Ayantika Chatterjee
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Ayantika Chatterjee has a total of 31 co-authors. Here are the top co-authors by number of shared publications:
- Debdeep Mukhopadhyay
Computer Science and Engineering, Indian Institute of Technology Kharagpur
7 shared publications Khin Mi Mi Aung
A*STAR
6 shared publications- Tanusree Parbat
Advanced Technology Development Centre, Indian Institute of Technology Kharagpur, Kharagpur, WB, India
4 shared publications - Anirban Chakraborty
Max Planck Institute for Security and Privacy, Bochum, Germany
3 shared publications - Bhuvnesh Chaturvedi
Indian Institute of Technology Kharagpur
3 shared publications - Indranil Sengupta
Department of Computer Science and Engineering, Indian Institute of Technology Kharagpur, Kharagpur, India
3 shared publications
Latest Publications
- Open Access
- 14 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 5806-5820
FHEMIST: A Maximum Independent Set Technique for Efficient Bootstrapping Reductions in FHE Evaluations
- 14 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 5806-5820
Open AccessFHEMIST: A Maximum Independent Set Technique for Efficient Bootstrapping Reductions in FHE Evaluations
- Open Access
- 14 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 7369-7371
POSTER: Universal Defensive Perturbations: A Cluster-Based Ensemble Defense
- 14 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 7369-7371
Open AccessPOSTER: Universal Defensive Perturbations: A Cluster-Based Ensemble Defense
- Open Access
- 23 September 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Internet Technology
- Vol. 26 (4) , 1-30
EHEIR: Efficient Homomorphic Encrypted Inference via Architectural Redesign
Encrypted inference under homomorphic encryption (HE) imposes stringent constraints on computation, making the direct deployment of deep neural networks infeasible on resource-constrained edge devices. Existing approaches, such as FHEEDGE and UniHENN, primarily focus on improving the efficiency of encrypted computation through algorithmic and systems-level optimizations. In contrast, we adopt a complementary perspective by redesigning neural architectures to align with the inherent cost model of HE operations. In this work, we present a framework that reformulates HE-aware model design as a constrained neural architecture search problem, where the objective is to identify architectures that are both cryptographically feasible and computationally efficient while preserving task performance. Starting from a pretrained network, we perform a structured, iterative transformation process that includes HE-compatible operator substitutions, depth–width tradeoffs, and rotation-efficient design strategies such as successive average pooling. These transformations are guided by constraints on ciphertext slots, multiplicative depth, and rotation complexity, and are evaluated using short-horizon training to efficiently explore the architecture space. We provide a theoretical analysis showing that the transformed network approximates the original model with bounded error under layer-wise Lipschitz propagation, ensuring controlled performance degradation. As a case study, we apply the proposed framework to encrypted crack detection for structural health monitoring. The optimized model achieves an execution time of 86.5 seconds for single-threaded CPU inference in the encrypted domain, compared with an estimated 3.5 hours required by the original encrypted convolutional pipeline. Furthermore, deployment on a Raspberry Pi demonstrates end-to-end encrypted inference within approximately 7.5 minutes. These results highlight that architecture-level optimization, when framed as a constrained search problem, complements existing algorithmic acceleration approaches and provides a practical pathway for enabling efficient and privacy-preserving inference on edge devices across a broad class of sensitive AI workloads.
- 23 September 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Internet Technology
- Vol. 26 (4) , 1-30
Open AccessEHEIR: Efficient Homomorphic Encrypted Inference via Architectural Redesign
Encrypted inference under homomorphic encryption (HE) imposes stringent constraints on computation, making the direct deployment of deep neural networks infeasible on resource-constrained edge devices. Existing approaches, such as FHEEDGE and UniHENN, primarily focus on improving the efficiency of encrypted computation through algorithmic and systems-level optimizations. In contrast, we adopt a complementary perspective by redesigning neural architectures to align with the inherent cost model of HE operations. In this work, we present a framework that reformulates HE-aware model design as a constrained neural architecture search problem, where the objective is to identify architectures that are both cryptographically feasible and computationally efficient while preserving task performance. Starting from a pretrained network, we perform a structured, iterative transformation process that includes HE-compatible operator substitutions, depth–width tradeoffs, and rotation-efficient design strategies such as successive average pooling. These transformations are guided by constraints on ciphertext slots, multiplicative depth, and rotation complexity, and are evaluated using short-horizon training to efficiently explore the architecture space. We provide a theoretical analysis showing that the transformed network approximates the original model with bounded error under layer-wise Lipschitz propagation, ensuring controlled performance degradation. As a case study, we apply the proposed framework to encrypted crack detection for structural health monitoring. The optimized model achieves an execution time of 86.5 seconds for single-threaded CPU inference in the encrypted domain, compared with an estimated 3.5 hours required by the original encrypted convolutional pipeline. Furthermore, deployment on a Raspberry Pi demonstrates end-to-end encrypted inference within approximately 7.5 minutes. These results highlight that architecture-level optimization, when framed as a constrained search problem, complements existing algorithmic acceleration approaches and provides a practical pathway for enabling efficient and privacy-preserving inference on edge devices across a broad class of sensitive AI workloads.
- Open Access
- 21 September 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Cyber-Physical Systems
- Vol. 10 (4) , 1-33
SPlatoon: Secure Platoon with Lightweight Authenticated Encryption
Vehicle platoons improve road capacity and safety by enabling cooperative driving with small inter-vehicle gaps. Such coordination requires a timely wireless exchange of vehicle state information (speed, position, and acceleration). However, vehicular networks are vulnerable to attacks that can manipulate shared information and compromise platoon safety. To secure message exchange, the Authenticated Encryption (AE) scheme is employed. AE transforms Cooperative Awareness Message (CAM) contents into ciphertext, and appends security fields such as nonce, authentication tag, and security header. These additions increase the encrypted CAM size and channel load, making packets more susceptible to delay in dense wireless environments, degrading platoon performance. One way to reduce this overhead is to lower the numerical precision of the CAM payload before encryption, thereby reducing the encrypted packet size. Our results show that platooning remains safe for certain reduced precision levels, indicating that full-precision CAMs are not always necessary to maintain safety. However, an arbitrary reduction of the CAM payload precision leads to inaccurate control updates that propagate along the platoon and may compromise safety. This work highlights the design tradeoff between security overhead and platoon control safety. To address this challenge, we propose SPlatoon , a tool that consists of: (i) a precision-aware Computer-Aided Design (CAD) framework, and (ii) a platoon simulator . The CAD-based framework performs design space exploration to systematically evaluate how different precision levels affect platoon safety and scalability. It helps in selecting the smallest safe reduced-precision encrypted CAM size, along with the maximum safe platoon size. The platoon simulator validates the CAD framework outputs under realistic highway scenarios by simulating vehicular traffic with secured message exchanges using a combination of OMNeT++, PLEXE, and SUMO libraries. The simulator incorporates ASCON, a lightweight AE scheme, inside the network stack of OMNeT++ for secure vehicular communication.
- 21 September 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Cyber-Physical Systems
- Vol. 10 (4) , 1-33
Open AccessSPlatoon: Secure Platoon with Lightweight Authenticated Encryption
Vehicle platoons improve road capacity and safety by enabling cooperative driving with small inter-vehicle gaps. Such coordination requires a timely wireless exchange of vehicle state information (speed, position, and acceleration). However, vehicular networks are vulnerable to attacks that can manipulate shared information and compromise platoon safety. To secure message exchange, the Authenticated Encryption (AE) scheme is employed. AE transforms Cooperative Awareness Message (CAM) contents into ciphertext, and appends security fields such as nonce, authentication tag, and security header. These additions increase the encrypted CAM size and channel load, making packets more susceptible to delay in dense wireless environments, degrading platoon performance. One way to reduce this overhead is to lower the numerical precision of the CAM payload before encryption, thereby reducing the encrypted packet size. Our results show that platooning remains safe for certain reduced precision levels, indicating that full-precision CAMs are not always necessary to maintain safety. However, an arbitrary reduction of the CAM payload precision leads to inaccurate control updates that propagate along the platoon and may compromise safety. This work highlights the design tradeoff between security overhead and platoon control safety. To address this challenge, we propose SPlatoon , a tool that consists of: (i) a precision-aware Computer-Aided Design (CAD) framework, and (ii) a platoon simulator . The CAD-based framework performs design space exploration to systematically evaluate how different precision levels affect platoon safety and scalability. It helps in selecting the smallest safe reduced-precision encrypted CAM size, along with the maximum safe platoon size. The platoon simulator validates the CAD framework outputs under realistic highway scenarios by simulating vehicular traffic with secured message exchanges using a combination of OMNeT++, PLEXE, and SUMO libraries. The simulator incorporates ASCON, a lightweight AE scheme, inside the network stack of OMNeT++ for secure vehicular communication.
- 8 August 2026
- Published by Springer Nature in Journal of Hardware and Systems Security
- Vol. 10 (1) , 18
Transferability of Evasion and Backdoor Attacks in Encrypted Domain
- 8 August 2026
- Published by Springer Nature in Journal of Hardware and Systems Security
- Vol. 10 (1) , 18
Transferability of Evasion and Backdoor Attacks in Encrypted Domain
- 13 November 2025
- Published by Springer Nature
- p. 405-424
WiperSentinel: HPC Based Wiper Detection with Enhanced AutoEncoder
- 13 November 2025
- Published by Springer Nature
- p. 405-424
WiperSentinel: HPC Based Wiper Detection with Enhanced AutoEncoder
- Open Access
- 24 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 13-29
An Efficient Circuit Synthesis Framework for TFHE via Convex Sub-graph Optimization
- 24 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 13-29
Open AccessAn Efficient Circuit Synthesis Framework for TFHE via Convex Sub-graph Optimization
- 15 March 2025
- Published by Springer Nature
- p. 351-383
IND-CPA$^{{\textbf {C}}}$: A New Security Notion for Conditional Decryption in Fully Homomorphic Encryption- 15 March 2025
- Published by Springer Nature
- p. 351-383
IND-CPA$^{{\textbf {C}}}$: A New Security Notion for Conditional Decryption in Fully Homomorphic Encryption- 5 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Services Computing
- Vol. 18 (2) , 1026-1038
SHELDB: Client Storage Aware Homomorphic Encrypted Database Processing Framework With Low Communication Overhead
Database as a service (DBaaS) in cloud raises severe concern in terms of data security. Storing data in encrypted form may confirm data confidentiality. But, database processing cannot be supported in this encrypted form. Theoretically, homomorphic encryption is a solution to support direct encrypted data processing. In this work, we highlight one of the major challenges of FHE- encrypted query processing that demands huge data transfer requirements from cloud to client for final decryption at the end of SQL query execution. We show in light of Chosen Plaintext Attack (CPA) that in spite of performing conditional filtering through SQL queries over encrypted databases, the size of the resultant dataset cannot be less than the original size of the database. In this work, we make an effort to propose a new encrypted query processing framework termed as SHELDB which supports client storage compatibility and low communication overhead using block-wise final result transmission from cloud to client by extending the concept of TOP operator implementation in standard SQL. However, it is to be noted that realization of such optimization is not straightforward because of circuit-based implementation requirements with FHE gates. Our experimental demonstration shows that the proposed framework is capable of executing all TPC-C standard SQL queries with the aid of 8-core parallel processing within $\sim 12.65$ minutes for an encrypted database of $768 \times 9$ size with 16-bits elements each. Though the computation time is linear with the number of rows, we have explored map-reduce type parallel processing techniques to reduce the timing requirements for databases with larger rows. Consequently, our new query processing framework reduces the communication overhead from m to $\delta k$ rows ( $1 \leq \delta \leq block$) where the encrypted database contains m rows, $\delta$ is the number of blocks to be transmitted each time with $k$ = ($m/block$) rows. In spite of $k$ being a controllable parameter according to client storage and $\delta$ is dependent on the query parameters, final security analysis explains why the proposed technique is general database attack-resistant.
- 5 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Services Computing
- Vol. 18 (2) , 1026-1038
SHELDB: Client Storage Aware Homomorphic Encrypted Database Processing Framework With Low Communication Overhead
Database as a service (DBaaS) in cloud raises severe concern in terms of data security. Storing data in encrypted form may confirm data confidentiality. But, database processing cannot be supported in this encrypted form. Theoretically, homomorphic encryption is a solution to support direct encrypted data processing. In this work, we highlight one of the major challenges of FHE- encrypted query processing that demands huge data transfer requirements from cloud to client for final decryption at the end of SQL query execution. We show in light of Chosen Plaintext Attack (CPA) that in spite of performing conditional filtering through SQL queries over encrypted databases, the size of the resultant dataset cannot be less than the original size of the database. In this work, we make an effort to propose a new encrypted query processing framework termed as SHELDB which supports client storage compatibility and low communication overhead using block-wise final result transmission from cloud to client by extending the concept of TOP operator implementation in standard SQL. However, it is to be noted that realization of such optimization is not straightforward because of circuit-based implementation requirements with FHE gates. Our experimental demonstration shows that the proposed framework is capable of executing all TPC-C standard SQL queries with the aid of 8-core parallel processing within $\sim 12.65$ minutes for an encrypted database of $768 \times 9$ size with 16-bits elements each. Though the computation time is linear with the number of rows, we have explored map-reduce type parallel processing techniques to reduce the timing requirements for databases with larger rows. Consequently, our new query processing framework reduces the communication overhead from m to $\delta k$ rows ( $1 \leq \delta \leq block$) where the encrypted database contains m rows, $\delta$ is the number of blocks to be transmitted each time with $k$ = ($m/block$) rows. In spite of $k$ being a controllable parameter according to client storage and $\delta$ is dependent on the query parameters, final security analysis explains why the proposed technique is general database attack-resistant.
- 29 September 2024
- Published by Springer Nature
- p. 96-117
On the Security of Privacy-Preserving Machine Learning Against Model Stealing Attacks
- 29 September 2024
- Published by Springer Nature
- p. 96-117
On the Security of Privacy-Preserving Machine Learning Against Model Stealing Attacks