Junji Takemasa

Junji Takemasa

Auto-generated profile, based on publicly available information

Junji Takemasa

Graduate School of Information Science and Technology, Osaka University, Osaka, Japan

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
19 May 2023 in IEEE Access
Verified
Journal Article
Open access
HMHiroki Masuda
Hiroki Masuda
KKKentaro Kita
Kentaro Kita
YKYuki Koizumi
Yuki Koizumi
JTJunji Takemasa
Junji Takemasa
THToru Hasegawa
Toru Hasegawa
Privacy-preserving and Byzantine-resilient machine learning has been an important research issue, and many centralized methods have been developed. However, it is difficult for these methods to achieve fast learning and high accuracy simultaneously. In contrast, federated learning based on local model masking like Byzantine-Resilient Secure Aggregation (BREA), is a promising approach to simultaneously achieve them. Despite the advantage of light computation of randomizing local models of users for privacy preservation, the verification of shares generated from local models in BREA, which mitigates Byzantine attacks, still incurs large complexity in communication. The paper designs a share verification method for BREA to offload some parts of the share verification process from users to a semi-honest server, which avoids broadcasting large-size commitments to shares. In addition, to mitigate the increase in computation time due to computations offloaded to the server, our method makes the verification algorithm running on the server efficient and executes the server and user computations in parallel. In our experiments, our method provides a speedup of up to 15× on low-bandwidth networks like mobile networks. Our method also preserves BREA’s resilience against Byzantine attacks.

9 June 2020 in IEEE Access
Verified
Journal Article
Open access
YYYoji Yamamoto
Yoji Yamamoto
JTJunji Takemasa
Junji Takemasa
YKYuki Koizumi
Yuki Koizumi
THToru Hasegawa
Toru Hasegawa
Increase in the size of content due to the emergence of bandwidth-intensive applications leads to a problematic phenomenon of packet-level caching which is called looped replacement. This causes a waste of cache usage and results in caching performance degradation. The goals of the paper are twofold: identifying the root cause of looped replacement and mitigating it. In order to achieve the goals, we analytically investigate not only how looped replacement occurs, but also how cache hit probability decreases. According to the analysis, we propose to use cache admission, which only admits packets arriving frequently to a cache, as a countermeasure against both problems. Finally, we develop an analytical model of packet-level caching with cache admission, and theoretically proves that cache admission mitigates these problems and improves cache hit probability.

11 May 2023 in IEEE Transactions on Network and Service Management
Verified
Journal Article
Closed access
YYYutaro Yoshinaka
Yutaro Yoshinaka
KKKentaro Kita
Kentaro Kita
JTJunji Takemasa
Junji Takemasa
YKYuki Koizumi
Yuki Koizumi
THToru Hasegawa
Toru Hasegawa
Consumer privacy leakage from data names poses a serious threat to Content-Centric Networking (CCN) networks. Obfuscating names is a promising countermeasure, and anonymizers with deterministic encryption schemes have been proposed to provide data privacy while enabling CCN features, such as in-network caching. Existing studies assume a weak threat model in which anonymizers are honest, and their obfuscation schemes are not resilient against privacy attacks such as name guessing attacks. This paper designs a name obfuscation framework based on the realistic assumption that anonymizers are semi-honest. The framework strengthens data privacy using multiple keys and separates obfuscation for prefixes and suffixes, and is implemented on a P4 switch to provide Tbps forwarding speed.

21 August 2020 in IEEE Access
Verified
Journal Article
Open access
JTJunji Takemasa
Junji Takemasa
ATAtsushi Tagami
Atsushi Tagami
YKYuki Koizumi
Yuki Koizumi
THToru Hasegawa
Toru Hasegawa
Equal load balancing for dispatching incoming packets to multiple threads is a crucial requirement in the stateful forwarding of multi-threaded software routers to achieve high-speed forwarding and low packet loss simultaneously. However, equal load balancing is not trivial for Named Data Networking (NDN) routers because of their stateful forwarding. In other words, the consistency of flow states should be maintained so that multiple threads do not access the states simultaneously. Sharding, wherein packets of the same flow are dispatched to the same thread while keeping loads of threads equal, has been proposed; however, in this study, we reveal that heavy hitters like popular content packets cause load imbalance, which may eventually cause packet losses. This study proposes a load balancing mechanism for NDN routers by exploiting the fact that states of flows need not be rigorously maintained when content packets are returned from caches at intermediate routers.

30 June 2025 in IEEE Transactions on Networking
Verified
Journal Article
Closed access
YYYutaro Yoshinaka
Yutaro Yoshinaka
YKYuki Koizumi
Yuki Koizumi
JTJunji Takemasa
Junji Takemasa
THToru Hasegawa
Toru Hasegawa
Programmable switches offer a promising platform for fast and flexible in-network computing. However, a standard mechanism, packet recirculation, degrades throughput due to bandwidth consumption caused by the loopback of not only packet headers but also cumbersome payloads. This paper proposes P4QRS, a mechanism that reduces payload recirculation by retaining payloads within the switch. Specifically, P4QRS bifurcates packets into headers and payloads, which undergo the computation process through pipelines and the buffering process leveraging the switch’s queue behavior, respectively. The headers and payloads then rendezvous for reassembly into complete packets to be sent out. To validate its effectiveness, we evaluated P4QRS through both an analytical model and implementation on state-of-the-art hardware programmable switches. Our results demonstrate that P4QRS operates stably and significantly accelerates complex in-switch computations. Moreover, we address packet reordering, which is a fundamental concern arising because P4QRS performs payload buffering on a per-packet basis rather than per-flow. After identifying the mechanism for developing packet reordering in three stages, we design, implement, and evaluate mitigation strategies targeting each stage. These strategies effectively suppress packet reordering in both flow-aware and flow-unaware cases.

5 November 2024 in IEEE Access
Verified
Journal Article
Open access
HMHiroki Masuda
Hiroki Masuda
KKKentaro Kita
Kentaro Kita
YKYuki Koizumi
Yuki Koizumi
JTJunji Takemasa
Junji Takemasa
THToru Hasegawa
Toru Hasegawa
Federated learning is a promising collaborative learning system from the perspective of training data privacy preservation; however, there is a risk of privacy leakage from individual local models of users. Secure aggregation protocols based on local model masking are a promising solution to prevent privacy leakage. Existing secure aggregation protocols sacrifice either computation or communication costs to tolerate user dropouts. A naive secure aggregation protocol achieves a small communication cost by secretly sharing random seeds instead of random masks. However, it requires that a server incurs a substantial computation cost to reconstruct the random masks from the random seeds of dropout users. To avoid such a reconstruction, a state-of-the-art secure aggregation protocol secretly shares random masks. Although this approach avoids the computation cost of mask reconstruction, it incurs a large communication cost due to secretly sharing random masks. In this paper, we design a secure aggregation protocol to mitigate the tradeoff between the computation cost and the communication cost by complementing both types of secure aggregation protocols. In our experiments, our protocol achieves up to 11.41 times faster while achieving the same level of privacy preservation and dropout tolerance as the existing protocols.