Research Article

Breaking the Black Box: Securing and Auditing Edge-Deployed LLMs via Shard Traceability

by  Gururaj Shinde, Ritu Kuklani, Varad Vishwarupe
journal cover
International Journal of Computer Applications
Foundation of Computer Science (FCS), NY, USA
Volume 187 - Issue 27
Published: August 2025
Authors: Gururaj Shinde, Ritu Kuklani, Varad Vishwarupe
10.5120/ijca2025925483
PDF

Gururaj Shinde, Ritu Kuklani, Varad Vishwarupe . Breaking the Black Box: Securing and Auditing Edge-Deployed LLMs via Shard Traceability. International Journal of Computer Applications. 187, 27 (August 2025), 44-49. DOI=10.5120/ijca2025925483

                        @article{ 10.5120/ijca2025925483,
                        author  = { Gururaj Shinde,Ritu Kuklani,Varad Vishwarupe },
                        title   = { Breaking the Black Box: Securing and Auditing Edge-Deployed LLMs via Shard Traceability },
                        journal = { International Journal of Computer Applications },
                        year    = { 2025 },
                        volume  = { 187 },
                        number  = { 27 },
                        pages   = { 44-49 },
                        doi     = { 10.5120/ijca2025925483 },
                        publisher = { Foundation of Computer Science (FCS), NY, USA }
                        }
                        %0 Journal Article
                        %D 2025
                        %A Gururaj Shinde
                        %A Ritu Kuklani
                        %A Varad Vishwarupe
                        %T Breaking the Black Box: Securing and Auditing Edge-Deployed LLMs via Shard Traceability%T 
                        %J International Journal of Computer Applications
                        %V 187
                        %N 27
                        %P 44-49
                        %R 10.5120/ijca2025925483
                        %I Foundation of Computer Science (FCS), NY, USA
Abstract

EdgeShard represents a significant advancement in edge-based large language model (LLM) inference, enabling efficient, accurate, and privacy-preserving deployment by intelligently partitioning and scheduling computation across multiple edge devices. This collaborative approach outperforms traditional quantization and unstable cloud-edge methods. However, distributing inference across heterogeneous and potentially unreliable devices introduces new risks for robustness - such as increased vulnerability to device failures and attacks, and challenges for auditability, including fragmented execution logs and difficulties in tracing and verifying the end-to-end inference process.

References
  • Ouyang, L., Wu, J., Jiang, X., et al. (2022). Training language models to follow instructions with human feedback. arXiv preprint arXiv:2203.02155.
  • Vishwarupe, V., Zahoor, S., Akhter, R., Bhatkar, V. P., Bedekar, M., Pande, M., Joshi, P. M., Patil, A., & Pawar, V. (2023). Designing a human-centered AI-based cognitive learning model for Industry 4.0 applications. In Industry 4.0 Convergence with AI, IoT, Big Data and Cloud Computing: Fundamentals, Challenges and Applications (pp. 84–95). Bentham Science Publishers.
  • Dwork, C., & Roth, A. (2014). The algorithmic foundations of differential privacy. Foundations and Trends in Theoretical Computer Science, 9(3–4), 211–407.
  • Sayyed, H., Alwazae, M., & Vishwarupe, V. (2025). BlockSafe: Universal blockchain-based identity management. In B. Alareeni (Ed.), Big Data in Finance: Transforming the Financial Landscape (Vol. 169, pp. 101–118). Springer. https://doi.org/10.1007/978-3-031-80656-8_6
  • Vishwarupe, V., Maheshwari, S., Deshmukh, A., Mhaisalkar, S., Joshi, P. M., & Mathias, N. (2022). Bringing humans at the epicentre of artificial intelligence: A confluence of AI, HCI, and human-centered computing. Procedia Computer Science, 204, 914–921. https://doi.org/10.1016/j.procs.2022.08.111
  • Rayson Laroca, R., Severo, E., Zanlorensi, L., Oliveira, L., Gonçalves, G., Schwartz, W., & Menotti, D. (2018). A robust real-time automatic license plate recognition based on the YOLO detector. arXiv preprint arXiv:1802.09567.
  • Vishwarupe, V., Bedekar, M., Pande, M., & Hiwale, A. (2018). Intelligent Twitter spam detection: A hybrid approach. In X. S. Yang, A. Nagar, & A. Joshi (Eds.), Smart trends in systems, security and sustainability (Vol. 18, pp. 157–167). Springer. https://doi.org/10.1007/978-981-10-6916-1_17
  • T. Li, Sahu, A. K., Talwalkar, A., & Smith, V. (2020). Federated learning: Challenges, methods, and future directions. IEEE Signal Processing Magazine, 37(3), 50–60.
  • Vishwarupe, V., Joshi, P. M., Mathias, N., Maheshwari, S., Mhaisalkar, S., & Pawar, V. (2022). Explainable AI and interpretable machine learning: A case study in perspective. Procedia Computer Science, 204, 869–876. https://doi.org/10.1016/j.procs.2022.08.105
  • The Syslog Protocol. (2001/2009). RFC 3164/5424, Internet Engineering Task Force (IETF).
  • Wani, K., Khedekar, N., Vishwarupe, V., & Pushyanth, N. (2023). Digital twin and its applications. In Research Trends in Artificial Intelligence: Internet of Things (pp. 120–134). Bentham Science Publishers.
  • Xie, C., Koyejo, O., & Gupta, I. (2020). Fall of empires: Breaking Byzantine-tolerant SGD by inner product manipulation. In Proceedings of the International Conference on Machine Learning (ICML).
  • Vidgen, B., Harris, A., & Emmery, C. (2021). Challenges and frontiers in abusive content detection. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing.
  • Vishwarupe, V., Bedekar, M., Pande, M., Bhatkar, V. P., Joshi, P., Zahoor, S., & Kuklani, P. (2022). Comparative analysis of machine learning algorithms for analyzing NASA Kepler mission data. Procedia Computer Science, 204, 945–951. https://doi.org/10.1016/j.procs.2022.08.115
  • Blanchard, P., El Mhamdi, E. M., Guerraoui, R., & Stainer, J. (2017). Machine learning with adversaries: Byzantine tolerant gradient descent. In Advances in Neural Information Processing Systems (NeurIPS).
  • Vishwarupe, V. (2022, February 10). Synthetic content generation using artificial intelligence. All Things Policy. IVM Podcasts. https://shows.ivmpodcasts.com/show/all-things-policy-Rx64RVpQImivrNQ8/episode/synthetic-content-generation-and-chinas-worries-ja9s-I7rfgZE3IhXRg2Fk
  • Kairouz, P., McMahan, H. B., et al. (2021). Advances and open problems in federated learning. Foundations and Trends in Machine Learning, 14(1–2), 1–210.
  • Sable, N. P., Rathod, V. U., Mahalle, P. N., & Birari, D. R. (2022, March). A multiple stage deep learning model for NID in MANETs. In 2022 International Conference on Emerging Smart Computing and Informatics (ESCI) (pp. 1–6). IEEE.
  • Common Event Format (CEF); JSON Logging Standards. ArcSight.
  • Shi, W., Cao, J., Zhang, Q., Li, Y., & Xu, L. (2016). Edge computing: Vision and challenges. IEEE Internet of Things Journal, 3(5), 637–646.
  • Zahoor, S., Bedekar, M., Mane, V., & Vishwarupe, V. (2016). Uniqueness in user behavior while using the web. In S. Satapathy, Y. Bhatt, A. Joshi, & D. Mishra (Eds.), Proceedings of the International Congress on Information and Communication Technology (Vol. 438, pp. 229–236). Springer. https://doi.org/10.1007/978-981-10-0767-5_24
  • Vishwarupe, V., Bedekar, M., & Zahoor, S. (2015). Zone-specific weather monitoring system using crowdsourcing and telecom infrastructure. In 2015 International Conference on Information Processing (ICIP) (pp. 823–827). IEEE. https://doi.org/10.1109/INFOP.2015.7489495
  • Zahoor, S., Bedekar, M., & Vishwarupe, V. (2016). A framework to infer webpage relevancy for a user. In S. Satapathy & S. Das (Eds.), Proceedings of First International Conference on Information and Communication Technology for Intelligent Systems: Volume 1 (Vol. 50, pp. 173–181). Springer. https://doi.org/10.1007/978-3-319-30933-0_16
  • Gehman, S., Gururangan, S., Sap, M., et al. (2020). RealToxicityPrompts: Evaluating neural toxic degeneration in language models. arXiv preprint arXiv:2009.11462.
  • Zhang, M., Cao, J., Shen, X., & Cui, Z. (2024). EdgeShard: Efficient LLM inference via collaborative edge computing. arXiv preprint arXiv:2405.14371.
  • Deoskar, V., Pande, M., & Vishwarupe, V. (2024). An analytical study for implementing 360-degree M-HRM practices using AI. In Intelligent Systems for Smart Cities: Select Proceedings of the 2nd International Conference, ICISA 2023 (pp. 429–442). Springer Nature.
  • Kuznetsova, A., Rom, H., Alldrin, N., Uijlings, J., Krasin, I., Pont-Tuset, J., Kamali, S., Popov, S., Malloci, M., Kolesnikov, A., Duerig, T., & Ferrari, V. (2020). The Open Images Dataset V4: Unified image classification, object detection, and visual relationship detection at scale. arXiv preprint arXiv:1811.00982.
  • Vishwarupe, V., et al. (2021). A zone-specific weather monitoring system. Australian Patent No. AU2021106275. Australian Government, IP Australia. https://patents.google.com/?inventor=Varad+Vishwarupe
  • Vishwarupe, V., Bedekar, M., Joshi, P. M., Pande, M., Pawar, V., & Shingote, P. (2022). Data analytics in the game of cricket: A novel paradigm. Procedia Computer Science, 204, 937–944. https://doi.org/10.1016/j.procs.2022.08.114
  • Vishwarupe, V. V., & Joshi, P. M. (2016). Intellert: A novel approach for content-priority based message filtering. In 2016 IEEE Bombay Section Symposium (IBSS) (pp. 1–6). IEEE. https://doi.org/10.1109/IBSS.2016.7940206
  • Vishwarupe, V., et al. (2025). Predicting mental health ailments using social media activities and keystroke dynamics with machine learning. In B. Alareeni (Ed.), Big Data in Finance: Transforming the Financial Landscape (Vol. 169, pp. 63–80). Springer. https://doi.org/10.1007/978-3-031-80656-8_4
  • Zahoor, S., Akhter, R., Vishwarupe, V., Bedekar, M., Pande, M., Bhatkar, V. P., Joshi, P. M., Pawar, V., Mandora, N., & Kuklani, P. (2023). A comprehensive study of state-of-the-art applications and challenges in IoT and blockchain technologies for Industry 4.0. In Industry 4.0 Convergence with AI, IoT, Big Data and Cloud Computing: Fundamentals, Challenges and Applications (pp. 1–16). Bentham Science Publishers.
Index Terms
Computer Science
Information Sciences
No index terms available.
Keywords

Large Language Models Edge AI RLHF LLMs Distributed AI Black Box Models Shard AI ML Human-Centered AI

Powered by PhDFocusTM