Author name cluster

Ankit Sharma

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

5 papers

1 author row

EAAI Journal 2025 Journal Article

Dual replay memory reinforcement learning framework for minority attack detection

Ankit Sharma
Manjeet Singh

Cyberattacks are becoming more frequent as the number of systems connected to the Internet has grown significantly. Intrusion detection systems based on machine learning have become essential to cyber security because they can evaluate and find patterns in huge amounts of data. A branch of machine learning called reinforcement learning has shown promise in identifying new types of cyberattacks, but the class imbalance problem remains unresolved. Addressing this gap, this work presents a novel offline reinforcement learning framework that significantly enhances the capabilities of traditional Deep Q-network based reinforcement learning models for network intrusion detection. The innovation lies in using dual replay memory, which prioritizes learning minority attack classes. Experiments with and without dual-replay memory on the same offline reinforcement learning framework highlighted the significant improvements in evaluation metrics like accuracy, precision, recall and F1-Score. The research investigated the influence of various ratios of replay memories on learning through the grid search method. It evaluated their performance for each class using receiver operating characteristics and area under the curve. Experimental results using the one-vs-rest and aggregated approaches demonstrated the effectiveness of handling the class imbalance problem with high accuracy. Compared to existing methods, the technique performs better in detecting attacks and reducing false positive rates. The proposed model represents significant progress in intrusion detection, offering a robust and highly effective solution for detecting common and minority network intrusions.

Details DOI

EAAI Journal 2024 Journal Article

Batch reinforcement learning approach using recursive feature elimination for network intrusion detection

Ankit Sharma
Manjeet Singh

In the era of escalating cyber threats and increased Internet dependence, enhancing intrusion detection systems is essential. Concerns are emerging regarding the viability and sustainability of existing approaches based on machine and deep learning. Prior work shows limited ability to detect low-frequency network intrusions and shows limited consideration of reinforcement learning for intrusion detection. The rising need for human intervention and the difficulty in detecting low-frequency attacks underscore the necessity for reinforcement learning models. Therefore, a model is proposed based on deep reinforcement learning framework and feature selection through the recursive feature elimination method for effective intrusion detection. Reinforcement learning's innovative way of learning through interactions and rewards helps in adaptive learning, while recursive feature elimination reduces computational load by reducing the number of features, making the model more efficient than traditional methods. The research investigates the influence of diverse feature selection techniques namely, correlation-based, information gain-based and recursive feature elimination on the accuracy of detecting various attacks. The experimental results demonstrated the effectiveness of the feature selection techniques within the deep reinforcement learning networks, achieving a maximum accuracy of 99. 12% and an F1-score of 99. 2%. Remarkably, the proposed model also excelled in detecting low-frequency attacks, utilizing only 60% of the dataset features. The study also provides new insights into how different discount factors in batch reinforcement learning affect intrusion detection system performance. These findings underscore the novelty of the model in improving intrusion detection system performance through the use of reinforcement learning and feature selection techniques.

Details DOI

AAAI Conference 2023 Conference Paper

Channel Regeneration: Improving Channel Utilization for Compact DNNs

Ankit Sharma
Hassan Foroosh

Overparameterized deep neural networks have redundant neurons that do not contribute to the network's accuracy. In this paper, we introduce a novel channel regeneration technique that reinvigorates these redundant channels by re-initializing its batch normalization scaling factor gamma. This re-initialization of BN gamma promotes regular weight updates during training. Furthermore, we show that channel regeneration encourages the channels to contribute equally to the learned representation and further boosts the generalization accuracy. We apply our technique at regular intervals of the training cycle to improve channel utilization. The solutions proposed in previous works either raise the total computational cost or increase the model complexity. Integrating the proposed channel regeneration technique into the training methodology of efficient architectures requires minimal effort and comes at no additional cost in size or memory. Extensive experiments on several image classification and semantic segmentation benchmarks demonstrate the effectiveness of applying the channel regeneration technique to compact architectures.

PDF Details DOI

EAAI Journal 2016 Journal Article

Social network regularized Sparse Linear Model for Top-N recommendation

Xiaodong Feng
Ankit Sharma
Jaideep Srivastava
Sen Wu
Zhiwei Tang

Social recommendation techniques have been developed to employ user׳s social connections for both rating prediction and Top-N recommendation. However, they are mostly using social network enhanced matrix factorization (MF) where the objective is to minimize the prediction error of rating scores, which makes it impractical and unsuccessful for Top-N recommendation. This paper thus focuses on developing more effective methods to utilize social network information for Top-N recommendation. Social network regularized Sparse LInear Model (SocSLIM) with its extensions incorporating local learning (LocSocSLIM) to improve efficiency are proposed. SocSLIM learns sparse coefficient matrix for users by solving a sparse representation problem over user-item rating/purchase matrix and user–user social network׳s adjacency matrix at the same time by sharing coefficient matrix. The coefficient matrix is used to predict the recommendation scores, which are then combined with a proposed item based Distance regularized Sparse LInear Model (DSLIM) to generate recommendations for the users. The experimental results demonstrate that SocSLIM effectively uses the social information to outperform the state-of-the-art methods by at least 12%. Moreover, the local weight learning extension LocSocSLIM significantly improves the efficiency up to 10 times as compared to SocSLIM as the original SLIM while achieving the close performance guarantees.

Details DOI

AAAI Conference 2010 Conference Paper

Asymmetric Spite in Auctions

Ankit Sharma
Tuomas Sandholm

In many auctions, agents bid more aggressively than selfinterest would prescribe. This can be explained by spite, where the agent’s utility not only increases in the agent’s surplus but also decreases as the other bidders’ surpluses increase. Spite can stem from long-term benefits from making competitors worse off and from inherent psychological effects. There have been important recent game-theoretic analyses of spiteful bidding assuming all agents are equally spiteful. We present, to our knowledge, the first auction analysis in the more realistic setting where bidders may be spiteful to different extents. We show that the equilibrium bidding function can still be written in the same form—except that the spite factor is replaced by an ‘expressed’ spite factor. This leads to bidders expressing spites that are higher or lower than their true spite depending on others’ spite. Perhaps surprisingly, in the two-bidder case, the mapping from true spite to expressed spite is the same across all common auction mechanisms. Furthermore, even with two bidders, important properties of symmetric-spite settings cease to hold: the allocation can be inefficient and the revenue ranking may reverse between first- and second-price auctions. We also show that in sealed-bid auctions under asymmetric valuation distributions, there can be a “bargaining problem” in selecting bids. Finally, we study the generalization where agents can have different extents of spite toward different other bidders.

PDF Details