Have I Met You Before? Using Cross-Media

Transcription

1 Have I Met You Before? Using CrossMedia Kumiko Ono and Henning Schulzrinne Columbia University

2 Outline 1. Introduction Results of our quick survey Legitimate calls from those with weak social ties Legitimate calls with no Caller IDs 2. Proposal: Labeling calls using crossmedia relations 3. Conclusions

3 Introduction Survey of Incoming Calls Cell phones Landline phones Fewer unsolicited calls to cell phones 6% 52% 48% Effect of national regulations 94% Legitimate calls Unsolicited calls Data: Calls to students in our lab in Feb (246 calls to cell, 136 calls to landlines) 3

4 Introduction Survey of Incoming Calls Cell phones Landline phones 1729 % could not be labeled legitimate before answering. 13% 4% 6% 78% 52% 18% 29% Known caller IDs Unknown caller IDs from known persons No caller IDs from known persons Unsolicited calls Data: Calls to students in our lab in Feb (246 calls to cell, 136 calls to landlines) w/o caller IDs: international calls w/unknown caller IDs: from persons/companies having prior contact via web or Crossmedia relations 4

5 Introduction Typical SPIT Filter Receive a request for realtime communication (audio/video/text) SPIT Filtering at server/user agent NG Set unavailable at Caller ID Caller ID Authentication OK Reject Yes On a black list? No Reject False positives Challenge: How can we reduce false positives? A voic box No On a white list? Yes Accept 5

6 Introduction Black List and White List Black list: to reject requests Block list Known SPIT caller addresses at a reputation service White list: to accept requests Buddy list Addresses of outgoing requests in call history Directory service, e.g., LDAP server Limits to strong ties 6

7 Introduction Problem and Approaches Problem How to label incoming requests Legitimate calls from those with weak social ties Legitimate calls with no Caller IDs Approaches Focus on prior contact when delivering callee s contact address through different media Enhance white list beyond caller IDs Crossmedia relations 7

8 Introduction AntiSPIT Work A collection of solutions is needed. I: Places burdens on callers DoNotCall registry service Turing tests to detect a prerecorded message II: Labels incoming requests at callees or service providers 8 Blacklists/Whitelists based on Caller ID: how to learn/ share/expand lists Suspicious traffic I&II: Prior permissions from callee to caller in SIP Using HTTP responses to publish caller IDs for future calls

9 Proposal: Using Crossmedia Relations Insert data User (Callee) Web Browser Insert data HTTPS Web Server Potential Callers Filter conditions Crossmedia relations 1: Information offered by potential callers Contact addresses (plain) Crossmedia relations 2: Weaklysecret information offered by callee Outgoing MessageIDs Contact addresses (hashed) Tokens in customized callee s addresses 9

10 Proposal CrossMedia Relations Offered by Callers Callee Web client HTTP request Potential caller Web server e.g., HTTP response CorrespondenceURI: tel: Potential caller client e.g., Callee client Message ContentType: text/directory; profile= vcard Forms: Caller s contact addresses in plain text or hashed Hashed are only for labeling. Better for callers privacy to avoid publishing routable addresses Containers An HTTP header in response Vcard in message body 10

11 Proposal CrossMedia Relations Offered by Callees Callee Web client HTTP response Potential caller Web server e.g., HTTP POST columbia.edu&phone2=tel: Potential caller client e.g., Callee client Message requesting realtime comm. Message to accept MessageID:<004301c9b17f257f6a 11 Weaksecret as a proof of prior contact Subaddressing in callee s contact address in HTTP POST form A field in signup form in message body Similar to subaddressing in addresses or extensions of phone numbers MessageIDs of outgoing s

12 Proposal Calls w/crossmedia Relations User Web Browser Web Server Potential Callers Caller e.g., INVITE From: INVITE To: INVITE References:004301c9b17f257f6a rel= Filter Conditions CrossMedia Relations including CallIDs Query Inbound SIP Proxy Server w/filter User (Callee) 12

13 Proposal Proposed SPIT Filter Receive a request of realtime communication (audio/video/text) Spam Filtering at server/user agent Set unavailable at Caller ID NG Sender ID Authentication OK Reject Yes On a black list? No Reject To has a valid Token? No Reference has a valid Message IDs? No On a white list? (plain/hashed) Yes Yes Yes Accept 13

14 Conclusions Summary Goal: Reduce false positive filtering by enabling callees to lable incoming requests from weak social ties Approach: Use crossmedia relations as a proof of prior contact Next Step Caller ID in plain text or hashed format, customized callee s ID, and MessageID Evaluation of usability and effectiveness using prototype 14