Juniper Networks Inc v. Packet Intelligence LLC

1. Case Identification

• Case #: IPR2020-00339
• Patent #: 6,954,789
• Filed: February 4, 2020
• Petitioner(s): Juniper Networks, Inc. & Palo Alto Networks, Inc.
• Patent Owner(s): Packet Intelligence LLC
• Challenged Claims: 1-2, 13-17, 19-20, 42, 44, and 48-49

2. Patent Overview

• Title: Method and Apparatus for Identifying Flows in a Computer Network
• Brief Description: The ’789 patent discloses a system for monitoring and classifying network data packets into "conversational flows." The invention purports to distinguish these from simple "connection flows" by identifying and linking multiple, separate connection flows that are part of a single, higher-level application activity.

3. Grounds for Unpatentability

Ground 1: Obviousness over Riddle in view of Ferdinand - Claims 1-2, 13-17, 19-20, and 42 are obvious over Riddle in view of Ferdinand.

• Prior Art Relied Upon: Riddle (Patent 6,412,000) and Ferdinand (WO 92/19054).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner argued that Riddle, a patent on classifying network traffic, discloses the core limitations of the challenged claims. Riddle teaches a traffic classifier that identifies and groups multiple network connections associated with a single application activity into what it calls a "service aggregate" (e.g., for FTP or PointCast traffic), which Petitioner contended meets the ’789 patent’s definition of a "conversational flow." Riddle also disclosed parsing packets, looking up flow information, and creating new flow entries when a new conversational flow is encountered. Ferdinand, which relates to monitoring communication networks, was cited for its disclosure of conventional and well-understood components of packet monitors, such as using a database (a STATS database) to store flow-entry information and using input buffers to temporarily store packets to prevent data loss.
  • Motivation to Combine: A Person of Ordinary Skill in the Art (POSITA) would combine Ferdinand’s conventional database and buffering teachings with Riddle’s classification system. The motivation was to improve Riddle’s system by adding robust, well-known data storage and management capabilities. Using a database as taught by Ferdinand would increase functionality by allowing for efficient searching, analyzing, and modifying of flow entries. Similarly, incorporating a buffer would improve performance and reliability by preventing packet drops, a common objective in network monitoring.
  • Expectation of Success: A POSITA would have had a high expectation of success, as the combination involved implementing known, standard components (databases, buffers) into a network monitoring system to perform their conventional functions in a predictable manner.

Ground 2: Obviousness over Riddle in view of Ferdinand and Hasani - Claims 44 and 48-49 are obvious over Riddle in view of Ferdinand and Hasani.

• Prior Art Relied Upon: Riddle (Patent 6,412,000), Ferdinand (WO 92/19054), and Hasani (Patent 5,805,808).

• Core Argument for this Ground:

  • Prior Art Mapping: This ground built upon Ground 1 and specifically addressed the limitations of claims 44 and 48-49, which required a "database of parsing/extraction operations" that includes information to determine protocol-dependent extraction operations. Petitioner asserted that Riddle’s "knowledge base" (relational database 306) already met this limitation by storing heuristics used to classify traffic. Hasani was added to further teach a packet parser that explicitly stores protocol-dependent parsing and extraction operations in a database to identify information from specific packet formats, such as IEEE 802.2 MAC packets.
  • Motivation to Combine: A POSITA would be motivated to incorporate Hasani's teachings into the Riddle/Ferdinand system to make the access and maintenance of protocol-dependent parsing operations more efficient. By storing such rules in a database as taught by Hasani, the system's ability to classify traffic based on specific, well-known packet standards would be enhanced, which was a known design goal for improving packet classifiers.
  • Expectation of Success: The combination was argued to be a predictable integration of known techniques. Using a database to store protocol-specific parsing rules was a known method for creating a more efficient and extensible packet classification system.

• Additional Grounds: Petitioner asserted additional obviousness challenges based on combinations including Yu (Patent 6,625,150) and RFC1945. Yu was cited for its teachings on flexible policy-based flow classification that links multiple streams into a single flow based on application data, further supporting the concept of a "conversational flow." RFC1945 (the specification for HTTP/1.0) was cited to show that using HTTP header fields like "Referer" to link related but separate connection flows was a well-known technique for identifying conversational flows.

4. Key Claim Construction Positions

• "conversational flow": Petitioner argued this term should be construed more narrowly than in prior proceedings, proposing it means "the sequence of packets that are exchanged in any direction as a result of specific software program activity, where such packets form multiple connection flows that are linked based on that activity." Petitioner supported this by citing the patent owner's statements in prior IPR and district court litigation, where the patent owner used examples like a Skype call (with separate audio, video, and control flows) to define the concept. This construction was central to Petitioner's argument that prior art concepts like Riddle's "service aggregates" meet the claim limitation.
• "new flow" / "existing flow": Petitioner contended these terms should be construed as "new conversational flow" and "existing conversational flow" based on the logical operation of the claims and prosecution history. This construction clarifies that the patent is concerned with identifying and tracking entire application-level conversations, not just individual network connections.

5. Arguments Regarding Discretionary Denial

• Petitioner argued against discretionary denial under 35 U.S.C. §325(d) and §314(a). They contended that the primary combination of Riddle and Ferdinand had never been considered by the USPTO during prosecution or in any prior IPR. They further distinguished their petition from prior IPRs filed against the ’789 patent. A prior petition by Sandvine was denied because it relied on a different primary reference (Engel) and a different theory for teaching "conversational flows." Another petition by Nokia was terminated by the parties before an institution decision was issued. Therefore, Petitioner argued the Board had never addressed the merits of the grounds presented in this petition.

6. Relief Requested

• Petitioner requested institution of an inter partes review and cancellation of claims 1-2, 13-17, 19-20, 42, 44, and 48-49 of Patent 6,954,789 as unpatentable.