributed training of a model across multiple data owners under the orchestration of a central server responsible for aggregating the models generated by the different clients. However, the original approach of FL has significant shortcomings related to privacy and fairness requirements. Specifically, the observation of the model updates may lead to privacy[…]-
Cryptography
-
SoSysec
-
Privacy
-
Machine learning
-
-
Black-Box Collision Attacks on Widely Deployed Perceptual Hash Functions and Their Consequences
Orateur : Diane Leblanc-Albarel - KU Leuven
Perceptual hash functions identify multimedia content by mapping similar inputs to similar outputs. They are widely used for detecting copyright violations and illegal content but lack transparency, as their design details are typically kept secret. Governments are considering extending the application of these functions to Client-Side Scanning (CSS) for end-to-end encrypted services: multimedia[…]-
Cryptography
-
SoSysec
-
-
Malware Detection with AI Systems: bridging the gap between industry and academia
Orateur : Luca Demetrio - University of Genova
With the abundance of programs developed everyday, it is possible to develop next-generation antivirus programs that leverage this vast accumulated knowledge. In practice, these technologies are developed with a mixture of established techniques like pattern matching, and machine learning algorithms, both tailored to achieve high detection rate and low false alarms. While companies state the[…]-
SoSysec
-
Intrusion detection
-
Machine learning
-