370 likes | 675 Vues
On the Distribution of Responsibility for Network Security. Scott Dexter Dept. of Computer and Information Science Brooklyn College. Overview. My Perspective My Argument Dramatis Personae Their Roles (and Malfeasances) and Their Eventual Rehabilitation. My Perspective.
E N D
On the Distribution of Responsibility for Network Security Scott Dexter Dept. of Computer and Information Science Brooklyn College
Overview • My Perspective • My Argument • Dramatis Personae • Their Roles (and Malfeasances) • and Their Eventual Rehabilitation
My Perspective • Theoretician by training and inclination • Dissertation: formal/logical analysis of security protocols • Implementation details sometimes regarded as abhorrent • Interested in relationship between individual empowerment and collective behavior • Deeply committed to (public) education
My Argument • Network/computer security is really interesting... • …mostly because of its social intractability: • Any actor, at any level, may (easily) cause a “security breach” • Probably due only to ignorance, human error, laziness, or even kindness • But could impact many others • No easy fix • Need some technical solutions • Need lots of different kinds of education
Network Technologies • “The Internet” • One of many technologies • (and can be said to be composed of many itself) • Designed for robustness and reliability (which is one aspect of “security”) • Designed to accommodate innovation • Also need • Proprietary/closed networks (e.g. bank machines) • Network applications “on top of” Internet
Actors • Designers • Implementers • Administrators • Users • (“Hackers”)
Design I: Internet Core Protocols • Info transmitted in sequence of datagrams • Data ‘payload’ plus complex array of control info • Achieves robustness and reliability in non-hostile environment • Possible to ‘craft’ illegal/bogus datagrams • Get information from nature of response • Circumvent firewalls & intrusion detectors • This is integral aspect of Internet!
Design II: Cryptographic Security • Appropriate use of cryptography has potential to solve many security problems • Can support many services: • Confidentiality (hide content from others) • Integrity (assure that content hasn’t changed) • Authenticity (demonstrate/confirm identity) • But appropriate is incredibly difficult….
One Scenario: “Key Distribution” Alice Bob
One Scenario: “Key Distribution” Alice Bob
One Scenario: “Key Distribution” K Alice Bob
Mallory One Scenario: “Key Distribution” K Alice Bob KA KB Solomon
Passive Attacks Eavesdropping Public Information Agent Identities Algorithms Protocol Design “Legitimate” Identity Active Attacks Modification Redirection Suppression Fake Messages Mallory’s Machinations
Defenses • Cryptography • Encryption {m}kencrypt message m with key k • Time • Nonces M,N random numbers “used only once” • “Handshake” function • (e.g. decrement) f(m) relates to “strong encryption”: assumption that {x}k and {g(x)}k (for any function g) are not easily computed from each other
The Needham-Schroeder Protocol AS: A, B, N A B S
The Needham-Schroeder Protocol AS: A, B, N SA: {N, B, K, {K,A}KB}KA A B S
The Needham-Schroeder Protocol AS: A, B, N SA: {N, B, K, {K,A}KB}KA AB: {K,A}KB A B S
The Needham-Schroeder Protocol AS: A, B, N SA: {N, B, K, {K,A}KB}KA AB: {K,A}KB BA: {M}K A B S
The Needham-Schroeder Protocol AS: A, B, N SA: {N, B, K, {K,A}KB}KA AB: {K,A}KB BA: {M}K AB: {f(M)}K A B S
The Needham-Schroeder Protocol AS: A, B, N SA: {N, B, K, {K,A}KB}KA AB: {K,A}KB BA: {M}K AB: {f(M)}K AB: {P}K once protocol completes, BA: {Q}K A and B can send messages over secure channel A B S
But . . . AS: A, B, N SA: {N, B, K, {K,A}KB}KA AB: {K,A}KB BA: {M}K AB: {f(M)}K What if Mallory • learns {K,A}KB • learns K ?
Mallory Attacks! AS: A, B, N SA: {N, B, K, {K,A}KB}KA AB: {K,A}KB BA: {M}K AB: {f(M)}K (M)B: {K,A}KB
Mallory Attacks! AS: A, B, N SA: {N, B, K, {K,A}KB}KA AB: {K,A}KB BA: {M}K AB: {f(M)}K (M)B: {K,A}KB B(M): {M}K
Mallory Attacks! AS: A, B, N SA: {N, B, K, {K,A}KB}KA AB: {K,A}KB BA: {M}K AB: {f(M)}K (M)B: {K,A}KB B(M): {M}K (M)B: {f(M)}K
Mallory Attacks! AS: A, B, N SA: {N, B, K, {K,A}KB}KA AB: {K,A}KB BA: {M}K AB: {f(M)}K (M)B: {K,A}KB B(M): {M}K (M)B: {f(M)}K . . . and now Bob thinks Mallory is Alice
Design Revisited • Such mistakes are common even today • Protocols disarmingly simple • Error may not be discovered (by the “good guys”) for a long time
Implementation • Perfect protocol design is worse than useless if implemented incorrectly • Myriad opportunities for that: • “Back doors” [ improper software engineering ] • Buffer overflow [ unsafe programming ] • Hobbled random number generators • Improper use of crypto algorithms • [ under-rated ] • [ fundamentals ]
Administration • (Where to start?) • Enforcing security policy, e.g. • Password policy • Rogue modems • Internal threats • Intrusion prevention and detection • Information Gathering • Information Analysis • Response
Information Gathering • ‘Normal’ Profiles • ‘Abnormal’ Signatures • Inbound Traffic Analysis • Audit Trail • On-the-Fly • Daily operational analysis (‘intuitive’)
Information Analysis • What is an intrusion? • Traffic Analysis Indicators • Analysis Techniques
Traffic Analysis Indicators • Repetition • Vulnerability Exploits • Mysterious Behavior / Problems • Unexpected/Inconsistent Activity
Analysis Techniques • Pattern-matching & Signatures • Dynamic Association • Statistical Profiling • Audit Reduction
Response • False Positives • Traceback & Anonymity • Offensive Action & Traps
Intrusion Detection Revisited • Technical component: • Hardware to scan traffic • Software to process traffic • Crafty component: • Clear understanding of normal traffic • Finger on the pulse • Sensitive viscera • Constant re-education
Users • Increasingly must be network admins at home • Personal privacy/security • Often must trust integrity of arcane infrastructure • Major target of “social engineering” • Critical thinking • Analytic problem-solving
Conclusions… • Improving security is fundamentally not technical • Must provide access to resources and knowledge • Must not let technology of security undermine permissive structure of Internet