SVM-based Analysis for Predicting Success Rate of Interest Packets in Information Centric Networks

Figures & data

Figure 1. The diagram demonstrates the flow of Interest packet to search content. There are two clients C₁ and C_2, requesting content N₁ and N₂ respectively. P₁ and P₂ are the generators of N₁ and N₂. Moreover N₁ is cached in CR₁ and hence C₁ is may be served by CR₁ and C₂ is served by the actual producer of N₂ (as it is not cached in any CR). Our algorithm tries to predict the appropriate location of the cache store using ML models.

Figure 2. System model for applying ML in ICN routing is presented in the diagram. Flow of operation starting from data collection till the findings of the proposed work as comparison of predicted and actual success rate of Interest packet is depicted.

Table 1. A comparison of similar work with proposed forwarding strategy

Reference	Objective	Contribution of the work	Comparison with our work	Use of ML model
Banerjee et al. (2017)	A routing strategy based on characteristic time of content in a cache to improve content search latency.	The proposed routing is dependent on the underlying caching strategy. Provides a probabilistic approach for content retrieval.	This work does not use any heuristic knowledge to predict content location. In our work we use a priori knowledge recorded in dataset.	No machine learning approach is used in this work like in our approach.
Zhu and Afanasyev (2013)	Exploits the features of the NDN naming rules from condensed cryptographic dataset which securely route data.	This work deals with security provisioning in forwarding and routing content.	Although they use heuristic dataset they apply the apriori knowledge to secure data routing. Like ours, they do not deal with the access latency of contents.	No ML prediction models are used in the work.
Lehman et al. (2016)	An Adaptive Smoothed RTT-based Forwarding (ASF) mechanism to mitigate hierarchical routing path selection.	An experimental investigation is demonstrated for packet delivery delay and overhead under and compared with NLSR.	No special care is taken while forwarding Interest packet to search content. In our approach, ML model helps in predicting the forwarding path.	ML models are not included in their discussions.
Proposed Work	A machine learning based forwarding strategy with prediction of data availability in cache locations.	A Interest forwarding approach guided by ML model to increase success rate of content retrieval and minimizes search latency.	– -	No ML prediction models are used in this work (Zhu and Afanasyev 2013).

Table 2. List of features in the dataset with their description. These are the features that captured at the time of collecting data from the simulation environment designed for ICN network using ndnSim. Later these features are analyzed to find the most important features to classify the dataset

Feature name	Description
nodeID	Unique identifier of the CR node that generates the Interest packet.
contentName	Unique name given to a content. Searching is done based on this name.
popularity	Every content has a popularity value computed based on its’ access frequency. The interest packet carries the last known popularity value for the requested content.
contentSize	The size in byte of the requested content and included in the Interest packet. The parameter ‘contentSize’ propagates through Content Advertisement (CA) message disseminated periodically by actual content producers. The detailed description of the CA process is not explained in this paper as this variable has no direct impact on the performance of the work presented here. Interested readers may refer to (Marandi et al. 2017; Quan et al. 2014).
ttlValue	The last known time of the requested content from the time of its caching in the network. This value is computed by the underlying caching mechanism.
inIF	ID of the interface through which the Interest packet is received.
outIF	ID of the interface through which the Interest packet is forwarded.
timeofReturn	Time period in which the response to the Interest is received by the CR.
queLength	Queue length of the outgoing interface through which the Interest packet is forwarded.
hopTraversed	Count of total number of hops traversed by the Interest packet so far including the current CR.
reqArrivalRate	The frequency of Interest packet arrival at the time of forwarding the packet.
dstCR	The ID of the CR from where the data is received.
Success	Target variable. Set to 0 initially to indicate not successful. Once the CR receives the response it sets to 1.

Figure 3. Correlation of features for determining the classification of data points. The two features namely ‘popularity’ and ‘ttl_value’ shows highly related correlation with values of 0.51 and 0.46, respectively. These two features are used for SVM classification in the later part of this work.

Table 3. Comparison of prediction and accuracy of the content retrieval for different features. This tables shows the best results with two features, popularity and ttl_value. (Parameters Gamma = 0.7, C = 100; all values are in %-age)

Feature count	Selected features	Failure prediction	Success prediction	Accuracy
	RBF kernel
1	ttl_value	84	48	78
1	popularity	84	46	78
2	popularity, ttl_value	86	52	80
3	popularity, ttl_value, QLen	82	69	81
12	All features	79	0	79
	Linear kernel
1	ttl_value	93	78	90
1	popularity	93	78	90
2	popularity, ttl_value	93	88	92
3	popularity, ttl_value, QLen	92	86	91
12	All features	93	78	90
	Polynomial kernel (with degree 3)
1	ttl_value	79	0	79
1	popularity	84	46	78
2	popularity, ttl_value	79	100	90
3	popularity, ttl_value, QLen	79	0	51
12	All features	79	0	79

Figure 4. Scatter plot of the ICN dataset to observe the separability for classification. correlation of features for determining the classification of data points. The data could be linearly separated with little misclassification. So, SVM is used for modeling.

Table

Step	Algo ml_model_for_prediction()
1	Input: Content_name, popularity, ttl_value Output: oIFace
2	Start
3	oIFace = null
4	If (popularity $\ge$ 0.6 and ttl_value $\ge$ 400) then
5	oIFace = lookup(popTab, popularity)
6	Else
7	oIFace = null
8	Endif
9	If (oIFace ! = null) then
10	Forward(Interest(Content_name, oIFace)
11	Else
12	Forward(Interest(Content_name))
13	Endif
13	Stop

Table 4. Simulation parameter

Parameter	Typical values
Network size	US26 network with 26 nodes
Number of contents	1000 contents
Content size	5000B
Cache size	50–150 contents
Betweeness centrality	1–5
Simulation time	100–650 Sec

Figure 5. A sample simulation topology based on US26 network. Approximately 1000 unique contents with a size of 5000B is taken for simulation. The cache store size is of 50–150 contents. Nodes with dotted circles have betweenness centrality (BC) value in the range 3–5 and results are collected in these CR nodes.

Figure 6. Interest packet success rate over simulation time. The plot considers those successful processing whose content is found in intermediate cache. Serving from the actual producer is not counted in the above results.

Figure 7. Interest packet success rate over request arrival rate. The plot considers only successful Interest requests served from the cache. Results are recorded after 100sec of simulation progress. By this time elapse most of the caches are filled with sufficient amount of data. Hence, there is a high possibility of getting data in cache.

Banerjee, B., A. Seetharam, A. Mukherjee, and M. Kanti Naskar. 2017. Characteristic time routing in information centric networks. Computer Networks 113:148–58. doi:10.1016/j.comnet.2016.12.009.

 Web of Science ®Google Scholar

Zhu, Z., and A. Afanasyev. 2013. Let’s ChronoSync: Decentralized dataset state synchronization in named data networking. Proceedings - International Conference on Network Protocols, ICNP, Texas, US. doi:10.1109/ICNP.2013.6733578.

 Google Scholar

Lehman, V., A. Gawande, B. Zhang, L. Zhang, R. Aldecoa, D. Krioukov, and L. Wang. 2016. An experimental investigation of hyperbolic routing with a smart forwarding plane in NDN. 2016 IEEE/ACM 24th International Symposium on Quality of Service, IWQoS 2016, Tokyo, Japan. doi:10.1109/IWQoS.2016.7590394.

 Google Scholar

Marandi, A., T. Braun, K. Salamatian, and N. Thomos. 2017. BFR: A bloom filter-based routing approach for information-centric networks. 2017 IFIP Networking Conference, IFIP Networking 2017 and Workshops, Stockholm, Sweden. doi:10.23919/IFIPNetworking.2017.8264842.

 Google Scholar

Quan, W., C. Xu, J. Guan, H. Zhang, and L. A. Grieco. 2014. Scalable name lookup with adaptive prefix bloom filter for named data networking. IEEE Communications Letters 18 (1):1. doi:10.1109/LCOMM.2013.112413.132231.

 Web of Science ®Google Scholar