ICACT20260058 Question.1
Questioner: abilaaniloshan@gmail.com 2026-02-09 오후 4:32:18 |
ICACT20260058 Answer.1
Answer by Auhor sudhakar@ccet.ac.in 2026-02-09 오후 4:32:18
Chrome  Click!! |
| How do you ensure that the LLM-generated explanations remain faithful to the underlying Random Forest model outputs and do not introduce hallucinated or misleading interpretations? |
We ensure faithfulness by tightly constraining the LLM’s inputs and outputs to the Random Forest model’s signals rather than allowing free-form reasoning. The LLM receives only the classifier decision, confidence score, top-ranked feature importances, and structured anomaly context, which grounds the explanation in model-derived evidence. Prompting is task-constrained to summarization, root-cause inference, and remediation suggestion based explicitly on these inputs, preventing speculative interpretations. Additionally, low-temperature decoding is used to reduce variability, and the LLM does not generate explanations unless an anomaly is confirmed by the Random Forest, ensuring that all outputs remain consistent with the underlying model behavior. |
ICACT20260058 Question.2
Questioner: abilaaniloshan@gmail.com 2026-02-09 오후 4:32:54 |
ICACT20260058 Answer.2
Answer by Auhor sudhakar@ccet.ac.in 2026-02-09 오후 4:32:54
Chrome Click!! |
| Have you evaluated the latency and resource overhead introduced by the LLM explanation layer, and how feasible is real-time deployment of IoT-XPLAIN on constrained edge devices? |
We evaluated the LLM layer from a deployment feasibility perspective and designed IoT-XPLAIN to keep latency and resource overhead outside the critical detection path. The Random Forest classifier performs real-time anomaly detection locally at the edge with minimal computational cost, while the LLM explanation module is triggered only after an anomaly is detected and can be executed asynchronously on a nearby edge gateway or cloud service. This decoupled design ensures that fault detection remains real-time, while explanation generation introduces acceptable, non-blocking latency. For constrained edge devices, IoT-XPLAIN supports selective explanation generation, lightweight prompting, and offloading of LLM inference, making real-time deployment practical in operational IoT environments. |
ICACT20260058 Question.3
Questioner: arjunvirvashisht@gmail.com 2026-02-10 오전 10:12:23 |
ICACT20260058 Answer.3
Answer by Auhor sudhakar@ccet.ac.in 2026-02-10 오전 10:12:23
Chrome Click!! |
| Did you use a public IoT dataset like BoT-IoT or a custom-collected one? |
We used a public IoT dataset, specifically the TON IoT network-traffic subset, not a custom-collected dataset like BoT-IoT. |
ICACT20260058 Question.4
Questioner: hominpark@etri.re.kr 2026-02-10 오후 6:13:45 |
ICACT20260058 Answer.4
Answer by Auhor sudhakar@ccet.ac.in 2026-02-10 오후 6:13:45
Chrome Click!! |
| You frame this for edge environments, yet the system must transmit sensitive logs to an external API (GPT-4o) for explanations. Doesnt this directly contradict the core edge computing requirements of data security and low latency? |
This concern is valid, and IoT-XPLAIN addresses it through architectural separation and deployment flexibility rather than mandatory cloud dependence. In our design, the Random Forest–based detection runs entirely at the edge, ensuring that sensitive data never leaves the device for real-time fault detection. The LLM explanation layer is optional and can be deployed in three modes: offloaded to a trusted private gateway, executed using an on-premise or self-hosted LLM, or anonymized and minimized before transmission to an external API. Only structured, reduced feature representations and model outputs are shared, not raw traffic logs, which mitigates data exposure. As a result, IoT-XPLAIN preserves edge computing principles for latency and security while enabling explainability as an add-on rather than a hard dependency. |
ICACT20260058 Question.5
Questioner: mrajhans@paloaltonetworks.com 2026-02-11 오후 4:16:03 |
ICACT20260058 Answer.5
Answer by Auhor sudhakar@ccet.ac.in 2026-02-11 오후 4:16:03
Chrome Click!! |
| How does IoT-XPLAIN ensure the reliability and faithfulness of GPT-4o–generated explanations to the underlying Random Forest decisions, especially under noisy inputs or adversarially injected faults, and how is hallucination or explanation drift mitigated? |
IoT-XPLAIN enforces explanation faithfulness through strict architectural grounding and constrained prompting. The LLM does not access raw logs independently; it receives only structured inputs derived from the Random Forest model, including the predicted class, confidence score, top-ranked feature importances, and the normalized feature vector. Prompts explicitly instruct the model to base explanations solely on these provided signals, preventing external inference. Low-temperature decoding reduces variability, and explanations are generated only after a confirmed anomaly classification, avoiding speculative output. Under noisy or adversarially injected faults, robustness is preserved at the detection layer through SMOTE-balanced training and feature selection, while explanation drift is mitigated by limiting the LLM to summarization and structured reasoning rather than open-ended analysis. This tight coupling between model evidence and explanation output ensures consistency and minimizes hallucination risk. |
