Research

Publications

Papers in computer architecture, hardware security, and AI.

Download all paper citations (.bib)

2026

arXiv preprint

CacheFit: The Rising Cost of Cache Replacement and Constraint-Driven Redesign of Last-Level Cache Replacement Policies

Bita Aslrousta, Kaushal Mhapsekar, Darsh Asher, Azam Ghanbari, Joshua Kalyanapu, and Samira Mirbagher Ajorpaz

Redesigning last-level cache replacement policies under latency, area, and power budgets using LLM proposals, ChampSim simulation, and HARCOM analytical cost modeling.

Public preprint · October 2026.

BibTeX reference
Download .bib
@misc{cachefit2026,
  author = {Bita Aslrousta and Kaushal Mhapsekar and Darsh Asher and Azam Ghanbari and Joshua Kalyanapu and Samira Mirbagher Ajorpaz},
  title = {{CacheFit: The Rising Cost of Cache Replacement and Constraint-Driven Redesign of Last-Level Cache Replacement Policies}},
  year = {2026},
  eprint = {2610.10579},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.10579},
  note = {Preprint}
}
arXiv preprint

CACHEFORGE: LLM-Guided End-to-End Generative Cache Replacement Policy for Performance and Hardware Efficiency

Kaushal Mhapsekar, Bita Aslrousta, Brijesh Kumar Bhayana, Paula Contreras, Azam Ghanbari, Ethan Goodman, Anna Andriiko, and Samira Mirbagher Ajorpaz

An LLM-guided generation and evaluation loop that develops C++ CPU cache replacement policies for performance and hardware efficiency.

Public preprint · October 2026.

BibTeX reference
Download .bib
@misc{cacheforge2026,
  author = {Kaushal Mhapsekar and Bita Aslrousta and Brijesh Kumar Bhayana and Paula Contreras and Azam Ghanbari and Ethan Goodman and Anna Andriiko and Samira Mirbagher Ajorpaz},
  title = {{CACHEFORGE: LLM-Guided End-to-End Generative Cache Replacement Policy for Performance and Hardware Efficiency}},
  year = {2026},
  eprint = {2610.07668},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.07668},
  note = {Preprint}
}
arXiv preprint

TranScope: What the Software Hides About LLM Training Data, the Hardware Reveals at Scale, and Accelerators Magnify

Joshua Kalyanapu, Darsh Asher, Kaushal Mhapsekar, Bita Aslrousta, Rushiraj Chaitanyakumar Sheth, Maharshi Mukeshkumar Oza, Achyuta Kannan, and Samira Mirbagher Ajorpaz

Studying how tokenizer and embedding locality produce cache, TLB, and hardware-counter footprints that reveal training-data distribution in locally executed models.

Public preprint · October 2026.

BibTeX reference
Download .bib
@misc{transcope2026,
  author = {Joshua Kalyanapu and Darsh Asher and Kaushal Mhapsekar and Bita Aslrousta and Rushiraj Chaitanyakumar Sheth and Maharshi Mukeshkumar Oza and Achyuta Kannan and Samira Mirbagher Ajorpaz},
  title = {{TranScope: What the Software Hides About LLM Training Data, the Hardware Reveals at Scale, and Accelerators Magnify}},
  year = {2026},
  eprint = {2610.06848},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.06848},
  note = {Preprint}
}
ASPLOS

CacheMind: From Miss Rates to Why—Natural-Language, Trace-Grounded Reasoning for Cache Replacement

Kaushal Mhapsekar, Azam Ghanbari, Bita Aslrousta, and Samira Mirbagher Ajorpaz

Interactive reasoning over cache traces, with a benchmark for evaluating explanations of microarchitectural behavior.

BibTeX reference
Download .bib
@inproceedings{cachemind2026,
  author = {Mhapsekar, Kaushal and Ghanbari, Azam and Aslrousta, Bita and Mirbagher-Ajorpaz, Samira},
  title = {{CacheMind: From Miss Rates to Why - Natural-Language, Trace-Grounded Reasoning for Cache Replacement}},
  booktitle = {Proceedings of the 31st ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2},
  year = {2026},
  pages = {307--322},
  doi = {10.1145/3779212.3790136},
  url = {https://doi.org/10.1145/3779212.3790136}
}
IEEE CAL

FeatureBleed: Inferring Private Enriched Attributes From Sparsity-Optimized AI Accelerators

Darsh Asher, Farshad Dizani, Joshua Kalyanapu, Rosario Cammarota, Aydin Aysu, and Samira Mirbagher Ajorpaz

How sparsity optimizations can expose private features through the timing of AI inference.

BibTeX reference
Download .bib
@article{featurebleed2026,
  author = {Asher, Darsh and Dizani, Farshad and Kalyanapu, Joshua and Cammarota, Rosario and Aysu, Aydin and Ajorpaz, Samira Mirbagher},
  title = {{FeatureBleed: Inferring Private Enriched Attributes From Sparsity-Optimized AI Accelerators}},
  journal = {IEEE Computer Architecture Letters},
  year = {2026},
  volume = {25},
  number = {1},
  pages = {77--80},
  doi = {10.1109/LCA.2026.3666823},
  url = {https://doi.org/10.1109/LCA.2026.3666823}
}
IEEE Micro · Top Picks · Early Access

GateBleed: Exploiting On-Core Accelerator Power Gating for High Performance and Stealthy Attacks on AI

Joshua Kalyanapu, Darsh Asher, Farshad Dizani, Azam Ghanbari, Rosario Cammarota, Aydin Aysu, and Samira Mirbagher Ajorpaz

IEEE Micro Top Picks recognizes research for its novelty and potential for lasting impact on computer architecture. GateBleed is one of 12 selections for 2026 from work published across 18 eligible conferences in 2025. The PDF bundle includes the Top Picks article, original MICRO paper, and impact summary.

Available in IEEE Micro Early Access. IEEE Micro Top Picks recognizes research for its novelty and potential for lasting impact on computer architecture. GateBleed is one of 12 selections for 2026 from work published across 18 eligible conferences in 2025. Bundle: IEEE Micro Top Picks article (13 pages), published MICRO 2025 paper (18 pages), and impact summary (3 pages). The individual Top Picks PDF is the author manuscript.

BibTeX reference
Download .bib
@article{gatebleedtoppicks2026,
  author = {Kalyanapu, Joshua and Asher, Darsh and Dizani, Farshad and Ghanbari, Azam and Aysu, Aydin and Cammarota, Rosario and Mirbagher Ajorpaz, Samira},
  title = {{GateBleed: Exploiting On-Core Accelerator Power Gating for High Performance and Stealthy Attacks on AI}},
  journal = {IEEE Micro},
  year = {2026},
  pages = {1--12},
  doi = {10.1109/MM.2026.3702482},
  url = {https://doi.org/10.1109/MM.2026.3702482}
}

2025

MICRO

GateBleed: Exploiting On-Core Accelerator Power Gating for High Performance and Stealthy Attacks on AI

Joshua Kalyanapu, Farshad Dizani, Darsh Asher, Azam Ghanbari, Rosario Cammarota, Aydin Aysu, and Samira Mirbagher Ajorpaz

A study of the security and privacy consequences of power gating in Intel AMX, and the tradeoffs involved in mitigation.

Three artifact-evaluation badges · MICRO 2025
GateBleed earned Artifacts Available; Artifacts Evaluated — Functional; Results Reproduced.

ACM Artifacts Available badgeACM Artifacts Evaluated — Functional badgeACM Results Reproduced badge
BibTeX reference
Download .bib
@inproceedings{gatebleed2025,
  author = {Kalyanapu, Joshua and Dizani, Farshad and Asher, Darsh and Ghanbari, Azam and Cammarota, Rosario and Aysu, Aydin and Ajorpaz, Samira Mirbagher},
  title = {{GateBleed: Exploiting On-Core Accelerator Power Gating for High Performance and Stealthy Attacks on AI}},
  booktitle = {Proceedings of the 58th IEEE/ACM International Symposium on Microarchitecture},
  year = {2025},
  pages = {308--325},
  doi = {10.1145/3725843.3756097},
  url = {https://doi.org/10.1145/3725843.3756097}
}
IEEE CAL

THOR: A Non-Speculative Value Dependent Timing Side Channel Attack Exploiting Intel AMX

Farshad Dizani, Azam Ghanbari, Joshua Kalyanapu, Darsh Asher, and Samira Mirbagher Ajorpaz

Characterizing how operand values influence matrix-accelerator timing and create information leakage.

BibTeX reference
Download .bib
@article{thor2025,
  author = {Dizani, Farshad and Ghanbari, Azam and Kalyanapu, Joshua and Asher, Darsh and Ajorpaz, Samira Mirbagher},
  title = {{Thor: A Non-Speculative Value Dependent Timing Side Channel Attack Exploiting Intel AMX}},
  journal = {IEEE Computer Architecture Letters},
  year = {2025},
  volume = {24},
  number = {1},
  pages = {69--72},
  doi = {10.1109/LCA.2025.3544989},
  url = {https://doi.org/10.1109/LCA.2025.3544989}
}
IEEE CAL

Exploiting Intel AMX Power Gating

Joshua Kalyanapu, Farshad Dizani, Azam Ghanbari, Darsh Asher, and Samira Mirbagher Ajorpaz

BibTeX reference
Download .bib
@article{intelamxpowergating2025,
  author = {Kalyanapu, Joshua and Dizani, Farshad and Ghanbari, Azam and Asher, Darsh and Ajorpaz, Samira Mirbagher},
  title = {{Exploiting Intel AMX Power Gating}},
  journal = {IEEE Computer Architecture Letters},
  year = {2025},
  volume = {24},
  number = {1},
  pages = {113--116},
  doi = {10.1109/LCA.2025.3555183},
  url = {https://doi.org/10.1109/LCA.2025.3555183}
}

2022

MICRO

EVAX: Towards a Practical, Pro-active & Adaptive Architecture for High Performance & Security

Samira Mirbagher Ajorpaz, Daniel Moghimi, Jeffrey Neal Collins, Gilles Pokam, Nael Abu-Ghazaleh, and Dean Tullsen

Adaptive hardware detection strengthened through the generation and study of evasive attacks.

BibTeX reference
Download .bib
@inproceedings{evax2022,
  author = {Ajorpaz, Samira Mirbagher and Moghimi, Daniel and Collins, Jeffrey Neal and Pokam, Gilles and Abu-Ghazaleh, Nael and Tullsen, Dean},
  title = {{EVAX: Towards a Practical, Pro-active \& Adaptive Architecture for High Performance \& Security}},
  booktitle = {2022 55th IEEE/ACM International Symposium on Microarchitecture (MICRO)},
  year = {2022},
  pages = {1218--1236},
  doi = {10.1109/MICRO56248.2022.00085},
  url = {https://doi.org/10.1109/MICRO56248.2022.00085}
}
ICCD

Composite Instruction Prefetching

Gino Chacon, Elba Garza, Alexandra Jimborean, Alberto Ros, Paul V. Gratz, Daniel A. Jiménez, and Samira Mirbagher Ajorpaz

Camera-ready author version. Gino Chacon and Elba Garza contributed equally.

BibTeX reference
Download .bib
@inproceedings{compositeinstructionprefetching2022,
  author = {Chacon, Gino and Garza, Elba and Jimborean, Alexandra and Ros, Alberto and Gratz, Paul V. and Jimenez, Daniel A. and Mirbagher-Ajorpaz, Samira},
  title = {{Composite Instruction Prefetching}},
  booktitle = {2022 IEEE 40th International Conference on Computer Design (ICCD)},
  year = {2022},
  pages = {471--478},
  doi = {10.1109/ICCD56317.2022.00076},
  url = {https://doi.org/10.1109/ICCD56317.2022.00076}
}
SBAC-PAD

Dynamic Set Stealing to Improve Cache Performance

Brady Testa, Samira Mirbagher Ajorpaz, and Daniel A. Jiménez

BibTeX reference
Download .bib
@inproceedings{dynamicsetstealing2022,
  author = {Testa, Brady and Mirbagher-Ajorpaz, Samira and Jimenez, Daniel A.},
  title = {{Dynamic Set Stealing to Improve Cache Performance}},
  booktitle = {2022 IEEE 34th International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD)},
  year = {2022},
  pages = {60--70},
  doi = {10.1109/SBAC-PAD55451.2022.00017},
  url = {https://doi.org/10.1109/SBAC-PAD55451.2022.00017}
}

2020

MICRO

PerSpectron: Detecting Invariant Footprints of Microarchitectural Attacks with Perceptron

Samira Mirbagher Ajorpaz, Gilles Pokam, Esmaeil Mohammadian-Koruyeh, Elba Garza, Nael Abu-Ghazaleh, and Daniel A. Jiménez

Using microarchitectural observations and compact hardware learning to detect attacks before information is leaked.

BibTeX reference
Download .bib
@inproceedings{perspectron2020,
  author = {Mirbagher-Ajorpaz, Samira and Pokam, Gilles and Mohammadian-Koruyeh, Esmaeil and Garza, Elba and Abu-Ghazaleh, Nael and Jimenez, Daniel A.},
  title = {{PerSpectron: Detecting Invariant Footprints of Microarchitectural Attacks with Perceptron}},
  booktitle = {2020 53rd Annual IEEE/ACM International Symposium on Microarchitecture (MICRO)},
  year = {2020},
  pages = {1124--1137},
  doi = {10.1109/MICRO50266.2020.00093},
  url = {https://doi.org/10.1109/MICRO50266.2020.00093}
}
MICRO

CHiRP: Control-Flow History Reuse Prediction

Samira Mirbagher Ajorpaz, Elba Garza, Gilles Pokam, and Daniel A. Jiménez

First place, 2020 IEEE Computer Society Lance Stafford Larson Writing Award (graduate category).

BibTeX reference
Download .bib
@inproceedings{chirp2020,
  author = {Mirbagher-Ajorpaz, Samira and Garza, Elba and Pokam, Gilles and Jimenez, Daniel A.},
  title = {{CHiRP: Control-Flow History Reuse Prediction}},
  booktitle = {2020 53rd Annual IEEE/ACM International Symposium on Microarchitecture (MICRO)},
  year = {2020},
  pages = {131--145},
  doi = {10.1109/MICRO50266.2020.00023},
  url = {https://doi.org/10.1109/MICRO50266.2020.00023}
}

2019

ISCA

Bit-level Perceptron Prediction for Indirect Branches

Elba Garza, Samira Mirbagher Ajorpaz, Tahsin Ahmad Khan, and Daniel A. Jiménez

BibTeX reference
Download .bib
@inproceedings{bitlevelperceptron2019,
  author = {Garza, Elba and Mirbagher-Ajorpaz, Samira and Khan, Tahsin Ahmad and Jim{\'e}nez, Daniel A.},
  title = {{Bit-level perceptron prediction for indirect branches}},
  booktitle = {Proceedings of the 46th International Symposium on Computer Architecture},
  year = {2019},
  pages = {27--38},
  doi = {10.1145/3307650.3322217},
  url = {https://doi.org/10.1145/3307650.3322217}
}

2018

ISCA

Exploring Predictive Replacement Policies for Instruction Cache and Branch Target Buffer

Samira Mirbagher Ajorpaz, Elba Garza, Sangam Jindal, and Daniel A. Jiménez

BibTeX reference
Download .bib
@inproceedings{predictivereplacement2018,
  author = {Mirbagher Ajorpaz, Samira and Garza, Elba and Jindal, Sangam and Jimenez, Daniel A.},
  title = {{Exploring Predictive Replacement Policies for Instruction Cache and Branch Target Buffer}},
  booktitle = {2018 ACM/IEEE 45th Annual International Symposium on Computer Architecture (ISCA)},
  year = {2018},
  pages = {519--532},
  doi = {10.1109/ISCA.2018.00050},
  url = {https://doi.org/10.1109/ISCA.2018.00050}
}

Student theses